Varieties of Understanding: New Perspectives from Philosophy, Psychology, and Theology 0190860979, 9780190860974

Citation preview

Varieties of Understanding

Varieties of Understanding New Perspectives from Philosophy, Psychology, and Theology Edited by

STEPHEN R. GRIMM

1

3 Oxford University Press is a department of the University of Oxford. It furthers the University’s objective of excellence in research, scholarship, and education by publishing worldwide. Oxford is a registered trade mark of Oxford University Press in the UK and certain other countries. Published in the United States of America by Oxford University Press 198 Madison Avenue, New York, NY 10016, United States of America. © Oxford University Press 2019 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, by license, or under terms agreed with the appropriate reproduction rights organization. Inquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above. You must not circulate this work in any other form and you must impose this same condition on any acquirer. CIP data is on file at the Library of Congress ISBN 978–​0–​19–​086097–​4 1 3 5 7 9 8 6 4 2 Printed by Integrated Books International, United States of America

Contents List of Contributors 1. Varieties of Understanding Stephen R. Grimm

vii 1

I .   P H I L O S O P H Y O F U N D E R S TA N D I N G 2. Perspectives and Frames in Pursuit of Ultimate Understanding Elisabeth Camp

17

3. The Epistemologies of the Humanities and the Sciences Richard Foley

47

4. On Literary Understanding Jennifer Gosetti-​Ferencei

67

5. Recasting the “Scientism” Debate Anthony Gottlieb

93

6. Firsthand Knowledge and Understanding Ernest Sosa

109

7. Toward a Theory of Understanding Linda Zagzebski

123

I I .   P SYC HO L O G Y O F U N D E R S TA N D I N G 8. Technology as Teacher: How Children Learn from Social Robots Kimberly A. Brink and Henry M. Wellman

139

9. Understanding Others to Learn and Help Others Learn:  Inferences, Evaluation, and Communication in Early Childhood Hyowon Gweon

167

10. How Do Partial Understandings Work? Frank Keil

191

11. Mechanistic versus Functional Understanding Tania Lombrozo and Daniel Wilkenfeld

209

vi Contents 12. Are Humans Intuitive Philosophers? Steven Sloman, Jeffrey C. Zemla, David Lagnado, Christos Bechlivanidis, and Babak Hemmatian

231

I I I .   T H E O L O G Y O F U N D E R S TA N D I N G 13. Religious Understanding and Cultured Practices Terrence W. Tilley

253

Index

275

Contributors Christos Bechlivanidis, University College London Kimberly A. Brink, University of Michigan Elisabeth Camp, Rutgers University Richard Foley, New York University Jennifer Gosetti-​Ferencei, Johns Hopkins University  Anthony Gottlieb, Oxford University Stephen R. Grimm, Fordham University Hyowon Gweon, Stanford University Babak Hemmatian, Brown University Frank Keil, Yale University David Lagnado, University College London Tania Lombrozo, Princeton University Steven Sloman, Brown University Ernest Sosa, Rutgers University Terrence W. Tilley, Fordham University Henry M. Wellman, University of Michigan Daniel Wilkenfeld, University of Pittsburgh Linda Zagzebski, University of Oklahoma Jeffrey C. Zemla, Brown University

1 Varieties of Understanding Stephen R. Grimm

Imagine that you’re trying to understand why your neighbor is acting in a certain way—​why he’s being less talkative than usual, perhaps. Does understanding your neighbor require something significantly different, from an epistemic point of view, than understanding an event in the natural world, such as a change in the tides or an eclipse? More generally, does understanding people differ in important ways from understanding physical events or phenomena? According to a long historical tradition, the answer to both of these questions is yes. Understanding comes in different varieties, it is said, and understanding people has a different epistemic profile than understanding the natural world—​it calls on different cognitive resources, for instance, and brings to bear distinctive normative considerations. Thus philosophers such as Wilhelm Dilthey and R. G. Collingwood have argued that in order to understand people we need to appreciate, or in some way sympathetically reconstruct, the reasons that led a person to act in a certain way.1 By comparison, when it comes to understanding natural events, like earthquakes or eclipses, no appreciation of reasons or act of sympathetic reconstruction is needed—​mainly because there are no reasons on the scene to even be appreciated, and no perspectives to be sympathetically pieced together. Instead, to understand these events, we simply need to identify the natural laws that gave rise to them, or perhaps to identify their place within the intricate causal networks that make up the world. For convenience, let us call the tradition associated with Dilthey and Collingwood the “humanistic tradition.”2 This tradition says that 1 See, for example, Dilthey (1996) and Collingwood (1946). 2 It has also been called the “verstehen tradition”; see, e.g., Martin and McIntyre (1994) and Stueber (2012). Verstehen is the German word for understanding, and the idea here (roughly) is that while things like eclipses might be explained, only human beings, properly speaking, can be understood. See Grimm (2016) for a sympathetic but critical discussion of this distinction.

2  Stephen R. Grimm understanding comes in different varieties, and in particular that there is a gulf between understanding human-​related phenomena and natural phenomena.3 Opposed to this view is what we might call the naturalistic or scientistic tradition. It says, roughly, that understanding comes in only one variety.4 While it might be true that we try to appreciate the reasons that are leading our neighbor to be less talkative, or that we might try—​in some sense—​to imaginatively enter into his or her shoes, the naturalistic tradition claims that these sorts of efforts are at best stepping stones along the way. When all goes well, they will simply lead to the same impartial framework of laws, causes, and mechanisms that we find in the natural sciences. In order to evaluate the dispute between the humanistic tradition and the naturalistic tradition, it will help first to clarify two different questions, which I will take up in the following section: 1. What would it take for understanding to count as having different varieties? 2. Even if we allow that understanding comes in different kinds or varieties, what is it that make them all cases of understanding—​i.e., different varieties of the same thing? In the course of addressing these questions, I will touch on several essays in the collection, although I will not try to offer detailed summaries. This introduction is rather meant to set the stage for later chapters in the volume, without attempting to be comprehensive.

1. Different Varieties Beginning with the first question, about what it would take for understanding to count as having different varieties, we can start with the simple observation that our understanding of the world is wide ranging. On what we might 3 Note that I am not suggesting with this distinction that human beings are non-​natural. It would therefore perhaps be clearer to say that the distinction is between human beings—​or more advanced agents generally—​and the rest of the natural world. But the traditional debate has generally been conducted in the former terms (human events vs. natural phenomenon), and I will continue that tradition here. 4 See, for example, Peter Atkins, who claims that “Science is the only path to understanding” (Atkins 2006:  124). Of course, there might be lively disagreement about what that one variety is (whether basically nomic, or causal, etc.).

Varieties of Understanding  3 call the larger scale, there are those who understand how fermentation takes place, why certain people are vulnerable to type 2 diabetes, why sea levels are rising, and why graduate students are increasingly reporting cases of depression and anxiety. On the smaller, everyday scale, we also understand countless little events. Walking downstairs in the morning, I understand why my dog is looking at me expectantly (because he wants his breakfast), why my coffee spilled (because the table wobbled), and why my morning paper was delayed (because of the ice storm). Noting the large number of things we understand, we can therefore ask: When would differences between these instances of understanding count as deep enough to mark distinct kinds or varieties of understanding?5 Consider by comparison how this question might look if asked of knowl­ edge, rather than understanding. What would it take for differences among instances of knowledge to count as deep enough to belong to different kinds or varieties? Suppose I visually scan my kitchen table before dinner, taking in its various forks, plates, and cups. I will thereby have acquired several new instances of knowledge—​that there is a fork here, a plate here, and cup over there—​but it nonetheless seems mistaken that I will have thus acquired several new varieties of knowledge: “fork knowledge,” for instance, as opposed to “plate knowledge” or “cup knowledge.” It seems more apt instead to say I have acquired several new instances of the same variety of knowledge: roughly, perceptual knowledge. This would, moreover, accord with traditional categories within epistemology, where distinct bodies of literature are dedicated to topics like perceptual knowledge, as opposed to memorial knowledge, as opposed to a priori knowledge, and so on.6 These all presumably count as varieties of knowledge because they share some common property. But the way in which this common property is attained or acquired seems different enough that these instances merit their own special category. For the sake of argument, suppose these all count as instances of knowl­ edge because they share the common property of reliably getting things right.7 “Reliably getting things right” would then be the genus encompassing

5 For more on the notion of a “deep difference” with respect to epistemic categories, see Grimm (2017). 6 These distinctions are then usually replicated in epistemology textbooks or anthologies. See, for instance, the distinct chapters in Audi (2011), Bernecker and Dretske (2000), and O’Brien (2006). 7 For more in defense of this claim, see Grimm and Ahlstrom-​Vij (2013). According to Kremer (2016), this also seems to have been Gilbert Ryle’s view.

4  Stephen R. Grimm all instances of knowledge, and there would be different varieties or species of “reliably getting things right” if the various ways of getting things right were distinctive enough. For instance, using the faculty of memory to get things right seems distinctive enough that it gives rise to its own variety of knowledge—​memorial knowledge. And similarly for perception, a priori knowledge, and so on. By contrast, using perception to reliably distinguish plates from cups from forks presumably does not count as distinctive enough. In short, it seems that when the cognitive resources or powers of the mind in virtue of which we get things right are different enough, we have distinct varieties of knowledge.8 We might also venture a little more, for arguably the reason why distinctive cognitive resources or powers come into play is because the objects known in these areas are themselves importantly different, and hence require different cognitive resources or powers to track them. In the case of past objects or events, a power of the mind like memory; in the case of present nearby objects, a power of the mind like perception; in the case of necessary truths, a power of the mind like rational insight, and so on. Thinking this way would accord with the ancient idea, going back at least to Plato, that deep differences in the object of cognition require (or even give rise to) deep differences in cognitive processes themselves. Drawing the discussion back to understanding, we can therefore suggest that different varieties of understanding arise, in a way parallel to different varieties of knowledge, if the differences in objects of understanding are pronounced enough that they require different cognitive resources or powers of the mind in order to be grasped. Or if you would prefer to skip the appeal to the objects grasped, we might simply say that varieties of understanding arise when they draw on importantly or markedly different cognitive resources or powers of the mind. The humanistic tradition holds that this is not just an abstract possibility. Instead, it claims that the cognitive resources or powers of the mind required to understand human beings are different enough that they give rise to a distinctive sort of understanding, with a distinctive epistemic profile. To be compelling, this hypothesis would need to be supported by results from psychology that point to deep differences in cognitive resources when it comes

8 For more on the “powers of the mind” idea, see Grimm (2019). It is worth noting, too, that many powers of the mind blend in with each other—​thus perception will often be informed by memory, and so on. But I hope the general point claimed here is clear enough.

Varieties of Understanding  5 to understanding human beings.9 I  do not believe any of the psychology papers in this volume make such a case, and I am not sure how any of the authors individually would take sides in the debate between the humanistic tradition and the naturalistic tradition, but they nevertheless help to shed light on these questions by addressing “nearby” or related questions of great importance. I will now turn to these.

2.  Psychology I begin with the framework articulated in Lombrozo and Wilkenfeld’s chapter. They distinguish a mechanistic stance (or mode of construal) from a functional stance (or mode of construal). According to the former, we try to understand things in terms of their component parts, or proximal causes. For instance, we try to understand why the alarm clock is beeping in terms of its internal circuity—​the various bits and pieces of the alarm clock that make it capable of emitting such a noise. According to the latter, we appeal to purposes or goals. For instance, we try to understand why the alarm clock is beeping by looking for the purpose of the beeps—​most likely, the purpose of waking up the clock’s owner. Despite these different stances or modes of construal, Lombrozo and Wilkenfeld nonetheless argue that the objects of understanding in both of these cases are the same—​namely, dependence relations.10 When we take a mechanistic stance, we try to identify how the phenomenon depends upon its proximal physical causes—​what it is about the alarm clock’s internal circuitry, for example, that gives rise to the beeping. By contrast, when we take a functional stance, we try to identify how the beeping depends upon the psychological goals or purposes of the agents on the scene—​for instance, the goals that led the clock’s owner to set the alarm at this time. 9 Work on mirror neurons might be thought to suggest such a case, but it is not clear whether these are essential to understanding other people, or just one pathway among others. For discussion see Stueber (2018). 10 For more on dependence relations, see Kim (1994), Grimm (2016b), etc. On this view, dependence relations would be the genus within which things like causal relations might be a species (e.g., mereological relations, in which properties of the whole depend on properties of the parts, might be another type of dependence relation). Note the similarity between this approach and Linda Zagzebski’s view in this volume. According to Zagzebski, understanding always takes structure as its object. If one thinks structures are necessarily things that encode dependence relations, then the two views are basically identical. From her examples, it is not clear that Zagzebski would subscribe to this view of structure (many maps, for example, do not obviously encode dependence relations—​and maps are one of her key cases). But if she did, there would be great overlap here.

6  Stephen R. Grimm This gives us a picture on which the alarm clock’s beeping depends on two sorts of things: on the one hand, the internal circuitry, etc.; on the other hand, the goals and purposes of the person who set the clock. Depending on our interests—​or as Lombrozo and Wilkenfeld put it, our inferential goals—​we might then focus on one set of dependence relations rather than another. Alternatively, we might take one stance toward the clock rather than another. According to Lombrozo and Wilkenfeld, because there are two sorts of dependence relations on which we might focus, this picture supports what they call the “weak differentiation thesis,” according to which these stances differ in terms of the understanding they afford because they differ in terms of their objects. As they note, however, the weak differentiation thesis counts as “weak” because the objects at issue—​the dependence relations at issue—​seem to be quite similar. Or, perhaps better, even though at one level the objects look rather different—​physical dependence relations on the one hand vs. psychological dependence relations on the other—​it is not obvious that these differences amount to deep differences. After all, on the face of it the objects of perception seem rather different—​forks are not plates, after all—​but these differences do not seem deep enough to give rise to different varieties of knowledge. In short, from an epistemological point of view, the difference in dependence relations picked out by these different modes of construal might not be significant enough to require distinctive cognitive resources, and hence might not count as different enough to give rise to distinctive varieties of understanding. They then take up the question of whether there might be deeper differences at play in these cases—​in other words, whether there might be evidence for what they call the “strong differentiation thesis,” according to which the epistemic profile of understanding in these cases is deeply different, because of the sorts of cognitive resources involved, or the distinctive objects at issue, or both. Although their arguments on this score are exploratory, they suggest that the normative-​evaluative aspect of the intentional stance might yield a different variety of understanding. For instance, when I learn that someone’s behavior is due to certain beliefs and desires, I might then go on to criticize the behavior as a flawed expression of those beliefs and desires. Or I might turn to criticize the beliefs and desires themselves—​as inapt, or unfounded, or poorly supported. Proper understanding in these cases might therefore require a sense of what the person should have done, or could have done better. But none of this seems relevant to understanding dependencies in the natural world. I do not critique an eclipse as poorly supported or unjustified.

Varieties of Understanding  7 So, tentatively, the normative dimension of understanding people might require distinctive cognitive resources, of the sort suggested by the humanistic tradition. In addition to this question about possible varieties of understanding, Lombrozo and Wilkenfeld raise a number of other questions that are partly taken up by other psychologists in the volume: 3. Do people prefer one of these modes of construal to the other, perhaps because they take it to be richer or deeper or more informative than the other? 4. Are we better at acquiring some of these forms of understanding than others? On both questions, the evidence from Sloman et al. and from Keil paints an interesting picture. Although they do not explicitly contrast mechanistic and functional stances along the lines of Lombrozo and Wilkenfeld, they nonetheless find a preference for mechanism-​focused explanations in a variety of areas. Sloman et al., for instance, find that people will prefer more complex explanations to simpler ones, so long as more complex explanation offers information about mechanisms. This is surprising, Sloman et al. suggest, because according to philosophical lore, going back at least to Ockham, simpler explanations are to be preferred to more complex explanations. But things are not so straightforward, Sloman et al. provocatively argue, and our various preferences for things like simplicity, mechanistic information, coherence, etc. plausibly interact to form “a horribly complex system that doesn’t lend itself to simple explanations.” Keil too explores the various ways in which people—​especially young children—​show a preference for mechanistic information. If asked why a car is “the right one to buy,” for instance, children will be unimpressed by the answer that it has a very pretty color, or that it makes a cool sound. Instead, if you say things like, the car “is the best because the motor uses high-​quality wire that makes it run very smoothly. When you drive, you can speed up and stop quickly,” then children will be more satisfied. This example does not turn on the contrast between the mechanistic stance and the intentional stance, as described by Lombrozo and Wilkenfeld. But it does suggest a strong desire for mechanistic information—​perhaps above all other types of information. Keil also discusses a surprising corollary of this finding. Despite our longing for mechanistic information, human beings are remarkably bad

8  Stephen R. Grimm at storing and retaining this information, or even properly identifying mechanisms in the first place. With respect to the former point, Keil points to studies showing that our knowledge of mechanisms—​say, the knowledge we might acquire in a biology class, of how the Krebs cycle works—​declines “massively and monotonically over time.” With respect to the latter, a striking example involving bicycles (see Keil, this volume; and Lawson [2006]) suggests that we are not even good at distinguishing possible from impossible mechanisms when we see them, indicating that our ability to recognize mechanisms is weaker than we might suspect. Gweon, in her essay, and Brink and Wellman, in theirs, also note that a large part of our understanding of the world is secondhand: we learn about the world from others. But clearly some people are better sources of information than others—​more exactly, they are more informative than others, and hence are better teachers. How then do we discriminate the good sources from the bad? According to Gweon, the better sources of information are the ones that give us just the right amount of information: not too much, not too little. So, if a toy has three distinct functions, and a teacher only reveals one to a child, the child will judge this person to be a poor teacher. Or again, if a toy had a hundred buttons that all perform the same function, and the teacher laboriously presses all hundred buttons, the child will judge this person to be a poor teacher. (“I get it,” would be the child’s natural thought, after a dozen or so buttons with identical functions had been pressed.) Gweon also persuasively shows that we evaluate the quality of teaching based on assumptions about the teacher’s goals, beliefs, and desires. So, for instance, we naturally take it that a teacher’s goal will be to show us all of the important functions of a given device: not to leave any important ones out. At the same time, they do not need to “over-​show”—​that is, provide redundant information, once it is clear that it has already sunk in. We therefore naturally seem to reason along the following lines: given the teacher’s goals + her beliefs about the subject she wants to communicate + her beliefs about the state of the students’ information, she is /​is not achieving these goals well. In short, as Gweon puts it, our understanding of others thus helps us to discriminate the good teachers from the bad. Hence, our understanding of others helps us to better understand the world around us, because it allows us to identify the good teachers around us—​the informative ones who tell us what it is important to know, and who leave out the inconsequential or unimportant stuff.

Varieties of Understanding  9 Brink and Wellman further explore the way in which we evaluate robots as potential teachers, and hence as potential guides for understanding the world. What they have found is that young children (up to around the age of seven) prefer information from robots with more human-​like qualities, especially those which are able to respond contingently to children’s questions; indeed, not only do they prefer this information, but they seem to retain it better, thus learning more from such robots. As we age, however, the preference for receiving information from human-​like robots fades; past the age of seven or eight, we do not seem to learn more, or retain information better, from human-​like robots than from non-​humanoid robots. Perhaps most interestingly, there is also a point at which robots can appear “too” human, especially to older children and adults, so that they seem creepy or unnerving. This point has been dubbed the “uncanny valley,” and Brink and Wellman surmise that robots reaching this point will be less effective as teachers, for adults, than robots without humanoid traits.

3. Firsthand Understanding In relation to the articles by Gweon and by Brink and Wellman, I have been discussing our “secondhand” understanding of the world, as when we trust teachers to tell us how various things are related, or depend upon one another. According to Ernest Sosa’s essay, however, in some domains it is important to understand or know in a “firsthand” way, without deferring to others. Key among these are the domains of morality, aesthetics, and philosophy itself. With respect to morality, for example, to take a position on some issue simply based on the say-​so of others is to “neglect one’s rational nature” and to “yield[] the core of [our] humanity” (Sosa, this volume). But what would firsthand understanding in the case of morality, for instance, actually amount to? In normal empirical cases firsthand understanding or knowledge usually amounts to taking a look—​to seeing things for oneself. What might the corollary be in the case of morality? For Sosa, this seems to require appreciating the grounds or reasons for a certain judgment. Suppose you’re trying to make up your mind about capital punishment, and a trusted advisor tells you that it is morally wrong. You then believe accordingly. According to Sosa, there is something lacking in your attitude until you can appreciate the reasons or grounds for the belief. You would be yielding your autonomy if you were simply to defer.

10  Stephen R. Grimm More concretely, suppose it turns out that capital punishment is wrong because it compromises human dignity. A firsthand understanding of the claim would therefore involve seeing or grasping the way in which this compromise occurs, or perhaps what it is about capital punishment that triggers the compromise. It would also involve, presumably, seeing or grasping in an immediate, non-​inferential way that it is wrong to compromise human dignity; for that matter, it would seem to require seeing or grasping in an immediate way that human beings have a special dignity to begin with. And none of this seeing or grasping can be acquired, it seems, by sheer deference. Seeing or grasping is the sort of thing that one needs to do for oneself, if it is to happen at all. What is less clear about Sosa’s claim is why deferring—​at least, to begin with—​on matters related to morality or aesthetics should be improper, or at odds with our rational nature. Let us grant along with Sosa (and Aristotle, and others) that we naturally seek “the why” of things—​presumably, the reason or grounds of things. Why can we not defer or trust, and then seek the why? Perhaps the answer is that this sort of deference or trust might lead us down the wrong path—​so we might be seeking whys (reasons, grounds) where there are no whys to be found. This is a fair concern, but what exactly is the epistemic alternative—​a life lived without deference or trust, or with deep suspicion of others? As Sosa notes, this in fact has been the attitude of many philosophers, including Descartes and Locke, for whom a secondhand belief was “often dismissed as second rate or even worthless” (Sosa, this volume). But Sosa rightly distances himself from that tradition, recognizing that we float in a vast river of epistemic dependence. An epistemic life without dependence would therefore be neither realistic or even, it seems, preferable. If we have an epistemic duty or obligation or ideal in the neighborhood, therefore, it would seem to be the following: defer, by all means, and even on moral and aesthetic and normative questions. But do not simply defer, or rest content with deference. Instead, defer in a way that seeks further understanding. This therefore suggests a variation on the traditional formula, associated with figures such as St. Anselm: fides quarens intellectum, or faith seeking understanding. Instead, an updated version of the formula might be scientia quarens intellectum, or (roughly) knowledge seeking understanding. For if one thinks, as Sosa does, that one can obtain knowledge from testimony, then trusting or having faith in another, even with respect to normative questions, can indeed give rise to knowledge. But there is plausibly a duty

Varieties of Understanding  11 here not to be content with the knowledge: one needs, in addition, to seek the deeper reasons that underlie the claim known via testimony.11 In other words, one needs to seek understanding, and not just knowledge, at least when it comes to moral matters.

4.  Understanding in Literature and Religion Finally, a very brief word about two of the other possible varieties of understanding that receive special treatment in this volume—​literary understanding and religious understanding—​to give a sense of the range of topics at issue. Jennifer Gosetti-​Ferencei claims that understanding a literary work cannot be reduced to simply understanding or comprehending “its represented content”—​i.e., the characters, places, and events that constitute the work. In addition, she claims, it involves regarding these contents from the perspective from which these contents are presented. We are invited to take up the perspective of the narrator, or of particular protagonists, or indeed the “perspective of the work as a whole,” which may exceed that of any particular person within the story. But what does it mean to take up these sorts of perspectives? According to Gosetti-​Ferencei, in a way that echoes some of Elisabeth Camp’s acute observations in her essay (this volume) on perspective-​taking, this requires a “responsive undertaking,” and in particular an ability to feel the “imagined experience generated in and through the work.” For instance, in Virginia Woolf ’s novel Mrs. Dalloway, we read the following: For having lived in Westminster—​how many years now? over twenty,—​one feels even in the midst of the traffic, or walking at night, Clarissa was positive, a particular hush, or solemnity; an indescribable pause; a suspense (but that might be her heart, affected, they said, by influenza) before Big Ben strikes. There! Out it boomed. First a warning, musical; then the hour, irrevocable. The leaden circles dissolved in the air. (Woolf 2000: 4) 11 Alternatively, as Laura Callahan has suggested in conversation: another interesting possibility is that communities might disperse these epistemic obligations in different ways. Perhaps everyone is not obliged to seek understanding of moral matters, even if in some sense they could. Perhaps the epistemic good of the community as a whole would be best served if some sought this understanding, while others relied on pure testimony.

12  Stephen R. Grimm Gosetti-​ Ferencei argues that understanding this passage fully requires “imagined listening”—​one needs to feel, at some level, the tension before the clock strikes, and then the experience of the slow dissipation of the sound. If you cannot imaginatively respond in the appropriate way, there seems to be a level of understanding that is missing. Certain types of moral and aesthetic understanding are also made especially vivid through fiction, according to Gosetti-​Ferencei. With respect to religious understanding, Terrence Tilley begins by arguing that all of our language is at root metaphorical, or that “literal language is parasitic on figurative language” (Tilley, this volume). The idea here seems to be that many if not most uses of language, and thought more generally, point beyond themselves to connect to other phenomena. Our understanding thus increases when we recognize these connections and appreciate these relationships. According to Tilley, religious language in particular alleges to point beyond itself, to something transcendent—​God, or karma, or samsara, for example. Growth in religious understanding therefore involves, for Tilley, appreciating these connections between the mundane and the transcendent, and Tilley argues that this appreciation can often best be achieved by participating in religious practices—​for instance, in the established rituals and ways of life of a religious community. Tilley notes that this way of thinking about religious understanding—​ as deeply embedded in practices—​generates a problem, because many practices are based on illusion, so being enculturated into a practice will not lead to genuine understanding, but rather to things like brainwashing, error, or arrogance. In response, Tilley claims that while these risks are real, not all practices should be tarred with the same brush. Many practices, including many religious practices, are self-​correcting, in the sense that they are open to critique from sources (scientific, historical, etc.) outside the practice, and are responsive to truth-​related considerations. Engaging in these practices therefore involves risk, but they are not an entirely naive risks to take, according to Tilley, especially because certain practices open up the possibility of types of understanding that would otherwise be closed off.

References Atkins, Peter. 2006. “Atheism and Science.” In The Oxford Handbook of Religion and Science. Eds. Philip Clayton and Zachery Simpson. New York: Oxford University Press.

Varieties of Understanding  13 Audi, Robert. 2011. Epistemology:  A Contemporary Introduction to the Theory of Knowledge. 3rd ed. New York: Routledge. Bernecker, Sven, and Dretske, Fred. 2000. Knowledge:  Readings in Contemporary Epistemology. New York: Oxford University Press. Collingwood, R. G. 1946. The Idea of History. Oxford: Clarendon Press. Dilthey, Wilhelm. 1996 [1867–​68]. Hermeneutics and the Study of History. In Selected Works, vol. 4. Ed. and intro. by Rudolf Makkreel and Frithjof Rodi. Princeton, NJ: Princeton University Press. Grimm, Stephen. 2016a. “How Understanding People Differs from Understanding the Natural World.” Philosophical Issues (Noûs supplement) 26: 209–​225. Grimm, Stephen. 2016b. “Understanding and Transparency.” In Explaining Understanding: New Essays in Epistemology and Philosophy of Science. Eds. Stephen R. Grimm, Christoph Baumberger, and Sabine Ammon. New York: Routledge. Grimm, Stephen. 2017. “Why Study History? On Its Epistemic Benefits and Its Relation to the Sciences.” Philosophy 92: 399–​420. Grimm, Stephen. 2019. “Understanding as an Intellectual Virtue.” In The Routledge Companion to Virtue Epistemology. Ed. Heather Battaly. New York: Routledge. Grimm, Stephen, and Kristoffer Ahlstrom-​Vij. 2013. “Getting It Right.” Philosophical Studies 166: 329–​347. Kim, Jaegwon. 1994. “Explanatory Knowledge and Metaphysical Dependence.” Philosophical Issues 5: 51–​69. Kremer, Michael. 2016. “A Capacity to Get Things Right: Gilbert Ryle on Knowledge.” European Journal of Philosophy 25: 25–​46. Lawson, R. 2006. “The Science of Cycology:  Failures to Understand How Everyday Objects Work.” Memory & Cognition 34: 1667–​1675. Martin, Michael, and Lee McIntyre. 1994. “Introduction.” In Readings in the Philosophy of Social Science. Eds. Michael Martin and Lee McIntyre. Cambridge, MA: MIT Press. O’Brien, Dan. 2006. An Introduction to the Theory of Knowledge. Malden, MA: Polity Press. Stueber, Karsten. 2012. “Understanding versus Explanation? How to Think about the Distinction between the Natural and Human Sciences.” Inquiry 55: 17–​32. Stueber, Karsten. 2018. “Empathy.” The Stanford Encyclopedia of Philosophy. Ed. Edward N. Zalta. https://​plato.stanford.edu/​archives/​spr2018/​entries/​empathy/​. Woolf, Virginia. 2000 [1925]. Mrs. Dalloway. London: Vintage.

I

PHILO SOPH Y OF U N DE R STA NDI NG

2 Perspectives and Frames in Pursuit of Ultimate Understanding Elisabeth Camp

Our ordinary talk is rife with “framing devices”: expressions that function, not just to communicate factual information, but to suggest an intuitive way of thinking about their subjects. Metaphors are the most obvious instance: when a speaker calls their new home a “dump,” their old job a “jail,” their new colleague a “bulldozer,” a “gorilla,” or a “quarterback,” their classmate “a tailwagging lapdog of privilege” (Moran 1989: 90) or a prospective paramour “the Taj Mahal” (Bezuidenhout 2001: 161), part of their intended effect is to evoke a host of unstated properties which fit together, along with images and feelings, into a coherent interpretive whole (Camp 2006). But metaphors are far from unique in doing this. Slurs like “kike” (Camp 2013) and thick terms like “foodie,” “bourgeois,” “wanton,” and “valor” also promulgate perspectives, as do truthful “telling details” like that Barack Obama’s middle name is “Hussein” (Camp 2008), and fictional or apocryphal “just-​so stories,” such as that Donald Trump was denied admission to Harvard as an undergraduate (Camp 2009). These rhetorical tropes differ in important ways, in their conventional status, assertoric force, and interpretive mechanisms and effects. But they all perform a recognizably common, and crucial, communicative function. The use of such framing devices isn’t confined to casual conversations. Frames are also ubiquitous in political, pedagogical, and scientific discourse. Lakoff (2004) illustrates the pervasiveness of “strong father” and “two-​parent egalitarian” parenting models in driving conservative and liberal policies and propaganda. Coll et  al. (2012) document widespread use of models and analogies, like enzymes as lock and key or photosynthesis as making a cake, in science education. And philosophers and historians of science have long noted the ubiquity of models, fictions, metaphors, and analogies in scientific investigation and explanation (Hesse 1993; Giere 1998; Elgin 2006;

18  Elisabeth Camp Godfrey-​Smith 2006; Camp 2019b). Finally, in addition to their use in communication, framing devices can also play an important role in individual cognition, as slogans, precepts, and models that guide inquiry, explanation, and memory. At the same time, framing devices are double-​edged swords (Glynn 2008). In their communicative role, they can mold our minds into a shared pattern—​even when we as hearers would rather resist (Camp 2017a). They can foster confusion, when speakers and hearers unknowingly focus on different unmentioned properties, images, and feelings. And they can provide cover for cowardly insinuation and innuendo (Camp 2018a). The risks in individual cognition are potentially greater, insofar as the intuitive power of a frame can blind us both to known features that don’t fit easily within the frame, and also to “unknown unknowns” we have not yet encountered. Thus, perhaps Locke is right to disavow such “eloquent inventions” as “perfect cheats” that “insinuate wrong ideas, move the passions, and thereby mislead the judgment” (1689/​1989: 34). I think this is the wrong conclusion to draw. The metaphor of double-​ edged swords is indeed apt; but that is because frames are tools for thought, which, like any tool, can be used well or badly, and for good or for ill—​not because they fall outside the realm of rationality altogether. In §1, I describe how framing devices express open-​ended perspectives, which produce structured intuitive characterizations of particular subjects. In §2, I argue that frames can make effective, distinctive epistemic contributions in the course of inquiry. And in §3 and §4, I argue that the cognitive structures that frames produce can contribute to, and even constitute, epistemic achievements in their own right, even in highly idealized circumstances at the nominal end of inquiry. Throughout, I will focus especially on scientific understanding, because it serves as a paradigm case of rational inquiry, from which frames and perspectives are most likely to be excluded. The case for other modes of inquiry and understanding, such as psychology and history, is comparatively easier to establish; I will not address the important differences among them here.

1.  Characterizations, Perspectives, and Frames As I will be using the terms, frames are representational vehicles with the function of expressing perspectives. Perspectives in turn are open-​ended

Perspectives and Frames  19 dispositions to interpret, and specifically to produce intuitive structures of thought about, or characterizations of, particular subjects. These are relatively familiar terms, used to describe familiar phenomena. I will articulate them in the specific way that I have found most theoretically fruitful, one that both overlaps and departs from those offered by other theorists describing the same broad range of phenomena. Some of these other taxonomies exclude those phenomena from the realm of rationality and the achievement of epistemic value by definition, a classification I am concerned here to challenge. But even leaving the question of epistemic status to the side, I believe these alternative taxonomies fail to do full descriptive justice to the way that the actual phenomena function, either in ordinary life or in more systematized contexts like historical and scientific inquiry.

1.1.  Characterizations Much of our everyday cognition involves complex, intuitive ways of thinking about specific subjects, which I call characterizations. Stereotypes are the most familiar case; but where stereotypes are culturally ubiquitous, characterizations can be restricted to a sub-​discipline, a particular conversation, or an isolated individual. In many cases, characterizations are close to what philosophers call “conceptions”: a set of intuitive beliefs about an individual or a kind, which need not be extension-​determining, or constitutive of conceptual competence, or even reflectively endorsed; but which are easily evoked and provide the standard “mental setting” (Woodfield 1991: 551) for thinking about a subject. Characterizations thus differ markedly from concepts, at least as philosophers typically think of them.1 Prescinding from as much detail as possible, characterizations differ from concepts in at least three key respects: their content, structure, and stability (Camp 2015). First, concepts are (or at least entail, or involve) abilities to re-​identify certain individuals and kinds: they determine what is being thought about, and in particular that the same individual or kind is represented across those variations in the circumstances of activation and evaluation that are irrelevant to the concept’s 1 I believe that philosophers have identified a significant cognitive role for concepts as they construe them, but that the phenomena that psychologists study under the rubric of “concepts” are also important, and that much of the apparent disagreement can be reconciled by interpreting psychologists as often primarily concerned with characterizations instead (Camp 2015).

20  Elisabeth Camp applicability. As such, concepts abstract away from many details in our experience and knowledge about the subject, including especially perceptual details and affective responses. By contrast, characterizations are informationally, experientially, and affectively rich, integrating as much data as possible into an intuitive whole. So, for instance, my intuitive characterizations of Barack Obama and Donald Trump include, inter alia, facts about their backgrounds, families, psychological traits, and past actions; what each of them look like, including how they walk, talk, and gesture; and my emotional and evaluative responses to these details and to them more generally. Second, where concepts have (at most) minimal internal structure, of deductive and material inference, characterizations connect the many constituent features that they attribute to their subjects into a complex multidimensional structure, reflecting the different ways in which a feature can matter in an agent’s characterization of a given subject. Two dimensions of “mattering” are especially crucial. First, some of the features ascribed to a subject are more prominent than others: more initially noticeable and quicker to recall. I take prominence to be equivalent to what Tversky (1977) calls “salience,” which he in turn analyzes as a function of two factors, each of which is contextually relative in a different way. On the one hand, a feature is diagnostic to the extent that it is useful for classifying objects in a given context, as the elliptical shape of a snake’s pupils might be useful for determining whether it is poisonous. On the other hand, a feature is intense to the extent that it has a high signal-​to-​noise ratio. What counts as background “noise,” and hence in turn how intense a given feature is treated as being, varies widely, both in how locally sensitive and in how cognitively mediated it is. So, for instance, the perceptual intensity of a light’s brightness relative to the ambient lighting is fixed by a background that is highly local and directly physical; while the perceived intensity of a pigment’s tonal saturation in a painting is likely to be measured not just in contrast to the other colors in that particular picture, but also against the agent’s assumptions about typical saturation levels in other paintings within that genre or from other historical periods. Where prominence selects which features matter, centrality determines how they matter, by connecting features into explanatory networks, such that more central features are more richly connected to other features. Causal connections are a paradigm basis for explanation. And a decent, albeit rough, measure of imputed causal centrality is mutability: how much the agent’s overall thinking about the subject would need to alter if they no longer

Perspectives and Frames  21 attributed a given feature f to the subject (Murphy and Medin 1985; Thagard 1989; Sloman, Love, and Ahn 1998). However, connections may be logical or metaphysical without being causal. We also often intuitively connect features on grounds that are emotional or ethical: in terms of the “tick-​tock” of what would be satisfying, or more generally which features “fit” together (DeSousa 1987; Kermode 2000). (Indeed, we sometimes impute causes in order to justify such intuitions.) Prominence and centrality are structurally distinct ways for a feature to matter intuitively. So, for instance, Barack Obama’s ears or Donald Trump’s hair may be highly prominent in my thinking just because they are unusually protruding and swooping, respectively, even if I do not take those features to be at all explanatory. But the two dimensions are not entirely disconnected. In particular, when a feature f’s intensity departs markedly from a contextually determined baseline—​especially if it is highly intense, but also sometimes if it is unusually low—​then this fact calls out for explanation. Sometimes we dismiss such departures from baseline as anomalies; but often we seek or posit an explanation of the departure which is grounded in the subject’s central features. And even when no plausible justificatory connection is forthcoming, we may still intuitively feel that the feature “fits” together with more central features, at least aesthetically. Caricatures often seem apt, and funny, because they play on such aesthetic connections:  for instance, by linking Obama’s protruding ears to his Spock-​like nerdiness, or Trump’s swooping hair to his grandiosity. The third key difference between concepts and characterizations, besides the richness and structure of their contents, is their stability. A core function of concepts is to underwrite connections between distinct thoughts, both synchronically via inference, and diachronically via recall and revision of beliefs, desires, and other attitudes. The possibility of re-​deploying the same belief on multiple occasions, let alone of changing my mind about its truth, requires being able to represent the same content on multiple occasions. Likewise, integrating multiple pieces of information about the same entity requires being able to represent that entity as the same. Thus, concepts, qua cognitive entities that enable agents to represent, collect, and use information about various entities and kinds, must be fairly cross-​contextually stable (Camp 2015). We do regularly use concepts in this logical, stabilizing way in actual life—​ for instance, when we individually plan a route for performing a series of errands or a budget for buying a house, or when we collectively plan and

22  Elisabeth Camp build a bridge or a political system. However, much of our intuitive thinking is also contextually malleable, as the vast experimental literatures on framing and affective and cognitive priming demonstrate (e.g., Kahneman, Slovic, and Tversky 1982; Levin, Schneider, and Gaeth 1998; Musch and Kluer 2003; Bartels 2008). In particular, ethical and emotional responses, as well as judgments about causal structure and statistical probability, all both influence and are influenced by our currently operative characterizations. And although contextual priming is sometimes driven by relatively innate impulses, as “dual systems” theorists maintain, it often activates and modulates repertoires of interpretation and evaluation that are rich, sophisticated, and highly enculturated. Because many of these context-​specific effects on what we notice and how we explain and respond to it are implicit, they often go unnoticed until we are prompted to reflect on our inconsistencies. The sort of contextual malleability displayed by characterizations underlines another key difference from concepts. Many philosophers take it to be characteristic, even constitutive of conceptual thought that it is subject to critical reflection and revision (McDowell 1994; Korsgaard 2009). In this sense, it is part of applying concepts, at least for normal adult humans, that one can hold them “at arm’s length,” in order to scrutinize their applicability and the transitions they underwrite. By contrast, characterizations are by their nature intuitive patterns of thought, which guide what an agent just does naturally notice, what explanatory connections they do tend to form, and how they immediately respond in cognition and action. As such, reflectively endorsing the appropriateness of a given characterization, as specified at arm’s length, is neither necessary nor sufficient for actually characterizing a subject in the relevant way. There is an important sense in which concepts are also intuitive, insofar as concept-​constituting inferential transitions are “primitively compelling” to a competent agent who entertains the relevant thoughts (Peacocke 1992). However, in the conceptual case, this “compulsion” arises because the agent takes the transition to be correct, or at least right in a way relevant to sorting and classification (Ginsborg 2011). By contrast, characterizations can remain intuitively compelling despite our rejecting their aptness. Often, of course, our conceptual and characterizing “compulsions” operate in tandem. And in particular cases, there may be no clear, independent fact of the matter about whether the intuitive attribution of a feature, or association between features, “really” belongs to the agent’s concept or to their characterization. However, the two classes of intuitive attribution and association can, and demonstrably

Perspectives and Frames  23 do, come apart, in ways that are answerable to different cognitive pressures and evaluative norms; and our theoretical taxonomy should recognize this (Camp 2015). In particular, it (all-​ too) frequently happens that one intuitively characterizes a subject in a way one wishes one didn’t, as with in stereotype threat or unwelcome slurs or insinuations (Camp 2013). At the same time, we also sometimes willingly try on characterizations that differ from those we take to be genuinely apt. This is perhaps most obvious in reading fiction (Camp 2017b). But at least sometimes in the course of political and personal debate, we attempt, and sometimes succeed, in “getting” how someone else construes the subject, “from the inside.” Still, in such cases, merely wanting or intending to try on a characterization doesn’t suffice to “get” it: even if we do reflectively endorse its appropriateness, the characterization may fail to function at an intuitive level. Reflective endorsement without intuitive applicability is perhaps especially palpable on first encounter with idealized scientific models, such as Feynman diagrams. Characterizing, as the complex intuitive construal of a subject, is thus in gen­ eral partly but not entirely under voluntary control. We do have antecedent, default intuitive characterizations of many subjects; but these are modulated for us, often without our noticing it and sometimes actively against our will, by external and internal contextual factors. We can also modulate our characterizations intentionally, by directing our attention toward some particular features and away from others, and by actively entertaining certain concepts and assumptions. When we are attempting to cultivate a given characterization, any one bit of information about which features play what role in it—​that this is a highly notable feature, which in turn explains that one—​ may help us to “get” it at an intuitive level. But much as with the classic cases of gestalt perception, ultimately the relevant structures of attention, explanation, and response just do govern our intuitive thinking—​or don’t. And just as with gestalt perception, while a new characterization sometimes “clicks” into place, accompanied by a phenomenology of sudden illumination, it may also dawn gradually. The analogy with gestalt perception brings out a final feature of characterizing: the sense in which it is a holistic affair. Altering the prominence or centrality of a single feature can induce pervasive, complex alterations to the structural relations among other elements, by “tipping” them into new clusters of explanatory relations and new weightings of prominence. The effects of applying a new frame may also extend beyond

24  Elisabeth Camp structural realignment, producing a paradigm shift that alters the significance of the basic features themselves. Thus, a spatio-​temporally equivalent gesture can seem threatening or merely awkward, depending on one’s characterization of the person performing it. This is especially obvious in cases where we know little about the person, and so where race and other general demographic features can have a strong effect (Devine 1989; Duncan 1976). But it extends to cases that encompass rich bodies of information: for instance, in our interpretations of an intimate partner’s specific behaviors while we are embedded within, and then after leaving, the relationship.

1.2.  Perspectives Thus far, I’ve been focusing on individual characterizations and spelling out the ways in which they differ from concepts, especially in terms of their contextual malleability. However, characterizations are rarely isolated, on a given occasion or across time. Rather, agents have default propensities to form certain types of characterizations of multiple subjects. A perspective is an open-​ended disposition to characterize: to encounter, interpret, and respond to some parts of the world in certain ways. We of course have our own individual interpretive dispositions, which may be more or less stable and more or less encompassing. But we also have an intuitive, sometimes quite nuanced, sense of other people’s perspectives. When we know someone well, we can predict how they will construe and respond to new subjects, and how they will assimilate new information about old subjects into their existing frameworks for them. Similarly, part of the power of psychologically rich fictions lies in the way that we as readers come to anticipate, not just how events in the fictional world will unfold and be described, but how the narrator or authorial figure would interpret the actual world if they were to encounter it (Camp 2017b). This open-​ended quality distinguishes perspectives from propositional attitudes even more strongly than characterizations are distinguished from concepts. A characterization does have a content, albeit a complex, often messy, and mostly unarticulated one: it attributes a set of features to a particular subject, and embeds those features in a multidimensional structure of prominence and centrality. At the same time, as we just saw, it is neither necessary nor sufficient for characterizing a given subject in the relevant

Perspectives and Frames  25 way that one explicitly endorse or even entertain this complex content. Rather, what matters is that one’s own intuitive patterns of attention, explanation, and response actually implement this structure, at least for the moment. In this sense, characterizations are not (just) attitudes toward propositions: they are, instead, intuitive cognitive implementations of complex structures.2 Perspectives inherit this intuitive implementational aspect from the characterizations that they generate. But they are also non-​propositional in a stronger sense. In principle, given sufficient reflection and effort, characterizations’ contents are fully specifiable in propositional terms—​ so long as those propositions can include demonstrative reference to images, experiences, emotions, and evaluations, as well as representations of higher-​order structural relations of relative prominence and centrality. It’s just that having a characterization involves more than having an attitude of entertainment or endorsement toward those propositions. By contrast, perspectives lack contents:  having a perspective need not require endorsing, or even intuitively attributing or connecting, any particular features of a subject. Perspectives are in their essence tools for thinking, not thoughts per se. Some perspectives may well be essentially connected with certain propositions: for instance, the perspective of evangelical Protestantism may essentially involve commitment to the divinity of Jesus and to the Bible being the Word of God. However, many perspectives, such as political liberalism or conservatism, lack any essential doctrines. Moreover, adherents of a perspective needn’t endorse the doctrines that might appear to be central or even essential to it. For instance, many self-​professed evangelical Protestants reject the claim that the only route to eternal life is belief in Jesus’s divinity, even though church leaders almost universally take this claim to be required for being a good evangelical Christian.3 But they still identify as evangelical 2 Perhaps at least some beliefs (and make-​beliefs) are not fully propositional attitudes, given this specification of “propositional attitudes.” I am inclined toward a view of beliefs as standing commitments to treat the world as being a certain way, where some such commitments may involve complex characterizations. High-​level interpretive beliefs, such as those expressed by “Men are more violent than women,” “Equal work merits equal pay,” or “Precision is more important than insight,” are the clearest candidates here. I leave an analysis of belief, and the relations between belief and concepts and characterizations, for another occasion. 3 A 2008 Pew study found that 51% of evangelical Christians believe that belief in religions other than Christianity can lead to eternal life, and that 26% allow that atheists can achieve salvation; while according to a 2011 Pew survey, 95% of evangelical Christian leaders say that these beliefs are incompatible with being a good evangelical Christian.

26  Elisabeth Camp Christians, insofar as they notice, care about, explain, and respond to many aspects of the world in many of the same ways as their more orthodox brethren. Thus, where characterizations are intuitive construals of a given subject at a given time, perspectives are open-​ended intuitive dispositions to interpret. They are open-​ended in at least two senses: first, they provide a way of updating a given characterization over time, as new information and experiences come in; and second, they generate characterizations of multiple, perhaps indefinitely many, specific subjects. In both dimensions, although which particular information, images, and feelings are attributed to a subject varies, there is a stable perspective just in case there is substantive stability in the sorts of features the agent tends to notice, the sorts of explanatory connections they tend to draw, and the sorts of predictions and emotional and evaluative responses they tend to have. While the effects of variation in context on particular characterizations are typically messy, they are also fairly clearly demonstrable. Given a stable set of information, it is possible to measure, in fairly specific and controllable ways, how priming for distinct emotions, explanations, and purposes, and/​or changing the order, vividness, or terms of description, alters an agent’s characterization of that information. In this way, in the case of characterizations we can at least begin to get a grip on the sense in which different agents, or the same agent at different times, have distinct characterizations of a common set of facts. By contrast, because of their essentially open-​ended nature, it is much harder to identify and individuate sameness and difference in perspective, and to say when an agent is operating with a certain perspective. Within a given context, the wider the swath of an agent’s characterizational dispositions that fit with a certain perspective, the more plausible it is to say that they have temporarily adopted it. Likewise, the wider the range of contexts in which an agent manifests those dispositions, the more plausible it is to say that this is their own perspective. However, there is not always a principled answer to the question of whether an agent has only temporarily tried on an alternative perspective, or whether their temporary responses really are “their own.” Rather than seeking to identify absolute sameness and difference in perspectives, it is often more accurate to speak only of relative overlap and stability.

Perspectives and Frames  27

1.3.  Frames Characterizations, qua intuitive, holistic, contextually malleable ways of thinking that are often imagistically and affectively laden, are typically messy, inchoate, and at least partly tacit. Perspectives, qua open-​ended dispositions to characterize, are even more so. At the same time, perspectives affect our thinking in deep and pervasive ways, including by influencing our explicit judgments and beliefs. For these reasons, we often want to impose more coherence and stability on our intuitive thinking. We also often want to communicate our characterizations and perspectives to others, either just to help them understand our state of mind, or else to coordinate on a common assumptions. To accomplish both goals, of internal regulation and external expression and coordination, we frequently employ interpretive frames. As I will use the term, frames are representational vehicles—​most obviously linguistic vehicles like slogans, but also non-​linguistic vehicles like diagrams and caricaturing cartoons—​under an intended interpretation, where that interpretation itself functions as an open-​ended principle for organizing and regulating one’s intuitive overall intuitive thinking about one or more subjects. Frames crystalize perspectives into compact, explicit form. Not all perspectives can be adequately expressed by frames: some are just too multivalent to be captured by a single slogan or image, or no one has yet happened or needed to do so. But when and to the extent that a frame does express a perspective, it unifies it into a more cohesive whole and underwrites wider contextual stability. Further, as publicly accessible entities, frames can function as vehicles for communicating perspectives, by evoking a body of shared experiences and feelings. A frame can express a perspective, and apply to its subject, in various ways. The relevant interpretive principle may be explicit in, or follow fairly immediately from, the informational content of the frame itself, as in “He’s just not that into you”; or it may evoke an unstated characterization, as with a tautological saying like “Boys will be boys.” The association between vehicle and perspective may be conventional, as with slurs (Camp 2019a); or it may be pragmatic, as with metaphor (Camp 2017a). The frame’s informational content may be true, as with “telling details”; or false, as with apocryphal and just-​so stories (Camp 2008). And the imaginative transformation required to apply a literally false frame may operate at an interpretive level, as with metaphor, or at a metaphysical one, as with fiction (Camp 2009).

28  Elisabeth Camp All of these differences make a difference to the type of cognitive effort involved in comprehending them and to the sort of cognitive effect they produce—​including in the context of inquiry, especially scientific inquiry (Camp 2019b). For current purposes, however, these differences are ancillary. What matters, first, is that frames in general express perspectives, which function as open-​ended intuitive principles for attending to, explaining, and responding to a range of subjects as an agent encounters new information and experiences. And second, while frames, perspectives, and characterizations all utilize and influence concepts and judgments, they always go well beyond, and often depart significantly from, the deliverances of conceptual thought.

2.  Frames as Instruments for Inquiry As a descriptive matter, the ubiquity of intuitive, associative cognitive structures in everyday thought is well-​attested empirically, albeit not using exactly the terms and categories I  have employed. However, the normative, and specifically epistemic, status of characterizations, perspectives, and frames is highly contested. Theorists who distinguish intuitive associative thinking from more narrowly conceptual thought, as I  have done, often treat the two classes of phenomena as belonging to distinct cognitive “systems,” with the evolutionarily ancient System 1 functioning as a fast and dirty shortcut for the more robustly logical cognition undertaken by System 2 (e.g., Gendler 2008; Evans and Frankish 2009; Kahneman 2011). If we understand characterizations and perspectives as at best intuitive proxies for, and as at worst antagonistic to rational, conceptual thought, it is difficult to see how they could make any genuine epistemic contribution. Either they are mere noise, which we should filter out in the pursuit of genuine knowledge (Gendler 2007, 2011); or else they are intuitive heuristics whose deliverances must be independently tested, and ultimately eliminated as inquiry proceeds. By contrast, I  take characterizations to play a functional role distinct from that of concepts, but one that is still in the service of rational engagement with the world. While concepts provide stable classificatory principles which enable agents to think about the same entity or kind across a range of contexts, in order to connect and update thoughts in systematic ways, characterizations facilitate smooth interaction within contexts by guiding attention and response, and by synthesizing rich bodies of information and experience into intuitive wholes (Camp 2015). If we accept that concepts and

Perspectives and Frames  29 characterizations do have distinct functional roles, then it becomes an open question what epistemic status to assign to characterizations, perspectives, or frames. In this section, I argue that frames can play a distinctive, even essential role in the course of inquiry, by guiding attention and suggesting hypotheses and explanations in an open-​ended, flexible way that a fixed set of propositions by itself cannot do. In §3 and §4, I argue that although frames drop away as inquiry proceeds, perspectives and characterizations contribute essentially to ultimate understanding even at the (putative) end of inquiry. There is a relatively straightforward, uncontroversial sense in which frames can contribute to understanding. As communicative tools for expressing complex intuitive patterns of thought, frames help us to comprehend one another’s perspectives, in ways that narrowly informational statements can’t. They thus enable interlocutors in a given conversation to coordinate efficiently on a rich set of unstated assumptions, expectations, and evaluations, which they can then utilize, independently of the frame, in the course of investigating what is true or what to do. Even when interlocutors don’t end up endorsing a common set of assumptions, they may still achieve a kind of respectful engagement with the other’s point of view, or at least an ability to predict what the other will think, say, and do, which they could not have without trying on that perspective “from the inside.” In addition to assisting in intersubjective understanding, frames can also help us to understand ourselves. They may do this by providing cross-​ contextually stable handles that encapsulate the interpretive principles we care most about. Frames can confront us forcefully with new perspectives to resonate to—​or to reject—​and thereby help reveal to us what our antecedent, tacit interpretive principles had been. Finally, frames may provide us with aspirational touchstones which we can employ in order to actualize more fully at an intuitive level interpretive principles that we reflectively endorse in the abstract. While these are all valuable roles for frames to play, which contribute to understanding our own and others’ minds in a rich, intuitive, ongoing way, what’s primarily at issue in investigating the epistemic status of frames, perspectives, and characterizations is whether they can make a distinctive contribution in understanding the world as it is, independently of us. A first step in arguing that they do facilitate understanding centers on the role of perspectives as tools for thought. A  perspective provides an agent with an ability to navigate efficiently among a rich body of existing information and experiences. When I have a perspective on a domain, I “know

30  Elisabeth Camp my way about” that domain (Wittgenstein 1953: §123) in an intuitive way that facilitates retention, recall, and selection of relevant information from a larger body. This robust effectiveness of frames in fostering the efficient manipulation of information explains the strong emphasis on their utility in education, most notably in science (Coll et al. 2005). Moreover, by providing a principle for incorporating new information and experiences, and for predicting further information, a perspective enables me to “know how to go on” in updating and generating new characterizations of the focal subject, and often of other, related ones as well (Wittgenstein 1953: §179). By guiding the intuitive characterizations that an agent forms of a given domain, a perspective also thereby influences their outright judgments about it. It does this most directly by influencing their higher-​order structural judgments about base-​level information: by guiding what information they take to warrant explanation and what they dismiss as irrelevant or unreliable; which explanations they find compelling; and what predictions they make about counterfactual contingencies and future events. And in turn, these structural judgments can provide the justificatory ground for—​and sometimes themselves constitute—​statistical, explanatory, and evaluative judgments about that domain. At a deeper level, a perspective also affects an agent’s base-​level beliefs themselves, by determining which concepts they deploy in forming those base-​level judgments. A perspective presupposes a taxonomy of categories, which function to support a profile of theoretical and practical ends. It thereby individuates occurrences of features as relevantly “the same again,” and imposes boundaries between kinds of objects in virtue of their possessing those features. It also thereby assigns greater prominence to features that are diagnostic relative to that taxonomy. And it assigns explanatory structures and degrees of centrality that answer to the operative profile of practical and theoretical purposes. By guiding the classification of information in the formation of base-​ level and higher-​order judgments about the world, frames and perspectives function as genuinely epistemic tools. As such, they can be assessed, and criticized, in terms of their functional utility. At a minimal level, we can make sense of an agent mischaracterizing a given subject relative to their own interpretive standards, if their assignments of prominence and centrality come apart from the assignments that are warranted by their operative taxonomy, practical and explanatory purposes, and the actual distribution of features in the world. More robustly, we can assess frames and perspectives themselves

Perspectives and Frames  31 for epistemic aptness. Internally, frames that are more coherent enable agents to better navigate among their existing information. Externally, frames that are more apt presuppose taxonomies that better serve the agent’s operative practical and explanatory purposes and that individuate kinds of features and objects using categories that more closely track actual statistical distributions. Apter frames thereby provide agents with a more robust and reliable epistemic grip on the world: one that enables them to assimilate a wider range of new information smoothly into their existing interpretive structure, to make more accurate predictions, and to make interventions that serve their practical goals. Beyond helping them to navigate among existing information and assimilate information that an agent happens to encounter, the perspective expressed by a frame can also contribute to understanding in a more proactive way, by guiding their search for information. It can do this directly, by leading them to seek confirming evidence for a putative explanatory connection. But it can also generate hypotheses for investigation in more indirect ways. In particular, one reason that analogical and metaphorical frames are so cognitively fruitful is that they set up “analogical equations” which transfer structural “kernels” of related features from one domain to another. More specifically, with analogical frames, the fact that the target domain is already known to exhibit many features within the kernel suggests that it may also point toward as-​yet unknown features which would “complete” that system (Gentner and Jeziorski 1993). But metaphor and analogy are not unique in producing “system completion” (Camp 2019b). For instance, true frames, such as “telling details,” can guide investigation by suggesting the application of a cluster of features that are associated within the characterization of that detail, and by generating explanations that focus on those features. Likewise, “just-​so stories” and other literally false frames can generate hypotheses by suggesting that if the subject a, were, contrary to actual fact, to possess feature X, then it would also possess features x1, x2, and x3; since a is already known to actually instantiate these features, and since those features are associated by the frame with features y1, y2, and y3, then perhaps a possesses y1, y2, and y3 as well. In the simplest cases of feature introduction by “system completion,” the candidate features can themselves be straightforwardly specified in literal terms. When this is the case, the frame is dispensable, in the sense that it merely prompts a hypothesis that could have been articulated independently

32  Elisabeth Camp all along. The contribution of such frames to investigation is merely heuristic, if practically substantial. In a second, slightly more complex class of cases, there may be no antecedent term for the relevant feature in the language; but the analogical equation set up by the frame may still be sufficiently constrained that it effectively provides a reference-​fixing description which permits the introduction of a new literal term into the language (Camp 2006). Here, the frame does play an essential initial role, by establishing a relevant mapping between the two structures of the framing and target characterizations. However, the equation itself, once specified, is still precise enough to determine a substantive identificatory condition for the relevant base-​level features or explanatory structures, independent of the frame. Thus, we might want to say that for this second class, the frame’s contribution can be restricted to generating a “Ramsified” proposition which can then be investigated in a (more or less) straightforward way. Beyond this, though, there remain a third class of cases of feature introduction by system completion, for which the respects of similarity between frame and subject are not sufficiently well articulated to be cashed out by Ramsification. As Richard Boyd (1979/​ 1993) puts it, metaphors (and, I would add, other frames) of this class can play a “programmatic research-​ orienting” role in investigation, not despite, but because of their intuitive, open-​ended, indeterminate status. For instance, Evelyn Fox Keller (1995) argues that the “code” model of genetic action, and the “organism” model of machine systems, were reciprocally effective frames in early 20th century scientific inquiry precisely because the operative theories of the framing domains were in each case so inchoate that they left open a wide range of candidate mappings onto the other domain. For a frame to play a fruitful “programmatic research-​orienting” role, it must already be substantive enough to significantly constrain the search space of hypotheses. It needs to establish a profile of cognitive and pragmatic interests and goals, point toward a restricted region within a broader domain as worth probing, impose some differentiating structure on that region, and suggest potential causal and other explanatory connections to features within that region that are better understood. At the same time, though, such a frame can be genuinely generative to the extent that the possible solutions to those analogical equations are not yet themselves fully articulated in a way that leaves just the verification or falsification of precisely defined propositions to be carried out. Instead, the perspective’s intuitive,

Perspectives and Frames  33 holistic nature plays an essential role in parsing the domain of possibilities at a quite basic level, by guiding investigators first in what they intuitively notice as a feature at all, or treat as the same feature again, and second in how they hypothesize that disparate features might be connected.

3.  Perspectives at the “End” of Inquiry In §1 I  argued that although characterizations are often messy, inchoate, contextually malleable, and idiosyncratic, they are not essentially noisy and context-​bound. Frames can crystallize perspectives into stable interpretive principles which can be deployed by an individual agent across multiple cognitive contexts, and shared by multiple individuals within and across contexts. In §2 I cataloged a range of ways in which frames, and the perspectives they express, can function as epistemically valuable tools for thought:  by facilitating effective navigation among existing information; by guiding how an agent updates their overall thinking, including beliefs and other reflective attitudes, in light of new information; and by influencing how they seek out new information. In each of these respects, frames and perspectives can provide agents with a more robust and reliable epistemic grip on the world. And in all of these respects, perspectives accomplish this, not just despite but because they differ from straightforward conceptual thought in the ways identified in §1—​because they are open-​ended dispositions to produce intuitive, holistic characterizations. Negatively, these points undercut the objection that according epistemic value to perspectives and characterizations is inappropriate because they are mere associationist noise, too idiosyncratic and fickle to be amenable to systematic analysis, let alone to contribute to intersubjectively shared cognitive projects. Positively, they establish that perspectives can play a distinctive role in support of distinctively epistemic ends. Indeed, perspectives are arguably unavoidable for the kinds of epistemic beings we are. As cognitively sophisticated but limited agents embedded in complex environments, in flexible pursuit of multiple long-​term goals, we need broad, ongoing cognitive principles in order to select relevant details from a teeming manifold of stimuli within a given situation, to synthesize those details into more or less coherent wholes, and to mobilize for immediate longer-​term action. As I emphasized in §1, these functions support, but are distinct from, the core function of

34  Elisabeth Camp conceptual thought, which is to track, combine, and deploy information in systematic ways across contexts. And while frames are not cognitively unavoidable in the same way as perspectives, they do help to bring perspectives and characterizations within the fold of rationality, by increasing their intrapersonal stability and breadth and their interpersonal accessibility. Thus, given that we are creatures for whom intuitive, holistic, experientially-​and affectively-​laden patterns of thought are both deeply natural and rationally functional, the best course of action would seem to be to determine how to use frames and perspectives for epistemic good, rather than how to eliminate them from serious inquiry altogether. However, granting that frames can express epistemically valuable, perhaps even indispensable, heuristics for guiding cognitively limited agents in the process of acquiring and managing information is compatible with insisting that their epistemic value is merely instrumental and temporary. In this section, I argue that perspectives can contribute to understanding even at the (nominal) end of inquiry, by implementing characterizations that accurately reflect the structure of the world. One prima facie source of support for the claim that frames and perspectives are only instrumentally epistemically valuable comes from the shifting role that they typically play in actual scientific investigation. Scientists frequently and happily employ frames, but—​as attested by the examples of the “code” model of genetic action and the “organism” model of machine systems above—​typically only at the early stages of investigation. Scientists ultimately aim to eliminate those frames, or to restrict their role to pedagogical contexts, where they are treated as mere gateways to a more nuanced, genuine understanding. The prevailing attitude is aptly encapsulated in the dictum from the mathematician George Pólya (1954): “And remember, do not neglect vague analogies. But if you wish them respectable, try to clarify them.” Metaphors in particular characteristically exhibit a trajectory or “career” of increasing literalization (Bowdle and Gentner 2005), in both ordinary and theoretical discourse. As we saw in §2.1, a metaphor like “machine systems are organisms” can play a productive “research-​orienting” role in part because it is inchoate and intuitive. But investigation typically consists in a series of attempts to precisify the frame, by articulating and systematizing tacit assumptions about both the framing and subject domains, by identifying putative matches between specific features in the frame and subject, and by probing whether the subject actually possesses anything in the ballpark of the proposed match.

Perspectives and Frames  35 Sometimes the end result is that the metaphor dies into the new life of literal meaning, as has plausibly happened with the model of the mind as a computer. Sometimes it remains as a merely pedagogical tool, as with the model of the atom as a solar system. Often, it is discarded as a potentially misleading first approximation, as has arguably happened with the metaphor of DNA as software. But whatever the ultimate status of the framing metaphor, as the research program it encapsulates is implemented and its interpretation becomes increasingly articulated, stable, and precise, scientists’ attention increasingly focuses on the myriad actual details about the target domain that it enables them to identify, rather than on the suggestive powers of the frame itself. Further, although this transformation from indispensible to ancillary is most obvious in the case of metaphorical frames, a similar shift occurs even with frames that are literally true. Most complex natural phenomena are unlikely to be as systematically unified as a single frame suggests. As theorists develop an increasingly firm grip on the details in their own right and come to recognize details that are left out of or obscured by the frame’s simplifying handle, we should expect catchy sloganeering to give way to a more nuanced, multidimensional perspective that cannot be crystallized by any single proposition. At a more fundamental theoretical level, at the “end” of inquiry, when all the evidence is in, there is by stipulation no need to generate hypotheses, assimilate information, or make predictions; and all explanations and other connections between disparate bits of information have been established. Thus, at that point the open-​ended cognitive function of perspectives in general—​whether or not they are encapsulated by frames—​is rendered otiose. All that remains are a host of particular, complex characterizations, each embedding a manifold of specific facts and explanatory connections. Further, many theorists consider perspectives to essentially involve a kind of epistemic limitation: a partial and selective view from a particular “somewhere,” in contrast to an omnipresent, omniscient panorama (e.g., Currie 2010). On this construal, perspectives are naturally contrasted with an ideal theory that is completely general in scope and encompassing in detail—​more specifically, with a theory that embeds successive reductions of higher-​level properties and kinds to lower-​level implementations. While metaphors and other frames might initially aid us in pursuit of such a “top-​down research strategy” (Pylyshyn 1993:  557), once that strategy had been fully implemented, we would ultimately be left with a single unified set of claims,

36  Elisabeth Camp each of which is true simpliciter and all of which are related by epistemically familiar, straightforward relations like entailment. Fully addressing the argument for dispensability just sketched would take more space than I have here. But I think that this model of ultimate theory radically undersells the depth and robustness of perspectives’ epistemic contribution, not just in the course of actual inquiry by cognitively limited agents, but even on a highly idealized construal of understanding as such. First, as I said in §2, a perspective presupposes a conceptual taxonomy plus a set of practical and explanatory ends: a commitment to which distinctions are worth making, and why. Even if we set aside practical interests as not relevant to “genuine” understanding and focus exclusively on explanatory interests, different conceptual taxonomies will more directly instantiate, or at least more smoothly integrate with, different explanatory interests, by imposing classifications that more closely track relevant statistical distributions and that more directly entail relevant explanatory connections. Given this, even if we take taxonomies to be narrowly conceptual resources as opposed to the broadly “non-​conceptual” phenomena of characterizations, perspectives, and frames, a perspective still affects which of those resources an agent should employ, given their operative explanatory purposes. The operative sense of “should” here is one of serving distinctively epistemic, and not just practical, ends. But as Carnap (1932) and many others have argued, the choice of a conceptual taxonomy cannot itself be fully articulated, let alone justified, within that taxonomy. Thus, at least in this sense, a perspective is more basic than the set of true propositions determined by any theory, even an “ultimate” one. Second, it is highly unlikely that the sort of systematic reduction that would determine a univocal conceptual taxonomy is possible, even in principle. Even if we grant some kind of ontological unitarianism, according to which all there really is are fundamental physical forces and (perhaps) particles, we will almost certainly need to embrace explanatory pluralism—​not just in virtue of variations in practical purposes and pragmatic constraints on explanatory tractability, but simply in order to capture the highly diverse patterns of structural contingencies that actually obtain among distributed clusters of the basic forces and particles. The world is massively complex, with individual features multiply connected to one another at radically distinct temporal and spatial scales (e.g., Cartwright 1999; Mitchell 2003). Different disciplines—​ scientific and humanistic—​ don’t just operate on different domains and scales, but legitimately prioritize different types and degrees

Perspectives and Frames  37 of generalization (Ismael 2018). And when these differences in explanatory focus and purpose bottom out in differing taxonomies, they cannot always be reconciled just by embedding a single, consistent feature within multiple explanatory networks, since what counts as a feature at all relative to one taxonomy may depend on a pattern of commitment and neutrality with respect to other lower-​level features within that same taxonomy, and where this pattern may conflict with the principles of individuation that are employed by other taxonomies. Thus, even if we do reach a point where we have “all the facts,” we will still need multiple, irreducibly distinct perspectives in order to articulate and explain them.

4.  The Ultimate Characterization Suppose, however, that we grant not just ontological, but also explanatory unitarianism—​either from a faith in the ultimate conciliation of explanatory purposes, or from a severe restriction on which explanatory purposes we treat as ultimately “legitimate.” Under that assumption, the ideal end of inquiry will indeed, by stipulation, produce one ultimate, internally coherent, all-​inclusive perspective. Because the open-​ended, inquiry-​guiding function of perspectives will have been exhausted at that point, this ultimate perspective will in effect collapse into a maximally coherent and inclusive characterization, subsuming many layers of complexly linked subsidiary characterizations of increasingly specific subjects. Thus, our final question is whether this highly idealized characterization itself makes a distinctive epistemic contribution, beyond knowledge of the truth of the constituent facts that it subsumes. More specifically, given the assumption that the ultimate characterization will include “all the facts,” in the sense of an exhaustive catalog of attributions of base-​level features to subjects, what we need to determine is, first, what becomes of the structures of prominence and centrality that a characterization imposes on those base-​ level facts; and, second, what epistemic value that structure itself might have. In normal cognitive circumstances, prominence is closely tied to the allocation of attention: assignments of prominence reflect assumptions about what is worth paying attention to, and function to guide attention to those features that matter. Specifically, in §1.1, I defined prominence as a function of intensity and diagnosticity, where intensity is a context-​sensitive measure of the signal-​to-​noise ratio of the subject a’s possessing the feature X that

38  Elisabeth Camp depends both on the external environment and on the agent’s assumptions about how common X is for subjects of that type; and where diagnosticity is the evidential relevance of a feature for classifying the subject relative to a presupposed taxonomy and a profile of practical and explanatory purposes. The idealized context we are currently envisioning diverges radically from these normal cognitive circumstances. Limitations on attention have been lifted; a unified taxonomy has been achieved; and contextual variations in environmental conditions, and in practical and explanatory purposes, have been eliminated. Fully implementing these abstractions would thus necessitate a wholesale transformation of the notion of prominence, which I cannot undertake here. However, as we also noted in §1.1, a key component of the ordinary function of prominence is to track departure from a statistical baseline, insofar as a feature warrants attention when its instantiation is surprising or distinctive. Thus, perhaps we can grant that assignments of the prominence of a given feature in application to a given subject within the ultimate characterization will reflect, and perhaps reduce to, assignments of the statistical distribution of that feature’s occurrence and of its correlation with other features in application to other subjects of the same and other types. What about centrality? As the embeddedness of a given feature within causal, logical, and other explanatory connections, it requires a less radical idealization than prominence. In normal circumstances, those connections depend on and are answerable to the agent’s operative practical and explanatory purposes. By hypothesis, at the nominal end of inquiry those purposes have been purified and reconciled, producing a single overarching explanatory structure that embeds every feature attributed to every subject within a vast, intricate network of logical and counterfactual connections operating at different temporal and spatial scales. While this structure will depart dramatically in shape and complexity from the structure of any ordinary characterization, this departure is more a matter of degree than of kind. It should not be controversial that statistical distributions and explanatory connections are epistemically important even at the ultimate end of inquiry. Rather, the question is whether characterizations make any epistemic contribution beyond simply specifying or expressing those assignments. In §1.1, I  granted that the content of a characterization can be specified propositionally in terms of higher-​order structural relations, but insisted that having a characterization goes beyond entertaining or even endorsing that set of propositions, and instead involves actually instantiating the relevant structure in one’s intuitive thinking, so that prominent features really do jump to

Perspectives and Frames  39 attention and central features really do connect in explanatory associations that one is actually disposed to draw. Thus, what is currently under dispute is whether the intuitive, holistic grasp afforded by characterizations makes any epistemic difference over and above the assignments they embody and which can be used to specify them. The skeptic insists that the endorsed truth of the higher-​order structural propositions articulating those assignments is all that matters, while I  want to argue that the intuitive grasp of those assignments is also crucial. Jonathan Kvanvig (2009: 99) presses a similar point about the value of understanding, as “the grasped relatedness of the items that constitute a body of information possessed by the individual in question,” thus: [I]‌t is not enough that the explanatory connections exist or that they could be discovered easily by the individual with only a little effort or reflection. Understanding involves an already-​possessed awareness of the explanatory and other connections involved in the subject matter in question, an already-​mastered grasp that involves or generates the illumination of a subject we resort to the language of intelligibility and sense-​making to convey.4

The challenge for a defender of understanding is to articulate what “having” a characterization or “grasping” a holistic set of relations amounts to, in a way that is operative even at the putative end of inquiry and that does not reduce to grasp of a set of propositions. One commonly invoked candidate appeals to the feeling of insight that we experience when a perceptual or cognitive gestalt clicks suddenly into place, or dawns gradually over a domain. However, while this feeling is satisfying in its own right (Hills 2016: 678), it is easily dismissed as a mere subjective, phenomenological state—​and worse, one that is prone to mislead, in particular by producing epistemic complacency precisely because it is cognitively satisfying (Trout 2002). A more plausible candidate appeals to the role of understanding in generating open-​ended explanatory abilities. In §2, I emphasized the open-​ ended role of perspectives in acquiring and assimilating information. At 4 Cf. Catherine Elgin (2006:  202):  “Science seeks, and often provides, a unified, integrated, evidence-​based understanding of a range of phenomena. A list, even an extensive list, of justified or reliably generated true beliefs about those phenomena would not constitute a scientific understanding of them. Veritism, in concentrating on truth, ignores a host of factors that are integral to science. These factors cannot be dismissed as just instrumentally or practically valuable. They are vital to the cognitive contributions that science makes.”

40  Elisabeth Camp the end of inquiry, those functions have been exhausted; but even then a significant practical difference remains between merely endorsing a set of higher-​order structural propositions and actually having a characterization. Even putting aside the role of characterizations in guiding attention and facilitating recall as of merely instrumental value, it is only when a characterization is implemented as a cognitive structure that an agent can perform the sorts of tasks we treat as markers of understanding: tasks such as explaining why a given subject has the particular features it does, or what further permutations would be produced by a specific change to it, or how clusters of features within one subject compare with those in another. That is, actually having the characterization, as an implemented, intuitive, holistic cognitive structure, is what provides the “cognitive control” that is characteristic of understanding (Hills 2016: 663). Further, this cognitive control is inherently epistemically relevant, and not just instrumentally useful, insofar as such higher-​order explanations often constitute justifications of lower-​level claims. Thus, the ability to construct such explanations constitutes a distinctively epistemic ability, and one that makes a significant contribution to an agent’s knowledge of the lower-​level claims (cf. de Regt 2009). In response, the skeptic of understanding will want to insist that the epistemic value here resides not in the cognitive control or ability itself, but rather in the higher-​order propositions that it generates—​for instance, the answers to a set of “why” questions. Kvanvig (2009: 101) objects to this reductionist move by pointing out that at least in some cases, there cannot be an exhaustive set of answers for why things happen as they do, because indeterministic systems by definition lack any such answers. A more ambitious and general response would accuse the skeptic of over-​intellectualization. On the one hand, in many domains we credit an agent with understanding the subject when they display a flexible ability to navigate among, draw connections between, and imagine appropriate counterfactual modifications to particular facts, even if they are unable to explicitly articulate those connections in themselves, let  alone form higher-​order explanations for why those connections and counterfactual modifications obtain. And on the other hand, if an agent cites a particular higher-​order structural proposition in response to a particular “why”-​question, without displaying any more gen­eral ability to explain why that proposition is true, then this will call into question our ascription of both understanding and knowledge-​why to them. At a more theoretical level, then, the skeptic faces a threat of regress, insofar as he locates in endorsement of higher-​order propositions, which themselves

Perspectives and Frames  41 risk standing unjustified unless and until backed by a yet higher-​order justification (Carroll 1895). These considerations favor ascribing the value that we ordinarily accord to understanding to the actual instantiation of the relevant explanatory structure in an agent’s cognition, rather than to the grasp of the higher-​ order proposition that this is the appropriate structure in which to relate the base-​level propositions. Further, emphasizing the practical difference between actually implementing an explanatory structure within cognition and merely endorsing a proposition specifying that structure opens the way to a final, more theoretical avenue for ascribing inherent epistemic value to characterizations. Alison Hills (2016: 679) argues that just as part of the inherent epistemic value of true beliefs derives from their being accurate reflections of the world in their contents, so too can a set of beliefs have inherent epistemic value in virtue of accurately reflecting the world in their form: in the network of dependence relations in which the individual beliefs stand. While a mere statement of those relations does accurately reflect the content of the world in its content, accurate mirroring through structural instantiation reflects that structure in a deeper and more direct way. Hills claims that such structural reflection is valuable in its own right, and not just because of the cognitive control it underwrites. We might add that it is such direct instantiation of the structure that underwrites appropriate cognitive control, rather than the reverse. That is, just as a map is a reliable tool for navigating through the world because it represents spatial relationships between represented objects by directly instantiating those very spatial relationships, so that transformations of the representing relationships automatically reflect transformations of the represented relationships (Camp 2018b), so too is an apt characterization a reliable tool for navigating the explanatory structure of the world because it directly instantiates those very relations.

5.  Conclusion In assessing these last arguments for the epistemic value of characterizations, it is important to remember just how far we have come. We began with the thought that while frames are effective manipulators of attention and interpretation, they and the perspectives and characterizations they produce are at best quick and dirty heuristics for achieving a simulacrum of genuine

42  Elisabeth Camp rationality, and at worst distractions from genuinely rational thought. Against this, I argued first, that frames can play a programmatic, research-​orienting function precisely because they are open-​ended, intuitive, and indeterminate, and so can suggest the potential attribution of features, and even the individuation of features, in ways that cannot currently be cashed out in independent terms. Second, I  argued that while most frames are eventually transcended as simplifications, perspectives continue to play a role not in the acquisition and assimilation of information, by establishing a conceptual taxonomy that subserves a profile of practical and explanatory purposes. Thus, even someone who rejects a distinctive epistemic contribution for characterizations and perspectives, because they endorse the assumptions of ontological and explanatory unitarianism at the end of inquiry should still grant their epistemic relevance in all but the most ultimately idealized epistemic contexts. The fact that frames, perspectives, and characterizations can make genuine epistemic contributions does nothing to dislodge the point that they can also manipulate cognition and occlude understanding. As tools for thought, frames and perspectives are indeed double-​edged swords. Their potential epistemic gains are counterbalanced by commensurate epistemic risks—​ especially, the risk of epistemic complacency. Given the depth and pervasiveness of perspectives in human cognition, however, the solution is not to abjure them in favor of a fantasy of purely logical thought. Instead, we need to harness the powers of logical articulation and reflection, as well as the interpretive disorientation that is produced by trying on conflicting perspectives, in order to employ frames actively and productively, and in order to assess the characterizations that they produce critically.

Acknowledgments This work was supported by the generous support of the Templeton Foundation. Thanks for useful discussion to audiences at the “Varieties of Understanding” conference at Fordham University and “The Nature of Belief ” workshop at the University of Santa Barabara, as well as at Arché St. Andrews, Harvard University, Northwestern University, and the University of Wisconsin, Madison. Special thanks to Catherine Elgin, John Bengson, Stephen Grimm, Jenann Ismael, Deborah Marber, and Jessica Wright for especially in-​depth and fruitful conversations.

Perspectives and Frames  43

References Bartels, Dand (2008): “Principled Moral Sentiment and the Flexibility of Moral Judgment and Decision Making,” Cognition 108:2, 381–​417. Bezuidenhout, Anne (2001):  “Metaphor and What Is Said:  A Defense of a Direct Expression View of Metaphor,” Midwest Studies in Philosophy 25, 156–​186. Bowdle, Brian, and Dedre Gentner (2005):  “The Career of Metaphor,” Psychological Review 112:1, 193–​216. Boyd, Richard (1979/​ 1993):  “Metaphor and Theory Change:  What Is ‘Metaphor’ a Metaphor For?,” in Metaphor and Thought, 2nd ed., ed. A. Ortony (Cambridge: Cambridge University Press), 481–​532. Camp, Elisabeth (2006):  “Metaphor and That Certain ‘Je Ne Sais Quoi,’” Philosophical Studies 129:1, 1–​25. Camp, Elisabeth (2008): “Showing, Telling, and Seeing: Metaphor and ‘Poetic’ Language,” Baltic International Yearbook of Cognition, Logic, and Communication, vol. 3: A Figure of Speech: Metaphor [Online], 1–​24. Camp, Elisabeth (2009): “Two Varieties of Literary Imagination: Metaphor, Fiction, and Thought Experiments,” Midwest Studies in Philosophy: Poetry and Philosophy XXXIII, ed. Howard Wettstein (Oxford: Wiley-​Blackwell), 107–​130. Camp, Elisabeth (2013): “Slurring Perspectives,” Analytic Philosophy 54:3 (September), 330–​349. Camp, Elisabeth (2015):  “Logical Concepts and Associative Characterizations,” in The Conceptual Mind:  New Directions in the Study of Concepts, ed. E. Margolis and S. Laurence (Cambridge, MA: MIT Press), 591–​621. Camp, Elisabeth (2017a):  “Why Metaphors Make Good Insults:  Perspectives, Presupposition, and Pragmatics,” Philosophical Studies 174:1, 47–​64. Camp, Elisabeth (2017b):  “Perspectives in Imaginative Engagement with Fiction,” in Philosophical Perspectives, Philosophy of Mind 31:1, ed. J. Hawthorne (Oxford: Wiley-​ Blackwell), 73–​102. Camp, Elisabeth (2018a):  “Insinuation, Common Ground, and the Conversational Record,” in New Work in Speech Acts, ed. D. Harris, D. Fogal, and M. Moss (Oxford: Oxford University Press), 40–​66. Camp, Elisabeth (2018b): “Why Cartography Is Not Propositional,” in Non-​Propositional Intentionality, ed. A. Grzankowski and M. Montague (Oxford:  Oxford University Press),  19–​45. Camp, Elisabeth (2019a):  “Slurs as Dual-​Act Expressions,” in Bad Words, ed. D. Sosa (Oxford: Oxford University Press), 29–​59. Camp, Elisabeth (2019b): “Imaginative Frames for Scientific Inquiry: Metaphors, Telling Facts, and Just-​So Stories,” in The Scientific Imagination, ed. P. Godfrey-​Smith and A. Levy (Oxford: Oxford University Press). Carnap, Rudolf (1932): “The Elimination of Metaphysics through the Logical Analysis of Language,” Erkenntnis,  60–​81. Carroll, Lewis (1895): “What the Tortoise Said to Achilles,” Mind 104:416, 691–​693. Cartwright, Nancy (1999):  The Dappled World:  A Study of the Boundaries of Science (Cambridge: Cambridge University Press). Coll, Richard K., Bev France, and Ian Taylor (2012): “The Role of Models and Analogies in Science Education:  Implications from Research,” International Journal of Science Education 27:2, 183–​198.

44  Elisabeth Camp Currie, Gregory (2010): Narratives and Narrators (Oxford: Oxford University Press). de Regt, Henk W. (2009): “The Epistemic Value of Understanding,” Philosophy of Science 76, 585–​597. De Sousa, Ronald (1987): The Rationality of Emotion (Cambridge, MA: MIT Press). Devine, Patricia (1989):  “Stereotypes and Prejudice:  Their Automatic and Controlled Components,” Journal of Personality and Social Psychology 56:1, 5–​18. Duncan, Birt (1976):  “Differential Social Perception and Attribution of Intergroup Violence: Testing the Lower Limits of Stereotyping of Blacks,” Journal of Personality and Social Psychology 34, 590–​598. Elgin, Catherine (2006): “From Knowledge to Understanding,” In Epistemology Futures, ed. Stephen Hetherington (Oxford: Clarendon Press), 199–​215. Evans, Jonathan, and Keith Frankish (2009) eds.:  In Two Minds:  Dual Processes and Beyond (Oxford: Oxford University Press). Fox Keller, Evelyn (1995):  Refiguring Life:  Metaphors of Twentieth Century Biology (New York: Columbia University Press). Gendler, Tamar Szabó (2007):  “Philosophical Thought Experiments, Intuitions, and Cognitive Equilibrium,” Midwest Studies in Philosophy 31, 68–​89. Gendler, Tamar Szabó (2008): “Alief and Belief,” Journal of Philosophy 634–​663. Gendler, Tamar Szabó (2011): “On the Epistemic Costs of Implicit Bias,” Philosophical Studies 156: 33–​63. Gentner, Dedre, and Michael Jeziorski (1993):  “The Shift from Metaphor to Analogy in Western Science,” in Metaphor and Thought, 2nd ed., ed. A. Ortony (Cambridge: Cambridge University Press), 447–​480. Giere, Ronald (1988): Explaining Science: A Cognitive Approach (Chicago: University of Chicago Press). Ginsborg, Hannah (2011): “Primitive Normativity and Skepticism about Rules,” Journal of Philosophy 108(5), 227–​254. Glynn, S. M. (2008). “Making Science Concepts Meaningful to Students:  Teaching with Analogies.” In Four Decades of Research in Science Education: From Curriculum Development to Quality Improvement, ed. S. Mikelskis-​Seifert, U. Ringelband, and M. Brückmann (Münster: Waxmann), 113–​125. Godfrey-​ Smith, Peter (2006):  “The Strategy of Model-​ Based Science,” Biology and Philosophy 21, 725–​740. Hesse, Mary (1993):  “Models, Metaphors and Truth,” in Knowledge and Language, vol. 3: Metaphor and Knowledge, ed. F. R. Ankersmit and J. J. A. Mooij (Dordrecht, Netherlands: Springer),  49–​66. Hills, Allison (2016): “Understanding Why,” Nôus 50:4, 661–​688. Ismael, Jenann (2018): “Why (Study) the Humanities?,” in Making Sense of the World: New Essays on the Philosophy of Understanding, ed. S. Grimm (New York: Oxford University Press). Kahneman, Daniel (2011): Thinking Fast and Slow (New York: Farrar, Straus and Giroux). Kahneman, Daniel, Paul Slovic, and Amos Tversky (1982) eds.:  Judgment Under Uncertainty: Heuristics and Biases (Cambridge: Cambridge University Press). Kahneman, Daniel, and Amos Tversky (1981):  “The Framing of Decisions and the Psychology of Choice,” Science 211:4481, 453–​458. Kermode, Frank (2000):  The Sense of an Ending:  Studies in the Theory of Fiction (New York: Oxford University Press).

Perspectives and Frames  45 Korsgaard, Christine (2009): “The Activity of Reason,” Proceedings and Addresses of the American Philosophical Association 83:2, 23–​43. Kvanvig, Jonathan (2009):  “The Value of Understanding,” in Epistemic Value, ed. A. Haddock, A. Millar, and D. Pritchard (Oxford: Oxford University Press), 309–​313. Lakoff, George (2004):  Don’t Think of an Elephant! Know Your Values and Frame the Debate: The Essential Guide for Progressives (New York: Chelsea Green Publishing). Levin, Irwin, Sandra Schneider, and Gary Gaeth (1998): “All Frames Are Not Created Equal: A Typology and Critical Analysis of Framing Effects,” Organizational Behavior and Human Decision Processes 76:2, 149–​188. Locke, John (1689/​1989): An Essay Concerning Human Understanding, ed. P. Nidditch (Oxford: Clarendon Press). McDowell, John (1994): Mind and World (Cambridge, MA: Harvard University Press). Mitchell, Sandra (2003):  Biological Complexity and Integrative Pluralism (Cambridge: Cambridge University Press). Moran, Richard (1989):  “Seeing and Believing:  Metaphor, Image and Force,” Critical Inquiry 16(1), 87–​112. Murphy, G., and D. Medin (1985):  “The Role of Theories in Conceptual Coherence,” Psychological Review 92, 289–​316. Musch, Jochen, and Karl Christoph Klauer (2003) eds.:  The Psychology of Evaluation: Affective Processes in Cognition and Emotion (Hillsdale, NJ: Erlbaum). Peacocke, Christopher (1992): A Study of Concepts (Cambridge, MA: MIT Press). Pólya, George (1954): Mathematics and Plausible Reasoning, vol. 1: Induction and Analogy in Mathematics (Princeton: Princeton University Press). Pylyshyn, Zenon (1993):  “Metaphorical Imprecision,” in Metaphor and Thought, ed. Andrew Ortony (Cambridge: Cambridge University Press), 481–​532. Sloman, Steven, Bradley Love, and Woo-​Kyoung Ahn (1998):  “Feature Centrality and Conceptual Coherence,” Cognitive Science 22:2, 189–​228. Thagard, Paul (1989):  “Explanatory Coherence,” Behavioral and Brain Sciences 12, 435–​502. Trout, J. D. (2002), “Scientific Explanation and the Sense of Understanding,” Philosophy of Science 69, 212–​233. Tversky, Amos (1977): “Features of Similarity,” Psychological Review 84, 327–​352. Wittgenstein, Ludwig (1953):  Philosophical Investigations, trans. G. E.  M. Anscombe. (Oxford: Basil Blackwell). Woodfield, Andrew (1991): “Conceptions,” Mind 100:399, 547–​572.

3 The Epistemologies of the Humanities and the Sciences Richard Foley

In this essay, I’ll be providing an overview of some of the themes of my book The Geography of Insight: The Humanities, The Sciences, How They Differ, and Why They Matter (2018). The central questions I address in the book are, do inquiries in the sciences and the humanities typically have different intellectual aims, and should they? A pair of closely related questions is whether the values informing research in the sciences are usually different from those in the humanities and whether they should be. I think the answer to all these questions is yes. In particular, there are four major differences: 1. The sciences, and especially the basic sciences, place a special value on findings that are minimally indexical in the sense of not being about particular places, times, or things. The ideal is to arrive at generalizations accurate for everything and every time and place without exception, or at least with as few exceptions as possible. In the humanities, minimizing indexicality isn’t nearly so valued, and it shouldn’t be, since the issues are ordinarily ones about particular people and events at specific times and places. To exert pressure in the direction of greater generalization and less indexicality is to risk glossing over important details. 2. The sciences prize findings that are as independent as possible of the perspectives of those conducting the inquiry. Minimizing perspectival influences reduces the risks of distortion, and it also broadens the potential audience. It increases the chances of the results of inquiries being comprehensible to those with different perspectives, perhaps even to sufficiently intelligent creatures with other kinds of perceptual and

48  Richard Foley cognitive faculties, if there are such beings elsewhere in the universe. Reducing perspectival factors to a minimum, by contrast, shouldn’t be the aim in the humanities. Their issues require an understanding of the conscious states and points of view of humans. Human inquirers have a built-​in advantage in dealing with these issues. They have knowl­ edge of what human consciousness is like from their own case, which they should make use of instead of trying to conduct inquiries in ways that leave as few traces as possible of their own perspectives. But to the degree that the conclusions reached and the grounds for them reflect distinctive features of the inquirers’ perspectives, they are less likely to be intelligible to others whose locations, histories, faculties, and hence perspectives are quite different. The audience is thus always a more restricted one, and the inquiries themselves have to be conducted with the intended audiences and their distinctive perspectives in mind. 3. The aim of the sciences is to be entirely descriptive. Inquirers are expected to limit themselves to describing and explaining what is (or was or will be) the case, but this needn’t be the sole aim for inquirers in the humanities. They too have descriptive aims, but they’re also free to be concerned with prescriptive claims, which make recommendations about what should be (or should have been) the case. As such, they give expression to values. Thus, it isn’t uncommon or in any way inappropriate for historical accounts to describe the details of a war, political campaign, or social movement but also to make assessments of whether their consequences would have been better (or worse) had their participants made different decisions. 4. In the sciences, the most important goal is that of increasing the stock of collective knowledge. The individual knowledge most highly valued is that which makes a contribution to what is collectively agreed upon and known. In the humanities, by contrast, individual insight is highly valued for its own sake, independently of any contribution it’s capable of making to generating a consensus. Indeed, many of the issues are such that it’s not feasible to expect consensus, not even in the long run. In everyday contexts, the terms “knowledge,” “understanding,” and “insight” are largely interchangeable, and I too will be treating them as such. A number of contributions to this volume make suggestions about how to introduce precise distinctions among the terms. I applaud these efforts, but it’s best to regard them not as reports on actual usage but as attempts to improve

The Epistemologies of the Humanities and the Sciences  49 it. For my purposes, however, following the everyday loose usage creates a big tent and helps to avoid getting bogged down in terminological disputes. Now, for some additional details on the distinctions.

1.  Indexical vs. Non-​indexical When a claim is indexical, it’s about particular times, places, or things, whereas non-​indexical claims have no such restrictions. They are intended to be about all times, places, and things. Indexicality is not all-​or-​nothing. It comes in degrees. The fewer the exceptions, that is, the fewer times, places, and things that are excluded, the less indexical the claim. In the sciences, especially the basic sciences, there is pressure to arrive at knowledge that is as non-​indexical as possible. Consider the theory of gravitation. It aspires to describe the attraction that objects with mass have for one another not only on Earth, and not only in our Milky Way galaxy, and not only over the last thousand or million years, but for the entire universe at all times. The goal is nothing less than to remove all traces of indexicality. This has turned out to be a frustratingly difficult goal to achieve, but still, it’s the ideal and remains so even if there are limitations on fully satisfying it. As with other ideals that cannot be completely realized, the ambition is to approach it as closely as possible, which in this case means to keep trying to reduce the times, places, and things to which the account doesn’t apply. In the humanities, most of the issues are such that knowledge with a high degree of indexicality is just what is being sought. So in general, there’s nothing like the premium on minimizing indexicality one finds in the sciences. I say “in general” because philosophy may be an exception. In all of its fields, there’s a tendency to search for simple and extremely general principles. On the other hand, it’s precisely this penchant for the simple and gen­ eral that Wittgenstein in his late period famously regarded as responsible for the central shortcomings of philosophy as a discipline.1 But for purposes here, I’ll be putting philosophy and these disputes about it to one side. If one looks at the rest of humanities, one sees project after project that is seeking

1 Ludwig Wittgenstein, The Blue and Brown Books (London:  Blackwell, 1958). See also Paul Horwich’s treatment of these issues in Horwich, Wittgenstein’s Metaphilosophy (Oxford:  Oxford University Press, 2012).

50  Richard Foley insights about particular times and regions of the world, whether they be the causes and effects of the Hundred Years’ War between England and France, the characteristics of Athenian democracy, or the significance of the Harlem Renaissance. The knowledge being sought is highly indexical: knowledge of particular places, times, people, and institutions. This shouldn’t be surprising, given that the issues of greatest interest to the humanities are ones that arise within a limited subset of times and places, namely, those associated with modern human societies. These issues are so intertwined with the details of those societies that satisfactory accounts of them likewise have to be contextualized with information about specific times and places. How contextualized is always a relevant question, however. Among historians, there are those who search for understanding by focusing on a few especially prominent individuals and events. Others go down a level and concentrate more on the details of ordinary people in their everyday lives, while still others go up a level and direct their attention to motivating political ideas or underlying economic pressures. Yet others go up yet a level further and treat the issues under study in terms of a much larger, long-​term system in whose workings individual and collective agency seem not to be so central. Think of accounts that try to explain the differential paces of societal transformations in different regions of the world in terms of such factors as climate differences and diversity of plant and animal species instead of political or cultural factors.2 Historians pick the dimension they think is most appropriate to fix upon in order to make the past intelligible, where appropriateness, as always, is a function of context and purpose, including views about the value of doing history in general. As a result, there’s no agreed-​upon, single best way of providing historical understanding. In the search for historical understanding, there can always be questions of how far back to go, where to look, and how widely to look.3 In addition, once the dimension is settled, it’s still not the case that the only insights of interest are ones about the particular contexts or events being studied. The best works in the humanities escape their immediate 2 See Jared Diamond, Guns, Germs, and Steel: The Fates of Human Societies (New York: Norton, 1997); and Peter Burke, The French Historical Revolution: The Annales School, 1929–​2014, 2nd ed. (Palo Alto: Stanford University Press, 2015). 3 Daniel Little, “Philosophy of History,” The Stanford Encyclopedia of Philosophy (Winter 2012 Edition), Edward N. Zalta (ed.). URL = http://​plato.stanford.edu/​archives/​win2012/​entries/​history/​.

The Epistemologies of the Humanities and the Sciences  51 subject matters to reveal structures, patterns, and dependencies that can lead to insights elsewhere. There are limits, however. Although there are general understandings to be had, they don’t come in the form of hard-​ and-​fast generalizations, which can be used to deduce explanations and predictions about other situations. The insights are instead ones that are instructive in a this-​ is-​ often-​ the-​ case way when thinking about other situations, and even here there are limits on how widely they can be useful. As similarities between the contexts decrease, so too do the chances that insights about the one will be helpful in understanding the other. It’s sometimes said that a mark of a genuinely great work in the humanities is that its insights are universal, but this is an exaggeration. Even the most perceptive accounts of the social and political conditions in France around the time of French Revolution aren’t going to have much relevance for understanding, say, early Inuit societies or the Cro-​Magnon humans living in Europe 40,000 to 10,000 years ago.

2.  Perspectival vs. Non-​perspectival Inquirers have distinctive perspectives, or points of view, that influence the conduct of their inquiries, the conclusions reached, and the extent to which the conclusions, and the grounds for them, can be understood by others. These perspectives are the products of the cognitive abilities and sensory faculties the inquirers have, the particular positions in space-​time they occupy, the physical and social environments that surround them, and their own peculiar histories and experiences within these environments. All of these together shape the information they have, the concepts they use, and the values they adhere to. It sometimes is a fitting goal for inquirers, however, to seek insights that are as independent as possible of their own perspectives. On the other hand, other times this is neither a desirable nor a feasible goal. Projects in the sciences tend to fall in the former category, those in the humanities in the latter. The value the sciences put on minimizing the influence of perspective is on display in pretty much all of their core practices: the importance of experiments being replicable; the insistence on publicly observable and quantifiable results; the use of instruments to measure these quantities; the tradition of team-​conducted research; the importance of collective knowl­ edge along with its associated emphasis on knowledge that can be readily

52  Richard Foley transferred from one investigator to another; and the reliance on mathematics as the language of science. These staples of scientific research help counter risks of distortions that can be present in any given perspective, but they also potentially increase accessibility. By promoting the development of accounts that aren’t tightly bound to the locations, histories, and faculties of any single inquirer or group of them, they enhance the possibility of the findings being intelligible to sufficiently advanced inquirers whose locations, contexts, experiences, and personal histories are greatly different. “Sufficiently advanced” is intentionally broad. It’s meant to capture the possibility of the insights being comprehensible not only to other human inquirers whose locations and histories are different, provided they have the requisite intelligence, training, and background information, but conceivably even to intelligent creatures elsewhere in the universe with perceptual and cognitive faculties quite different from humans and who thus experience the world quite differently. Minimizing the influence of perspectival factors normally isn’t and shouldn’t be so valued in the humanities. The humanities are largely concerned with what Helen Small has called “the meaning making practices of human culture, past and present.”4 These practices are so contextually situated that insights about them require background information, concepts, and values associated with particular times, places, and societies. Highly indexical factors, in other words. But in addition, the issues are so entangled with human consciousness that successful treatments have to include also insights about the characteristic intentions, purposes, and points of views of the relevant individuals and groups, which inquirers try to provide by extrapolating from their own experiences and histories. There’s no reason to think that human inquirers are in a specially privileged position with respect to the issues of the basic sciences, but they do have advantages when it comes to those of the humanities (as the term itself might suggest). Human inquirers have firsthand understanding of what human conscious states and human points of view are like, and of what it’s like to live in human societies. We humans tend to take this knowledge for granted, but suppose again that somewhere there are sophisticated inquirers with non-​human cognitive and perceptual faculties living in societies quite different from those of humans. No matter how intelligent they may be, they 4 Helen Small, The Value of the Humanities (Oxford: Oxford University Press, 2013).

The Epistemologies of the Humanities and the Sciences  53 would face greater obstacles than we in understanding human lives and societies. For we but not they can draw upon our own human experiences and histories when addressing issues that are intertwined with the conscious states and points of view of other humans. But this is just to say that in dealing with such issues, human inquirers can, do, and should exploit their distinctive ways of apprehending the world as opposed to trying to escape from their perspectives to the extent possible. Extrapolating from their own perspectival resources, inquirers try to make inferences about the beliefs, attitudes, purposes, and points of views of those who are their subjects. Difficult as this may be, things get harder still when the project, as in the humanities, is to work up the inferences into an account that will be illuminating to others. For then it’s also necessary to take into consideration the perspectives of the intended audience. The task facing inquirers is to mobilize enough of their own experiences, background information, concepts, and values that have overlaps with those of the audience to make intelligible to this audience the consciousness-​inflected issues under study. The audience can be more or less broad, but unlike the sciences, whose claims are ideally intended for a contextually unrestricted audience, in the humanities the audience is always limited. The goal isn’t to arrive at insights that are as widely accessible as possible, certainly not to all individuals of comparably advanced cognitive abilities, however spatially, temporally, and culturally distant they may be and however different their faculties and histories are. It’s one of more local appropriation, of making the issues intelligible to us, where the “us” can be more widely or narrowly defined—​our group, our society, our age, or whatever. This isn’t the right approach for all the issues of interest to us about conscious states and the creatures that have them. Questions about the neurophysiological conditions associated with conscious phenomena can be addressed in as non-​perspectival and non-​indexical way as any other kind of scientific inquiry. That’s not the problem. The problem, rather, is that even a complete as possible neurophysiological account of a creature’s brain wouldn’t portray what it is like to be that conscious subject, what it is like to be a bat in a bat-​like environment, to use Thomas Nagel’s famous example.5 That’s not what a neurophysiological account is supposed to be about. This point is relevant, moreover, not just for bats but also for the vast set of phenomena that arise only because there are creatures in the world 5 Thomas Nagel, Mortal Questions (Cambridge: Cambridge University Press, 1979).

54  Richard Foley with conscious states, especially a particular kind of creature—​humans. A large proportion of these phenomena and the issues they generate are so intertwined with human experiences and purposes that satisfying treatments of them have to take into account what it is like to be these creatures, that is, what it is like to be human. All the issues most representative of the humanities are examples. Understanding human societies, arts, wars, legal systems, religions, and relationships require understanding what it is to experience the world as human. And when the issues concern particular individuals or groups of individuals in specific situations, as they commonly do, accounts need to take into account how those situations look from the points of views of these individuals.

3.  Prescriptive vs. Descriptive Descriptive claims are meant to express facts. They profess to be about what is the case, whereas prescriptive claims are about what should be the case. They give expression to values, sometimes moral values and other times political, aesthetic, or prudential ones. Like the other distinctions, this one is not all-​ or-​nothing. Many claims have both descriptive and prescriptive elements, at times so deeply interwoven with one another as to be virtually inseparable.6 Questions about the nature of prescriptive claims are as old as philosophy itself and still vigorously debated, but for purposes here what matters are two non-​nihilistic theses that are widely, if not universally, accepted. The first is simply that there is some distinction to be made between the descriptive and the prescriptive. Explaining what it amounts to may be difficult, but at the end of the day it has to be acknowledged that, for example, various ethical principles proposed over the centuries by philosophers such as Kant (act on maxims that can be universalized) and Mill (act to maximize utility) are more prescriptive and less descriptive than, say, Newton’s laws of motion. The second thesis is that there are standards for assessing prescriptive claims, and those meeting the standards are superior to those that do not. They are, depending on how one understands the nature of prescriptive claims,

6 See Bernard Williams’s discussion of treachery, cowardice, brutality, and other such “thick” moral concepts in his Ethics and the Limits of Philosophy (Cambridge: Harvard University Press, 1985).

The Epistemologies of the Humanities and the Sciences  55 more accurate or more coherent or more fitting or in some other way more defensible. It follows that prescriptive issues are also suitable subjects for inquiry, and it’s possible to have insights about them. The sciences cannot have the primary responsibility in providing these insights, however. Their aim is to be completely descriptive. Theories of gravitation seek to understand how bodies with mass attract one another, not how they might best do so. One can entertain fantasies about gravity being a weaker or stronger force than it is in our world, but this is the stuff of science fiction, not science. Like the sciences, the humanities have descriptive aims. An account of the Battle of Waterloo will seek to depict elements of the battle—​the sizes of the opposing forces, their weaponry, their positions, and so forth—​that help explain why it unfolded as it did. Similarly for projects about the roles that canals and trains played in the Industrial Revolution or the use of meter in different traditions of poetry. Here too the goal is accurate, detailed descriptions. There’s nothing to prevent inquiries in the humanities from being wholly occupied with descriptive matters, but many have prescriptive aspirations as well. Thus, it isn’t unusual or in any way out of place for historical accounts about the end of slavery in the United States to assess the trade-​offs that Lincoln agreed to in order to get the Emancipation Proclamation passed. To make such assessments, historians need to look in detail at the context in which Lincoln made his decisions and evaluate whether different decisions might have been preferable. These evaluations, in turn, involve value judgments about how best to think about political trade-​offs and political leadership in general. So too it with other studies in the humanities that have mixtures of descriptive and prescriptive aspirations, from those assessing T. S. Eliot’s status as a 20th-​century poet to ones appraising the wisdom of the reparations imposed upon Germany following World War I. Descriptive accuracy is a prerequisite of such projects, but they also defend evaluative conclusions. In developing views about such issues, inquirers have no choice but to enter the realm of the prescriptive. They have to make value judgments. They can be criticized for how they go about doing so. They may be relying on sloppily acquired information, or biases may be causing them to overlook relevant facts, or they may be giving too much weight to relatively inconsequential considerations. The criticism, however, cannot simply be that ethical, political, aesthetic, and prudential values are affecting their conclusions,

56  Richard Foley since about these issues, there’s no alternative but to make appeal to such values.

4.  Individual vs. Collective If there’s consensus within a group about the truth of a claim and the group has accurate and comprehensive enough information about it, then provided the claim is true, it can be regarded as part of the group’s collective stock of knowledge. One pathway to consensus is for enough individuals in the group to determine on their own that the claim in question is true. More interesting for purposes here, however, is that even when only a few members are in a position to determine its truth, the claim can still be collectively believed and known if these few have the requisite intellectual authority within the group and have vouched for it. This way of achieving consensus and collective knowledge is especially important in scientific fields that are so specialized that only a small number of highly trained individuals are able to understand the work being done, much less defend the results. Their discoveries can nonetheless expand the collective stock of knowledge available to others. Science is a collective enterprise driven forward by individual effort and achievement. The key tool for incorporating individual contributions into the collective enterprise is the division of intellectual labor. The system operates by breaking up problems into components, incentivizing investigators to develop highly refined expertise in these narrowly defined areas, and requiring them to make their findings and data publicly available. This allows other specialists to verify the results, but it also allows those working in adjacent fields to make use of the findings even if they themselves are not in a position to provide firsthand defenses of them. Collective knowledge plays a far less central and institutionalized role in the humanities, even though they too are communal enterprises. The fields making up the humanities and the standards for conducting inquiries in them have been developed over time by generations of scholars and students. Insights coming out of these inquiries are thus not to be understood simply as the products of solitary scholars doing their jobs in isolation. They’re also products of a social practice that has been collectively constructed.

The Epistemologies of the Humanities and the Sciences  57 There are also more everyday ways in which the humanities are communal enterprises. Scholars rely on research done by peers, past and present. They use archival materials that have been collected and preserved by others. There are schools of thought that encourage the use of preferred methodologies, and movements that direct inquirers toward neglected areas of research. In addition, there are cooperative efforts of a more limited sort, for example, anthologies, book series, and conferences. Even so, there’s nowhere near the overall emphasis on consensus and collective knowledge one finds in the sciences. Nor do the humanities have the elements that make this emphasis work so well in the sciences. Sharply honed divisions of intellectual labor are often not as feasible, for example. Nor is it appropriate to require that hypotheses be capable of generating observable, quantifiable predictions. In addition, there’s simply less pressure in the humanities to achieve consensus. The sciences have well-​defined pragmatic as well as intellectual agendas. The aim is not just to understand the world but also to control it. Science goes in hand in hand with engineering and technology, with the former being the supplier of information for the latter, but to carry out their projects, engineers and technicians need be able to draw upon a collectively agreed-​upon stock of knowledge. By contrast, with respect to many of the topics of most interest in the humanities, consensus isn’t to be expected. There also isn’t as much room for deference. Works in the humanities, as in the sciences, are assessed in terms of how influential they are, but in the humanities this influence should ordinarily be the result by putting others in a position to apprehend for themselves the insights in questions. The most important role for experts to play in the humanities is that of exerting what can be called “Socratic influence,” not intellectual authority.7 Suppose you get me to believe a claim through a series of well-​thought-​out questions and instructions. Afterwards, I understand what you understand and believe what you believe, but my believing isn’t dependent upon you. I have come to know it and be able to defend it on my own. You have exercised Socratic influence over me but not authority. Like the other distinctions being discussed, this one comes in degrees. There are many shades of gray between sheer deference on the one hand and understanding completely on one’s own on the other. In any particular 7 The term “Socratic influence” is from Alan Gibbard, Wise Choices, Apt Feelings (Cambridge:  Harvard University Press, 1990).

58  Richard Foley case, it can be some combination of the two that leads someone to accept the conclusions of an expert. And more generally, it has to be acknowledged that we are all, as the historian Marc Bloch once memorably made the point, “permanently floating in a soup of deference” to others.8 Much of what we believe invariably comes from what others have said or written. There’s no escaping this. No escape in one’s everyday life, no escape in one’s work life, and if one is a scholar, no escape in one’s scholarly life. So, the thesis here isn’t that deference to others has no role at all to play in the humanities. It does, and I’ll be discussing in the section on intellectual authority when and why it does. Nor is the thesis that Socratic influence has no place whatsoever in the sciences. The thesis, rather, is the more nuanced one that in the humanities there’s nothing like the organized system of specialization and deference to the views of the experts that one finds in the sciences, and likewise nowhere near the emphasis on reaching consensus. Together, the preceding four distinctions provide a sketch of the epistemologies of the humanities and the sciences respectively. In addition, these core distinctions get expressed in a set of further differences, which like the primary ones, are intertwined with one another. I won’t have time to discuss all of these secondary distinctions, but some of the more important follow, along with some brief comments on them:

The humanities and sciences have different takes on the notion of there being a natural endpoint to inquiry Inquiry in the sciences is designed to move over time in the direction of a definitively accurate and comprehensive account on which there’s consensus. Such an account would represent a natural stopping point for inquiry. More accurately, this would be so for theoretical inquiry. Efforts to find new applications of the knowledge needn’t come to an end. It’s just that there wouldn’t be anything to motivate major revisions of the underlying theory. There needn’t be realistic prospects of achieving such an endpoint anytime soon or for that matter in the foreseeable future. A definitive account can nonetheless function as a regulative ideal, the ambition being to make movement toward it. 8 Marc Bloch, “Rituals and Deference,” in H. Whitehouse and J. Laidlaw (eds.) Rituals and Memory: Towards a Comparative Anthropology of Religion (London: Altamira Press, 2004).

The Epistemologies of the Humanities and the Sciences  59 In the humanities, on the other hand, this usually isn’t a suitable aim at all. With the exception of a few sharply defined issues, which I’ll be discussing subsequently in the section on intellectual authority, the notion of a definitive account on which there’s consensus cannot function as an ideal, not even one that might be approached in the long run. The nature of the issues being addressed, with their characteristic combination of indexical, prescriptive, and perspectival elements, means that inquiry into them is inescapably open ended. Open ended in the sense that major additional insights and revisions are always possible, always to be expected, and always to be sought.

Different notions of intellectual progress are appropriate for the sciences and humanities Progress in the sciences can be visualized as movement in the direction of an endpoint: an account that is not only fully accurate and comprehensive but also collectively agreed upon. But not so for the humanities. Advances there are better thought of as coming in the form of greater precision, breadth, and coherence, and with individual insights being prized even when they aren’t useful in producing a convergence of opinion. Think by way of analogy with the development of a body of law over time and the role played in this development by appellate judges. The cases coming before judges often arise from a need to revise current law to fit new circumstances. To arrive at decisions about these cases, judges frequently have to reach beyond the law and make use of extralegal norms. When doing so, they should draw upon norms they would also find acceptable outside of legal contexts.9 Even then, before finalizing their decisions, they should be able to articulate the bases of any disagreements between their prospective rulings and the rulings in similar cases. Their task is to fit all these elements—​the existing law, the possible revisions or extensions of it, the detailed facts of the case, the relevant extralegal norms, the bases of disagreements with other rulings—​into a coherent whole they are willing to stand behind. Ideally, the conclusion should be such that further reflection wouldn’t cause them to change their mind. It should in this sense be invulnerable at least to self-​criticism. 9 Ronald Dworkin, Law’s Empire (Cambridge:  Harvard University Press, 1986); and Ronald Dworkin, Justice for Hedgehogs (Cambridge: Harvard University Press, 2011).

60  Richard Foley “At least” because others may well be highly critical, but even if the decision doesn’t immediately produce disagreements, it remains open to further testing and revision as new circumstances emerge. So, it’s not as if there’s an expectation that the decision will settle the issue for all time. Nevertheless, when everything goes well, the result is a body of law of greater precision, breadth, and coherence. In other words, progress. Progress in the humanities can be conceived similarly, with humanities scholars playing roles analogous to judges in the law. The conclusions reached by humanities scholars do not, of course, have the binding character of rulings by judges, but in other respects the analogies are striking. Just as judges must master the elements of their cases, humanities scholars must master the materials relevant to their issues, whether these are archives or texts or whatever. And just as judges have a responsibility to understand the rulings of current and past peers and the bases of any disagreements with them, so too humanities scholars need to be conversant not only with the works of their contemporaries but also those of past scholars. This means that unlike scientists, they always have a reason to engage the works of their predecessors. Their interest in the histories of their fields ought not be merely antiquarian, as it can be and usually is in the sciences. There are additional parallels. One of the most familiar features of law is how changing circumstances can be the occasion for revised interpretations of established precedents. So too in the humanities, social and political developments can cause there to be fresh looks at entrenched views and ignored issues. The movement for women’s rights is but one example but a powerful one. It produced political and social reforms but has also led to major changes in the kinds of research projects undertaken in the humanities. Intellectual developments, in particular engagements with new scientific knowledge, can do the same. One example, but again an especially far-​ reaching one, comes from evolutionary biology, which has shown how the bodies and faculties of humans have been shaped by the same processes of natural selection that shaped other animal and plant life. To many (but to be sure, not all) this insight now seems unremarkable, but it represents a profound shift from earlier eras that insisted on a clear-​cut distinction between humans and other living things. The recognition that there aren’t such sharp distinctions to be made is one of the factors responsible for the increased interest one sees across the humanities in examining anew the similarities, differences, and appropriate relationships between humans and other living things.

The Epistemologies of the Humanities and the Sciences  61

Intellectual authority plays different roles in the sciences and humanities In the humanities classic works are relevant to its current inquiries, whereas classic works in the sciences aren’t especially relevant to its current inquiries. On the other hand, reliance on expert authority plays a far more central role in the sciences than in the humanities. These two observations might appear to be in conflict with one another, but two further observations help explain why they’re not. The first is that the continuing relevance of great historical figures in the humanities isn’t a matter of deference to their authority. More on this in a moment. The second is that although in the sciences there is extensive reliance on authority, the relevant authorities are contemporary experts, not the great figures of the past. Great works in the sciences are rarely a part of the training of students, whereas as in the humanities, students are still often expected to have a working familiarity with the classic works in their disciplines. This difference is the result of the different kinds of progress to be made in the two fields. In the humanities, given the nature of its issues, there in general cannot be collective progress in the direction of a definitive account, whereas in the sciences there can be and often has been. The works of Kepler, for example, are thus too dated to be of much direct use to contemporary astronomers. The division of intellectual labor in the sciences, however, does allow inquirers to have trust in and make use of the findings of contemporary experts, and to do so even if they themselves would be unable to defend the findings. Many subfields of the sciences are so specialized that only a few highly trained individuals are in a position to provide firsthand, credible defenses of the new knowledge being produced. This knowledge is nonetheless available to the larger scientific community. Researchers in other fields who are working on adjacent issues defer to the specialists in the subfield. This deference isn’t automatic or irrevocable. There are always decisions to be made about whom to trust and how much to trust them. The structure of the overall system, however, makes it reasonable for inquirers to have a general attitude of confidence in the findings of others. One of key features of the system that encourages such trust is its public character. The works of even the most celebrated authorities are subject to scrutiny. Even they must make the data they collect available to others, and even they must formulate their hypotheses so that others can test them. Still, in the real world of science, even when other scientists have the expertise to re-​confirm the

62  Richard Foley hypotheses, there are limits to how much time and effort they should spend in doing so. After all, one of the most important benefits of the division of intellectual labor is that it frees inquirers to devote more time on their own issues, which in turn allows them to make more rapid progress. Influence is exercised differently in the humanities (and the arts as well). It can continue to be direct over long stretches of time. “Direct” in the sense that the vehicles of influence can be the preserved works themselves. This is palpably true in the arts—​Shakespeare’s plays are still studied and performed, Michelangelo’s sculptures still analyzed, and Rembrandt’s self-​portraits still debated—​but it’s also the case in the humanities. The works of Thucydides, Gibbon, Carlyle, and Tocqueville are still taken seriously by historians and their students, as are those of Aristotle, Descartes, Locke, and Kant by current philosophers. On the other hand, as observed earlier, the central role for experts to play in the humanities is that of exerting Socratic influence, not intellectual authority. There isn’t as much room for reliance on the authority of experts, and this includes great figures of the past. So, although classic works remain relevant to current inquiries in the humanities in ways that they aren’t in the sciences, this is a matter of engagement with them as opposed to deference to them. In the humanities, the questions are in general ones that don’t admit of definitive answers. They’re open ended. They’re ones to which different ages, different cultures, and different individuals must fashion answers appropriate for their circumstances. The focus is on reinterpreting for one’s self, one’s culture, or one’s era many of the very questions that the great historical figures grappled with and the conclusions they reached about them. There is, accordingly, less room for reliance on the authority of experts, whether they are the great figures of the past or the leading contemporary thinkers. There’s more of an emphasis on intellectual self-​reliance. Even in the humanities, there are situations where accepting claims on the basis of authority isn’t at all out of place, but when and where this is so is telling. The contexts are ones where the issues are wholly descriptive and highly indexical, and where, consequently, the humanities, like the sciences, can profit from divisions of intellectual labor. These are also the issues in the humanities where it sometimes may be possible to arrive at something resembling a definitive account. Examples are the details of particular event (when it occurred, where it occurred, who its participants were, etc.) or the provenance of a particular work of art. Because there can be definitive

The Epistemologies of the Humanities and the Sciences  63 accounts of such issues, there can also be definitive authorities about them, authorities to whom others should defer. But with the exception of issues that are wholly descriptive and highly circumscribed indexically, the issues in the humanities leave scant room for the practice of borrowing opinions from other experts. The primary model of influence is instead expected to be Socratic.

The humanities and sciences have different working assumptions about simplicity and complexity The standard working hypothesis in the sciences is that the laws of nature are simple. All else being equal, the fewer the assumptions, hypotheses, and postulates, the better the theory. The case for simplicity, moreover, isn’t just aesthetic or pragmatic. Simpler theories may be more beautiful and easier to use, but the claim is that complexity is also a mark that a theory doesn’t have things quite right. The preference for simple theories is in service to the search for truth. In the humanities, by contrast, there normally isn’t a built-​in preference for simplicity, nor should there be. The more appropriate attitude is one of wariness, simplicity being a warning flag that important details may be getting overlooked, or worse, that ideology is at work. A taste for some complexity should be the norm. This isn’t to say, however, that it’s always out of place to press for greater breadth and simplicity when addressing issues associated with human lives and human societies. There is no single best way to address these issues. If one takes an Olympian viewpoint about the approaches human inquirers take when seeking insights about their fellow humans, they can be placed into two broad categories. There are those that are primarily interested in uncovering similarities, and those that are more focused on differences. This divide, in turn, corresponds roughly with the divide between the studies of humans one finds in the sciences versus those in the humanities. The sciences tell us that humans are extremely alike physically. They display relatively little genetic diversity compared to other mammal species. We now know that there are also striking genetic similarities between humans and other living things. These overlaps provide exceptionally fertile grounds for research. As a result, at most universities in the world one finds research being done on the genetics of non-​human life forms, the hope being that this

64  Richard Foley work will lead to a more complete understanding of comparable processes in humans. These projects follow the familiar scientific pattern of assimilation and simplification. Genetic processes found in human beings are matched with ones found in fruit flies, worms, and yeast in a search for generalizations that govern all of them. Applying the knowledge is something else. Here differences among species are important, as are differences among individuals within a species. It’s not just for purposes of applications that we dwell on these differences, however. We’re also just plain curious about them. Curious about the differences between ourselves and other living things, but especially curious about differences among ourselves. A sign of our preoccupation is that we make far finer distinctions about one another than anything else. We make discriminations about our capacities, emotions, personalities, and backgrounds. Also, our physical features, environments, languages, and beliefs. And this is just the beginning. We distinguish on the basis of ethnicity, gender, nationality, religious background, and age, and seem to have almost inexhaustible interest in finding social, cultural, economic, and other differences correlated with these categories. The availability of so many distinctions and the zeal with which we deploy them may make it appear as if people are enormously different from one another, but any careful look reveals this to be an exaggeration. Why is it that our differences so intrigue us? Perhaps because a mastery of them helps us with the intricacies of social relations, but whatever the explanation, it’s with these discriminations that the humanities (and the arts) come into their own. The pattern in the sciences is to assimilate seemingly disparate phenomena in efforts to find generalizations that apply ever more widely. The pattern throughout most of the humanities is the opposite. It’s to look for ways of distinguishing even seemingly similar phenomena in an effort to reveal hitherto unnoticed complexities. Once again, these are tendencies. The emphasis on reducing indexicality in the sciences varies from field to field. Physics is different from biology, and both physics and biology are different from viticulture or comparative anatomy. I’ve also already made reference to how in the humanities, philosophy may be a special case. And more generally, I’ve noted that there’s a range of options for inquirers in the humanities to take in deciding how broadly to cast the net in the search for understanding. Still, the overall propensities of the sciences and the humanities are poles apart. Assimilation and simplification on the one hand. Diversification and

The Epistemologies of the Humanities and the Sciences  65 complication on the other. Each has its place. There’s potentially enormous power in simplicity and potentially great richness in complexity. There’s obviously much more to be said about all these issues. More to be said, for example, about how the emphasis on minimizing indexicality varies across the sciences; about how virtually all the issues in the humanities are outgrowths of mentality but only a small segment of the issues in the sciences are; about philosophy’s relationships with the sciences on one hand and the rest of the humanities on the other; and about how the social sciences and arts fit in with all this. I take up these and other such issues in The Geography of Insight.

References Bloch, Marh. “Rituals and Deference.” In Rituals and Memory: Towards a Comparative Anthropology of Religion, edited by H. Whitehouse and J. Laidlaw. London: Altamira Press, 2004. Burke, Peter. The French Historical Revolution: The Annales School, 1929–​2014, 2nd ed. Palo Alto: Stanford University Press, 2015. Diamond, Jared. Guns, Germs, and Steel:  The Fates Human Societies. New  York:  Norton, 1997. Dworkin, Ronald. Law’s Empire. Cambridge: Harvard University Press, 1986. Dworkin, Ronald. Justice for Hedgehogs. Cambridge: Harvard University Press, 2011. Foley, Richard. The Geography of Insight: The Humanities, The Sciences, How They Differ, Why They Matter. New York: Oxford University Press, 2018. Gibbard, Alan. Wise Choices, Apt Feelings. Cambridge: Harvard University Press, 1990. Horwich, Paul. Wittgenstein’s Metaphilosophy. Oxford: Oxford University Press, 2012. Little, Daniel. “Philosophy of History.” The Stanford Encyclopedia of Philosophy, edited by Edward N. Zalta. Winter 2012 Edition. http://​plato.stanford.edu/​archives/​win2012/​ entries/​history. Nagel, Thomas. Moral Questions. Cambridge: Cambridge University Press, 1979. Small, Helen. The Value of the Humanities. Oxford: Oxford University Press, 2013. Williams, Bernard. Ethics and the Limits of Philosophy. Cambridge: Harvard University Press, 1985. Wittgenstein, Ludwig. The Blue and Brown Books. London: Blackwell, 1958.

4 On Literary Understanding Jennifer Gosetti-​Ferencei

Literary theorists for half a century have debated the problem of literary interpretation—​whether there can be a correct and satisfactory account of the meaning of a literary text, and how such an account may be validated.1 Meanwhile, philosophers of literature have addressed—​often with considerable skepticism—​whether literature can yield knowledge.2 Here I would like to take up a related, and perhaps more fundamental question, that of literary understanding. If we suppose that coming up with a satisfactory interpretation of a literary work, or drawing any significant knowledge from it, requires a reader’s having understood it, what does such understanding involve? Beyond the minimal requirements for basic comprehension, is there a dimension of understanding that is distinctly literary? Do our achievements in understanding literature exercise our understanding more generally, or aid our understanding of other, non-​literary matters? Several recent approaches to literature—​what I will describe as moral, aesthetic, and cognitive models of literary experience—​allow us to consider its relevance in epistemic terms. Through an examination of the insights and limits of these approaches, I will present the case for the experiential, generative, and expressive dimensions of understanding the literary work, and for their implications beyond literary reading. That literary understanding is experiential will mean that, beyond knowledge of what the text is about, one must have acquaintance with what it is like to undergo the imaginings prompted by the text. That literary understanding is generative means that what we understand in literary experience is not merely the objects or events 1 See for example E. D. Hirsch, Validity in Interpretation (New Haven: Yale University Press, 1967) and The Aims of Interpretation (Chicago: University of Chicago Press, 1976). 2 See Wolfgang Huemer, “Cognitive Dimensions of Achieving (and Failing) in Literature,” in Understanding Fiction: Knowledge and Meaning in Literature, ed. Jürgen Daiber, Eva-​Maria Konrad, Thomas Petraschka, and Hans Rott (Münster: Mentis Verlag, 2012), 26–​44. See also Peter Swirski, Of Literature and Knowledge: Explorations in Narrative Thought Experiments, Evolution, and Game Theory (New York: Routledge, 2006).

68  Jennifer Gosetti-Ferencei in the world from which the work may draw, but how these are transformed in the specific literary presentation created by the work. That literary understanding is expressive will mean that the object of understanding issues from, and brings us into contact with, a point of view, even if one known only through and as the work itself. These dimensions of literary understanding, I will suggest, are relevant beyond the experience of literature as such.

1.  From Literary Knowledge to Literary Understanding First, some preliminary qualifications are in order. “Literature” is sometimes taken to denote a broader category than the subject of discussion here. We will consider specifically the work of literary writing—​whether fiction, poetry, or some other genre—​as taken up in and considered through the minds of readers.3 By “literary,” I mean works the meaning and value of which are not determined by their correspondence to matters of fact, or by conveyance of literal truth, but through their evocation of an imagined experience for the reader. The debate about the relationship between literature and knowledge is older than Plato, who characterized literature as epistemically void, among other deficiencies metaphysical and moral. More recent epistemic skeptics of literature attack literature on several fronts. Bertrand Russell described literary works as at best cognitively trivial, at worse false, because they do not describe real objects.4 David Velleman has argued that narrative fails as a vehicle for knowledge insofar as its cadence imposes emotional transitions and a sense of closure that are typically mistaken for a causal-​explanatory process.5 Gregory Currie not only attacks the notion of literary knowledge as such—​literature in his view “militates against” effective reasoning and 3 On the history of reading as a mental process, see Paul Saegner, Space between Words: The Origins of Silent Reading (Stanford: Stanford University Press, 1997). See also Peter Kivy, The Performance of Reading (Hoboken, NJ: Wiley-​Blackwell, 2008), and “The Experience of Reading,” in A Companion to the Philosophy of Literature, ed. Garry L. Hagberg and Walter Jost (Oxford: Blackwell, 2010), 106–​19. On the significance of the written text for literary experience, see Christopher Collins, Neopoetics: On the Origins of the Literate Imagination (New  York:  Columbia University Press, 2016); Roman Ingarden, The Cognition of the Literary Work of Art, trans. Ruth Ann Crowley and Kenneth R. Olson (Evanston, IL: Northwestern University Press, 1973), 14; and Russell A. Berman, Fiction Sets You Free: Literature, Liberty, and Western Culture (Des Moines: University of Iowa Press, 2007), xvii. 4 Bertrand Russell, An Inquiry into Meaning and Truth (London:  Allen and Unwin, 1962), 277. Russell argues that “statements descriptive of Hamlet are all false because there was no such man,” 275. 5 David Velleman, “Narrative Explanation,” Philosophical Review 112 (2003): 1–​25.

On Literary Understanding  69 is the “opposite of knowledge-​enhancing”6—​but also rejects any association of literature with insight, learning, valuable acquaintance with others’ experiences, and the refinement of moral sensitivities.7 More weakly skeptical views may characterize knowledge associated with literature as inadequate in any number of ways. Even if literature deals in true beliefs, it does not offer justifications of them. Literary belief will extend only to belief in fiction, and so is not relevant beyond the work. If literary knowledge can be translated or paraphrased in propositions, the elements of literary form are unnecessary, and confounding. If the definition of knowledge can be expanded to allow, for instance, for vicarious experience, or “realizing by living through,” literature offers no guarantee that the experience has simulative veracity.8 Cautious advocates may grant that literature does not yield knowledge of the world, but of other objects of thought: knowledge of possibilities, and self-​ knowledge. Since literature relies on “a reader’s direct acquaintance with his or her mental states,”9 I may learn something about myself by my responses to literature’s imagined situations.10 These qualified allowances notwithstanding, the difficulties in this tradition may justify abandoning the search for literary knowledge in favor of some other cognitive or epistemic yield. For the view persists among readers, scholars, and some philosophers that literature can illuminate the full roundness as well as specificity of human experience, as it manifests both particularity and possibility.11 Many works of literary fiction are celebrated as forms of bearing witness. But witness to what, if their worlds are imaginary? The Holocaust writer Aharon Appelfeld explicitly defended his use of fiction—​even to skeptical fellow-​survivors who urged straightforward testimony—​as for him the only way to tell the effects 6 Gregory Currie, “Creativity and the Insight That Literature Brings,” in The Philosophy of Creativity: New Essays, ed. Elliot Samuel Paul and Scott Barry Kaufman (Oxford: Oxford University Press, 2014), 39–​61, here 51–​52. Currie there goes on to suggest that literature may be even epistemi­ cally deceptive, since the complexity of literature, as we expend effort in making sense of it, gives “an illusion of learning.” A literary work’s complexity, he claims, “may enhance its power to spread igno­ rance and error.” 7 Currie, “Creativity,” 44. 8 Dorothy Walsh supports the notion of virtual experience in Literature and Knowledge (Middletown, CT:  Wesleyan University Press, 1969), 101. For criticisms see Tilman Köppe, “On Making and Understanding Imaginative Experiences in Our Engagement with Fictional Narratives,” in Daiber et al., Understanding Fiction, 81–​95, here 91; Currie, “Creativity,” 44. 9 Kivy, “The Experience of Reading,” 106. 10 Robert Stecker claims that as readers we learn self-​knowledge through interaction with the text, in “Literature as Thought,” in Daiber et al., Understanding Fiction, 22. Köppe, in “On Making and Understanding Imaginative Experiences,” claims that we learn the phenomenal character of imagined experiences and “something about ourselves” in articulating them, both of which are moments of understanding, 93. 11 See Dick Attridge, The Singularity of Literature (New York: Routledge, 2004).

70  Jennifer Gosetti-Ferencei of what had happened to him in the camps. Only with fiction, he claimed, could he get beyond recounting events, and capture their human impact.12 Meanwhile, despite the challenges to the integrity of literary work in poststructuralism, literary works are nevertheless privileged objects of philosophical interpretation in Continental thought.13 In phenomenology and hermeneutics, literature and poetry has been associated with “revealing” reality, and thus with a form of truth.14 Several approaches to literature invite examination of literature’s relation to understanding. The moralist view asserts that literature depicts moral situations in a subtle and illuminating way, and through provoking empathy, exercises and refines responsive perception of others that can aid us in real-​ life moral contexts. Literary aesthetes, in contrast, argue that the achievement of literature is the generation of new kinds of experience irreducible to real-​worldly life. If literature aids our understanding of reality, it does so not by representing it, but through offering alternatives to it. Evidence for literature’s contribution to understanding in a broad sense is mounting from a cognitive perspective. In developmental psychology, it has been argued that imaginative fictions help children not to evade, but to cope with, reality.15 Cognitive theorists have begun to appreciate literature as cultivating the tools of cognition, from exercising the “theory” of other minds at multiple levels,16 to the kinds of problem-​solving and hypothetical thinking we use both in practical life and in science.17 Literary devices such as metaphor can be regarded as verbal affordances that help us develop complex thoughts,18

12 Aharon Appelfeld, Beyond Despair: Three Lectures and a Conversation with Philip Roth (New York: Fromm Intl, 1993). 13 Jacques Derrida, for example, ties literature to “the space of democratic freedom,” in On the Name, ed. Thomas Dutoit, trans. David Wood, John P. Leavey Jr., and Ian McLeod (Stanford: Stanford University Press, 1995), 28. 14 Martin Heidegger offers the notion of poetic truth as a form of revealing, in Poetry, Language, and Thought, trans. Albert Hofstadter (New  York:  Harper & Row, 1971). Hans-​Georg Gadamer articulates a hermeneutic variation on this position in “On the Contribution of Poetry to the Search for Truth,” in The Relevance of the Beautiful and Other Essays, ed. Robert Bernasconi (Cambridge: Cambridge University Press, 1986), 105–​15. Roger Scruton accepts, with some qualification, Heidegger’s notion of poetic revealing in “Poetry and Truth,” in The Philosophy of Poetry, ed. John Gibson (Oxford: Oxford University Press, 2015), 149–​61. 15 Paul Harris, The Work of Imagination: Understanding Children’s Worlds (Oxford: Blackwell, 2000). 16 Robin Dunbar, “Theory of Mind and the Evolution of Language,” in Approaches to the Evolution of Language:  Social and Cognitive Bases, ed. James R. Hurford et  al. (Cambridge:  Cambridge University Press, 1998), 92–​110; Lisa Zunshine, Why We Read Fiction: Theory of Mind and the Novel (Columbus: Ohio State University Press, 2006). 17 Swirski, Of Literature and Knowledge. 18 Terence Cave, Thinking with Literature: Towards a Cognitive Literary Theory (Oxford: Oxford University Press, 2016).

On Literary Understanding  71 or that reorient the knowledge we already have.19 Literature, it is argued, is a “technology” that helps us to reflect on language itself.20 Evolutionary anthropologists consider narrative and fiction as crucial to our cognitive and social evolution.21 Shifting our focus to understanding literature, it will not be sufficient to describe our grasp of a literary work’s content—​knowledge of what a novel, for instance, is about. Rather, understanding will require acquaintance with the experience literature evokes, a grasp of its formal and perspectival specificity. It may be that it is not in spite of, but because of, literature’s difference from reality, that literary understanding bears upon human understanding more broadly. But to get there we have to approach literature not only as a representative, but as a “transformative,” use of language, where the world is rendered, and so must be engaged, imaginatively.22 Understanding a literary work, of course, involves a grasp of meaning at several levels. Most basically, this requires knowledge of the literal meaning of the words and sentences, and further recognition of any figurative meanings evoked through such devices as metaphor. At a literal level the process of understanding may be, for the most part, unproblematic—​since, as Wittgenstein pointed out, words in literature mean the same as they do in life. In a literary work, however, how the subject matter is conveyed is essential to understanding what is conveyed. The understanding is not limited to its represented content—​its objects (characters, places, things), and events (action, plot)—​but includes also the perspective, or diverging perspectives, from which that content is presented. The reader is invited to take up, indeed to imagine, these objects and events from the perspective generated by the work as a whole, a perspective characterized through its formal, stylistic, and tonal qualities, its use of imagery and figurative language, its construction of repetitions and patterns. Some context—​knowledge of real-​worldly references, the conventions of genre, and of the era in which a work was written—​may supply important background. We can take as an initial example a passage from Virginia Woolf ’s novel Mrs. Dalloway. The novel opens on a fine day in London in June 1923. Having

19 Catherine Z. Elgin, “Art in the Advancement of Understanding,” American Philosophical Quarterly 39, no. 1 (2002): 1–​12. 20 Alva Noë, Strange Tools: Art and Human Nature (New York: Hill and Wang, 2015). 21 Brian Boyd, On the Origin of Stories: Evolution, Cognition, and Fiction (Cambridge, MA: Harvard University Press, 2010). 22 John Gibson, Fiction and the Weave of Life (Oxford: Oxford University Press, 2007), 131.

72  Jennifer Gosetti-Ferencei left the house to “buy the flowers herself,” for her party that evening, Clarissa sets off toward Bond Street, and hears the clock striking. For having lived in Westminster—​how many years now? over twenty,—​one feels even in the midst of the traffic, or walking at night, Clarissa was positive, a particular hush, or solemnity; an indescribable pause; a suspense (but that might be her heart, affected, they said, by influenza) before Big Ben strikes. There! Out it boomed. First a warning, musical; then the hour, irrevocable. The leaden circles dissolved in the air.23

The reader familiar with London will know that Big Ben is the bell of the clock on the tower of the Palace of Westminster, and so will infer Clarissa’s proximity, perhaps also symbolic, to that center of British establishment. But this both is, and is not, the London we can find on the map. Even the referential content—​Westminster, the clock—​is recognized through the imagining of a world generated under a particular perspective—​and in the case of the novel as a whole, through the interweaving of multiple perspectives. The narrative alignment in this passage with Clarissa’s point of view, the rolling accumulation of syntactical fragments, the variances of punctuation, the imagery, generate a cadence of thought and imagined movement in which the reader imaginatively participates. “There! Out it boomed,” for example, provokes responses in the reader that correspond to the imagined sound of a clock. The reader undertakes the activity of imagined listening (or as it has been called, “perceptual mimesis”)24 which conveys something both of Clarissa’s experience and of her surrounding world. There are further metaphorical suggestions. The suspense provoked by the clock striking is potentially confused by Clarissa with her heart affliction. As there are already hints in adjacent passages of Clarissa’s nostalgia for her youth, and her missing a former suitor about to come to town, the reference carries not only literal meaning but also metaphorical suggestion. The auditory resonances that follow each strike are described as “leaden circles.” There is an aesthetic element—​invisible sound being described in terms of a visual image—​and an emotional one, as lead evokes associations with heaviness,

23 Virginia Woolf, Mrs. Dalloway (London: Vintage, 2000), 4. 24 Elaine Scarry, “Imagining Flowers: Perceptual Mimesis (Particularly Delphinium),” Representa­ tions 57 (1997): 90–​115.

On Literary Understanding  73 and with coffins. The Londoners of Mrs. Dalloway, like the real Londoners of 1923, had, after all, just emerged from a devastating war. But there are other possibilities too. Woolf ’s language here may bring to mind another literary passage we can consider, the final quatrain from Emily Dickinson’s poem (372) “After Great Pain, a Formal Feeling Comes”: This is the Hour of Lead—​ Remembered, if outlived, As Freezing persons, recollect the Snow—​ First—​Chill—​then Stupor—​then the letting go—​25

Dickinson’s speaker’s association of “hour” and “lead,” and the staccato punctuation, seem to be echoed in Woolf ’s passage. The “solemnity” of Clarissa Dalloway’s anticipation of the striking clock, too, recalls the formal feeling described in Dickinson’s poem, for which “the nerves sit ceremonious, like tombs.” Just as Woolf ’s passage is not merely about Big Ben striking the hour on a day in June, Dickinson’s poem is not merely about the formal feeling that follows “great pain,” but generates an imagined embodiment of it. In coming to understand the passage from the novel or the poem, respectively, it will not be enough for the reader to already know what a clock sounds like, or what a painful experience feels like, and that the novel or poem refers to it. The reader will also undertake the specific phenomenal and emotional imaginings of the listening of Clarissa Dalloway, or the “formal feeling” Dickinson’s speaker evokes. This example has allowed us to introduce some elements of literary understanding. Grasping these passages and the larger works to which they belong will be a responsive undertaking. What we respond to is not merely a representation of reality, of a real London and person within it, but an imagined experience generated in and through the work, through linguistic, imagistic, and formal features, that contribute to revealing its subject matter from a particular perspective. The imagining it prompts, while drawing from elements of the real world, is not exhausted by that reference; but neither is the experience wholly otherworldly.

25 Emily Dickinson, The Poems of Emily Dickinson, ed. Ralph W. Franklin (Cambridge, MA: Belknap Press, 1998), no. 372.

74  Jennifer Gosetti-Ferencei

2.  The Moral Approach to Literary Understanding Our capacity for powerful responses to literature has been central to our very conception of it since Aristotle, and forms the basis of the moral approach. Appropriating Aristotle’s view of literary empathy, philosophers such as Martha Nussbaum have argued that moral understanding in life is not only analogous to that in literature, but finds therein “its most appropriate articulation.”26 Nussbaum develops a modern account based on the realist novel and its depiction of moral situations: for example, sacrifices made by a character on behalf of another, as depicted by Henry James in The Golden Bowl, the difficulties experienced by factory workers depicted by Charles Dickens in Hard Times, and racial oppression as suffered by African Americans in Richard Wright’s Native Son. Through such literature, she claims, we can learn to cultivate “a lucid apprehension” of others in a moral context.27 Similar claims for the ethical possibilities of realist fiction have been made by Jean-​Paul Sartre, Robert Pippin, and Wayne Booth.28 The moral apology for literature—​at least for some literature—​relies on the view that the narrative form, imagery, and other elements of literary description are essential to the meanings discovered by the characters and conveyed to the reader. Accordingly, the moral situation it illuminates could not be fully captured in paraphrase. Literature amplifies the “irreducibly particular character of a concrete moral context and the agents who are its components,” promotes a determination to “scrutinize” such particularity “with intensely focused perception,” and enjoins the reader “to care for it as a whole.”29 At the imaginative level, following characters’ relations with each other, and responding to them inwardly ourselves as readers, cultivates a kind of “know-​how,” for it requires “the ability to respond to and resourcefully interpret the particulars of . . . context.”30 Moral understanding here means a kind of responsive perception; it is “seeing a complex concrete reality in a

26 Martha Nussbaum, “Finely Aware and Richly Responsible: Moral Attention and the Moral Task of Literature,” Journal of Philosophy 82, no. 10 (1985): 516–​29, here 516. 27 Nussbaum, “Finely Aware,” 528. 28 Jean-​Paul Sartre, What Is Literature?, trans. Bernard Frechtman (New York: Routledge, 2001); Robert Pippin, Henry James and Modern Moral Life (Cambridge:  Cambridge University Press, 2001); Wayne Booth, The Company We Keep: An Ethics of Fiction (Berkeley: University of California Press, 1992). 29 Nussbaum, “Finely Aware,” 526. 30 Nussbaum, “Finely Aware,” 525.

On Literary Understanding  75 highly lucid and richly responsive way; it is taking in what is there, with imagination and feeling.”31 More recent arguments affirm the view that the feelings that are involved in literary empathy are relevant to understanding the work. Emotional responses provoked by literary stories, Jenefer Robinson writes, “can help us to understand them, to understand characters and grasp the significance of events in the plot . . . many of the emotions aroused by reading a novel are important—​even essential—​to a proper understanding of the novel.”32 The moral theorist’s view is that we feel emotions in literature on behalf of characters just as we feel them on behalf of others in real life, but in the context of cognitive suspension of our impulse to actively, rather than only inwardly, respond. The role of emotions in literary understanding has been acknowledged even by theorists who attempt to locate meaning within the boundaries of the text, rather than in the reader’s response to it. E.D. Hirsch, for example, having attempted to separate meaning from its extratextual significance, arrived at the view that meaning of a work itself cannot be restricted to “conceptual meaning.” For meaning “embraces not only a content of mind represented by written speech but also the affects and values that are necessarily correlative to such a content.” In fact, he argues, “one cannot have a meaning without its necessarily correlative affect or value” which is realized within the reader.33 Moral accounts of literary empathy rightly stress the experiential qualities of literary understanding. Yet those mentioned here all rely explicitly on the realistic portrayal of reality, while favoring literary works thought to achieve such realism. Sartre praises realistic prose over poetry for its straightforward language, through which reality can be transparently revealed. For Nussbaum and other philosophical interpreters, Henry James is the paradigmatic moral visionary because his writing is “thoroughly committed to the real”: “The realism in question is ‘internal’ and human; its raw material is human social experience. . . . But it is realism all the same.”34 One could object that this view leaves out a great deal of literature before the 19th-​century realist novel, and much of modernism as well.35 One could 31 Nussbaum, “Finely Aware,” 521. 32 Jenefer Robinson, “Emotion and the Understanding of Narrative,” in Hagberg and Jost, A Companion to the Philosophy of Literature, 71. 33 Hirsch, Aims of Interpretation, 8. 34 Nussbaum, “Finely Aware,” 528. 35 Ian Watt, The Rise of the Novel: Studies in Defoe, Richardson and Fielding (New York: Penguin, 1963), 32.

76  Jennifer Gosetti-Ferencei point out that earlier readers of Henry James, such as Kenneth Rexroth, complained that “there are no real people” in his novels,36 but caricatures inflected with class and other prejudice. Dickens was criticized even by his contemporaries for anti-​Semitic caricatures.37 A considerable amount of literary criticism in the last few decades has been devoted to exposing underlying prejudicial distortions in literary works, and their reflection of political or cultural oppression. Apart from whether the novel gets the reality “right,” we may question the underlying presumption about what a literary work is and does—​whether its value lies in illustrating life as we ordinarily live it. For even literary realism involves the “created illusion of a real referent.”38 This involves both selection and construction, for much that pertains to everyday human experience never appears in a realist novel, while much does appear that is never experienced in reality at all—​such as a narrator’s intimate knowledge of multiple other minds. Henry James himself recognized the creation of what he called “an air of reality” in his works, the implicit admission being that realism is itself a style, an artistic achievement.39 Moreover, coming to understand a work’s overall meaning, even its potential moral meaning, is not only episodic and immediate, as empathetic responses to characters and specific events can be, but also responsive to larger patterns that come through upon reflection on the world of the work as a whole. For example, the striking of the clock in Mrs. Dalloway is overheard by, and so connects, all of the major characters of the novel, who overhear it from different locations in London at different hours of the day. The reader is aware of an emerging pattern in ways the individual characters are not. The metaphoric suggestion of the “leaden circles” that follow the first striking of the clock is amplified when it is repeated toward the end of the novel in connection with the character Septimus Smith, a former soldier returned from the Great War who commits suicide early that evening. At her party Clarissa hears of the suicide of this man she does not know, and, as she later hears Big Ben strike out the hour, she directly compares herself to him:

36 Kenneth Rexroth, “Henry James and H. G. Wells,” in Assays (Norfolk: New Directions, 1961), 114–​17. 37 See Geoffrey Nunberg, The Way We Talk Now: Commentaries on Language and Culture (New York:  Houghton Mifflin Harcourt, 2001) 126. 38 Berman, Fiction,  6–​7. 39 Henry James, “The Art of Fiction,” Longman’s Magazine, September 1884, 4.

On Literary Understanding  77 She felt somehow very like him—​the young man who had killed himself. She felt glad that he had done it; thrown it away. The clock was striking. The leaden circles dissolved in the air.40

The repetition of the image of “leaden circles” may provoke the reader to notice how often thoughts of death and other forms of ruin have recurred despite Clarissa’s exuberance. This leads to an understanding of the situation of the novel as a whole as more precarious than the fine June weather and the social triumph of Clarissa’s party would betray. This understanding exceeds the emotional identification with a single character, and could not be won through realist depiction alone.

3.  The Aesthetic Approach to Literary Understanding While literary moralists hold up the understanding gained from realist literature as paradigmatic for the perceptive or emotive responsiveness we may need in living, the aesthetic approach resists confounding literature and life. By an aesthetic approach I mean the ontological or phenomenological identification of the literary work in terms of its constitution as a work of art. Such accounts invest in literary form, over and above content, as manifesting the work’s singularity or autonomy. In this vein, Dick Attridge claims that to defend the “singularity” of literature or “the specificity of literature and the uniqueness of the literary work is to raise the issue of form.”41 Russell Berman defends literature’s autonomy as “the capacity of the text to resist reduction to an external reality, while aspiring to an internal formal coherence.”42 Literature is regarded aesthetically when it is said, as John Gibson writes, to “give us a world, and . . . a unique object of appreciative and interpretive scrutiny” that is irreducible to life.43 The most radical formulation of this view is offered by Maurice Blanchot, who identifies the “space of literature” as severed from the real world: The book which has its origin in art has no guarantee in the world. . . . Hence the strange liberty of which reading—​literary reading—​gives us the prime

40 Woolf, Mrs. Dalloway, 186.

41 Attridge, The Singularity of Literature,  10–​11. 42 Berman, Fiction Makes You Free, xi–​xii.

43 Gibson, Fiction and the Weave of Life, 131.

78  Jennifer Gosetti-Ferencei example: a movement which is free insofar as it does not submit to, does not brace itself upon, anything already present.44

In Blanchot’s account, the “imaginariness” of literature is likened to an experience of nothingness, of the void. Yet while literary aesthetes would, in the main, not deny the validity of some use of literature for acquaintance with reality,45 they share in common an aim to undermine an instrumental view of literature as illustration of the world. Aesthetically approached, to understand literature is to grasp its generative nature, to experience it as the creation of new objects of thought. Influential for aesthetic models of literary understanding is the work of Roman Ingarden, for whom the reader’s consciousness participates in the “constitution of the polyphonic aesthetic value of the concretized work of art,” through engaging literature’s figurative and pictorial language, among other “aesthetically valent qualities.”46 For Ingarden, understanding the work issues from a complex process of constituting a unique object, comprised of several strata, from word sounds, to larger meaning units such as sentences and paragraphs, schematized aspects such as visual and other imagery, to represented entities and events. The work as a whole is a schematic formation full of indeterminacies that must be filled in by supplementation and inference on the part of the reader. Because of this indeterminacy, literary works, unlike scientific ones, allow for a variety of legitimate concretizations which may vary among readers, concretizations which thus do not map mimetically onto reality.47 Emphasizing the reader’s constitution of the literary work, Wolfgang Iser defends its irreducibility on aesthetic grounds. Literature has an “indefinable” quality, for “the meaning of a literary text is not a definable entity but, if anything, a dynamic happening.”48 The event of literature cannot be adequately described by stating what the work is about, or extracting its content:

44 Maurice Blanchot, The Space of Literature, trans. Ann Smock (Lincoln: University of Nebraska Press, 1982), 194–​95. 45 As Berman writes, “the experience of immediacy and vividness which we often gain from literary works of the past leads naturally to their being pressed into service as a source of evidence for lives led before ours or in foreign places,” Fiction Makes You Free,  6–​7. 46 Ingarden, Cognition of the Literary Work of Art, 152, 90. 47 Ingarden, Cognition of the Literary Work of Art, 164. 48 Wolfgang Iser, The Act of Reading: A Theory of Aesthetic Response (Baltimore: Johns Hopkins University Press, 1980), 22.

On Literary Understanding  79 The aesthetic effect is robbed of unique quality the moment one tries to define what is meant in terms of other meanings that one knows. For if it means nothing but what comes through it into the world, it cannot possibly be identical to anything already existing in the world.49

Literary signs therefore should not be understood as the reproduction of reality, but as the indications necessary for a production on the part of the reader. In Iser’s account, literary signs “fulfill their function to the degree in which their relatedness to identifiable objects begins to fade or is even blotted out.”50 Paul Ricoeur links this severance from reality to the generative quality of literature and its autonomy: “Because fictions don’t reproduce a previous reality, they may produce a new reality. They are not bound by an original that precedes them.”51 If the literary aesthete is right, however, that the literary work is severed from the world in this way, how does the literary work relate matters of the world we recognize, and thus merit our attention and interest? Iser acknowledges the mimetic function of literature by reference to what he calls the “repertoire,” the familiar territory within the text, including descriptions of recognizable objects we know from reality (Ingarden’s represented entities), reference to other literary works, to social and historical norms, to the culture and geography from which the text arises.52 We can thus acknowledge even on an aesthetic model the significance for Tolstoy’s novel War and Peace that Napoleon’s army in fact invaded Russia. The dire conditions of poor children in industrialized England are crucial to the fictional subject-​matter of Dickens’s Oliver Twist. The suicide of Septimus Smith in Mrs. Dalloway will be better grasped with the awareness that many soldiers returning from the Great War were traumatized by what became known as shell shock. The abolitionist movement in the United States remains essential to understanding the events and characters of Marilynne Robinson’s novel Gilead. Understanding even Kafka’s wholly unrealistic transformation of Gregor Samsa into a “horrible vermin” relies on our ordinary knowledge of species difference. More fundamentally, the experiential contours of space, time, and causality, although differently construed, condensed, and rearranged in literary reading, are inferentially constructed and

49 Iser, The Act of Reading, 65.

50 Iser, The Act of Reading,  65–​66.

51 Ricoeur, Lectures 14:9, see Taylor p. 97. 52 Iser, The Act of Reading, 82.

80  Jennifer Gosetti-Ferencei abstracted on the basis of pre-​literary experience. The repertoire, however faded, transformed or abstracted, services logistical and cultural inferences that allow the reader to make sense of the story. The aesthetic value of a literary work, however, will lie in the creative differences from reality it generates. Contrary to the literary moralist’s demand for (and presumption of) faithfulness to reality, the familiar territory of the text is for the aesthete “interesting not because it is familiar, but because it is to lead in an unfamiliar direction.”53 Literary mimesis on this model must be conceived more complexly than as straightforward representation of life.54 The transformation, rather than reproduction, of the familiar world repertoire, of the ordinary world’s sedimented expectations and conventions, can reveal by contrast, can serve indirectly as “a means of telling us about reality.”55 The world familiar from life can be seen in a new light, moreover, through the generated world of a literary work. This may occur through what Merleau-​Ponty called the “coherent deformation” of expressive language, where new meanings may be created through the distortion of ordinary habits of linguistic description.56 The Russian formalist Victor Shklovsky described this process within the arts in general as “defamiliarization.”57 As I have argued elsewhere, literature can help to illuminate how the familiar look of the world is constituted in the first place, and can rescue ordinary experience from the prejudices and dismissals of habit.58 Literary metaphor can reconfigure “a domain, drawing boundaries that cut across familiar distinctions, disclosing neglected aspects of the terrain, pointing thought in new directions.” Through such devices literature can engage possibility, promoting curiosity as well as discernment—​skills that lie at the root of all manner of human inquiry.59 Thus even in its aesthetic qualities, literature may be linked to the broader epistemic value reflected in the cognitive turn to literature.

53 Iser, The Act of Reading, 70. 54 See Jennifer Anna Gosetti-​Ferencei, “The Mimetic Dimension: Literature between Neuroscience and Phenomenology,” British Journal of Aesthetics 54, no. 4 (2014): 425–​48. 55 Iser, The Act of Reading, 55. 56 Maurice Merleau-​Ponty, “On the Phenomenology of Language,” in Signs, trans. R. C. McCleary (Evanston, IL: Northwestern University Press, 1964), 84–​97, here 91. 57 Viktor Shklovsky, “Art as Technique,” in Russian Formalist Criticism: Four Essays, trans. Lee T. Lemon and Marion J. Reis (Lincoln: University of Nebraska Press, 1965), 4. 58 See Jennifer Anna Gosetti-​Ferencei, The Ecstatic Quotidian:  Phenomenological Sightings in Modern Art and Literature (University Park: Penn State University Press, 2007). 59 Elgin, “Art in the Advancement of Understanding,” 4.

On Literary Understanding  81

4.  The Cognitive Approach to Literary Understanding The transformative or generative capacity of literature has also been appreciated in the “cognitive turn” in literary studies of the last decade. In the main, cognitive approaches tend to reject “the view of literature as belonging to a special (and consequently marginal) aesthetic domain,”60 examining instead ways in which literature reflects and fosters human thinking more broadly. Versions I would call empirical literary cognitivism aim to locate literature within the project of identifying its (often strictly limited) support for empirical or objective knowledge,61 or approach literary cognition as an object of empirical, often neuroscientific, study.62 Because these approaches do not focus on the experience of literature, I will set them aside in order to discuss briefly what I call literary literary cognitivism. This view recognizes an indirect role for literature in understanding, and the importance of specifically literary means through which such understanding is fostered. Such literary cognitivism may suggest that literature reshapes the mind or reconfigures existing knowledge;63 enables us to reflect on the nature of language itself;64 or uniquely evidences the embodied nature of cognition.65 Most cognitivists would accept that what literary works are about is part of what makes them distinctively literary. Yet most relevant from the cognitive standpoint is not what knowledge literature may offer about the world, but the ways thought-​experiences fostered by literary reading can shape the mind of the reader. Peter Lamarque, for example, describes how great literature can “reshape our minds or change our inner landscape in a positive manner . . . through bringing thoughts to mind.” What matters is not the content but “how we acquire those thoughts and how they can in subtle ways reorder our minds.”66 The process of having and reflecting on literary experiences exercises particular cognitive skills, as we construct, according to the work, 60 Cave, Thinking with Literature,  3–​4. 61 Swirski, Of Literature and Knowledge. 62 See Gerhard Lauer, “Going Empirical:  Why We Need Cognitive Literary Studies,” Journal of Literary Theory 3, no. 1 (2009): 145–​54; G. Gabrielle Starr, Feeling Beauty: The Neuroscience of Aesthetic Experience (Cambridge, MA: MIT Press, 2013); Semir Zeki, “Neural Concept Formation and Art: Dante, Michelangelo, Wagner,” Journal of Consciousness Studies 9, no. 3 (2002): 53–​76. 63 Peter Lamarque, “Thought Theory and Literary Cognition,” in Daiber et  al., Understanding Fiction; Elgin, “Art in the Advancement.” 64 This would include Heidegger, Poetry, Language, Thought; Gadamer, “On the Contribution of Poetry,”; and Noë, Strange Tools. 65 Cave, Thinking with Literature; Guillemette Bolens, The Style of Gestures:  Embodiment and Cognition in Literary Narrative (Baltimore: Johns Hopkins University Press, 2016). 66 Lamarque, “Thought Theory,” 69.

82  Jennifer Gosetti-Ferencei a perspective by which the particulars of the content are considered. To understand a literary work is on this view not to extract generalizable truths, but to grasp the “fictional world” within the work itself as “an end in itself.”67 Since what we “do” with the thus accumulated thoughts will vary among different readers, Lamarque stops short of endorsing a concrete cognitive yield, refusing “to read back” into the works themselves “any such variable, reader-​ relative, and instrumental cognitive gains.”68 A reshaping of cognition through literature may be primarily perceptual, as described in Rainer Maria Rilke’s novel The Notebooks of Malte Laurids Brigge, in which the protagonist’s study of modernist poetry and impressionist paintings demonstrably transforms his perception of the Paris streets and the faces of city dwellers around him.69 Literature may also provoke forms of introspection which may be likened to phenomenological insight. Woolf credited modern fiction with illuminating the processes of thinking as “the mind receives a myriad of impressions—​trivial, fantastic, evanescent, or engraved with the sharpness of steel,” arriving incessantly from all sides, while the “accent falls” variously according to the vagaries of attention.70 Literature may present hypotheses that force reconsideration of established cognitive or existential habits. Rilke’s well-​known poem “Archaic Torso of Apollo” describes an encounter with the fragment of an ancient Greek statue in the Louvre. The speaker begins the poem by invoking the first-​person plural: “We cannot know his legendary head.” The poem’s well-​ known ending—​“There is no place /​that does not see you. You must change your life”—​shifts to a second-​person imperative. Through syntactical as well as imagistic means, Rilke has so structured the work that the imperative applies equally for the reader of the poem as it does the (imagined) viewer of the statue. Whether this provocation prompts self-​examination for any given reader is another matter; but in coming to understand the poem, the possibility is unavoidable as a thought-​experience. The reshaping of thought may also involve coming to imagine experiences different from one’s own, or to regard a familiar experience from a new reflective standpoint. Claudia Rankine’s Citizen:  An American Lyric, for example, constructs a series of experiential reflections on race that invites both 67 Lamarque, “Thought Theory,” 78. 68 Lamarque, “Thought Theory,” 79–​80. 69 Gosetti-​Ferencei, The Ecstatic Quotidian, 110–​11. 70 Virginia Woolf, The Essays of Virginia Woolf, vol. 4:  1925 to 1928, ed. Andrew McNeille (London: Hogarth Press, 1984), 160.

On Literary Understanding  83 personal and political responses. One passage describes a moment of reflection that invokes epistemic scrutiny on the part of the speaker: The rain this morning pours from the gutters and everywhere else it is lost in the trees. You need your glasses to single out what you know is there because doubt is inexorable; you put on your glasses. The trees, their bark and leaves, even the dead ones, are more vibrant wet. Yes, and it’s raining. Each moment is like this—​before it can be known, categorized as similar to another thing and dismissed, it has to be experienced, it has to be seen. What did he just say? Did she really just say that? Did I hear what I think I heard?71

The reader recognizes from earlier passages the subtext of the questions, namely the experience of the speaker’s being the object of everyday, injurious racism. Yet the ambiguity in the description serves a cognitive and emotional purpose. At a literal level the speaker describes the experience of looking, a visual scene coming into focus, while at a metaphorical level it describes the process of trying to understand a social interaction. Without having yet looked closely, the speaker knows what is out there from the sound and a blurry visual perception, from the most proximal evidence. Our cognitive habits allow us to register familiar objects or statements—​“what you know is there”—​even if only partially perceived or perceived out of context. But in such categorization, necessary for our negotiation of the world, we may not see the particulars clearly, or under the aspect of a particular weather or perspective. We may need to adjust our vision to look more closely, as the speaker does putting on her glasses. As in Rilke’s poem, though more implicitly, the scrutiny the speaker demands of herself is asked of the reader. Although there is ambiguity here, it may describe the effort to counter habitual categorization that is characteristic of racial prejudice itself, and the effect of generalizations on the particular. The thought-​experience in this and surrounding passages shapes a distinctively simultaneous scrutiny and validation of the speaker’s (and potentially readers’) experiential knowledge. But this uncertainty may be different for readers familiar with the experiences Rankine describes than for readers who have not suffered what her speaker calls “the stresses of racism.” There is something here of what Sartre recognized in Richard Wright’s work as the genius of “double

71 Claudia Rankine, Citizen: An American Lyric (New York: Penguin, 2015) 7.

84  Jennifer Gosetti-Ferencei simultaneous postulation”: with the same words Wright could speak differently to readers familiar and unfamiliar with the described experiences, and “go beyond this split” to make a work of art.72 If coming to understand such literary works reshapes our minds through the thought-​experiences they provoke, this will have effects that will not be confined to the world of the work alone. The exercises of imagined perception, introspection, existential, epistemic, or other form of scrutiny a literary work may prompt are likely to have cognitive and other effects that overflow the boundaries of the work itself. So far this would appear to support the literary moralist’s claim for the direct relation between moral understanding in literature and that in life. Yet this relationship may be due not to the achievement of realism in literary content, even when the work aims to convey, and even where it does convey, an accurate representation of life as it is really experienced. For the assumption behind the moralist’s realism is that reality itself is available transparently to observation, and the representation of that reality need only be equal to it in order for it to affect the reader in edifying ways. There are, of course, objective facts to back up concern about racism in America, and many ways to convey those facts; this is undoubtedly the necessary context of Richard Wright’s Native Son as of Rankine’s Citizen. Yet literary works do not only record facts—​which are available in many other formats—​but generate frameworks of reflection—​ new possibilities for interpreting that reality, and for making features of it salient. Rankine’s work exposes reality not as a fixed given that determines our experience in universal ways, but as one the perspectival experience of which is important, and shaped by perception and misperception, categorization and miscategorization, familiarity and unfamiliarity. One of the larger themes of Citizen as a whole, for example, is that elements of reality, such as racially motivated violence, are glimpsed by some Americans only as headline news, while for others they persistently shadow and afflict everyday life. If Lamarque limits the relevance of cognition to the work itself, others refer such effect to its medium. Such a view is offered by Alva Noë, who in Strange Tools describes literature as a cognitive “technology” for reflection on the activity of writing itself, and more generally on language. The cognitive benefit of literature is, by making language “strange,” to reveal the ways we organize and filter experience through language. On this view, which echoes both Heidegger and his student Gadamer, Rankine’s poetry could be

72 Sartre, What is Literature?, 80.

On Literary Understanding  85 understood as a tool to help us think about the role of language, here specifically in racial categorization or oppression. Certainly that could be an important theme, but with reference to language alone, we could not account for the embodied and affective dimensions of the experiences she describes, and the ways in which the reader is provoked to imagine them. Analogously, if we understood Rilke’s poem as a tool to help us think about the nature of language, we could not adequately register the cognitive or existential impact of its provocation to the reader, even as that is enabled by particular syntactic moves. One problem with the technology theory—​and Noë’s particular twist on Heidegger’s theory of language as both estranging and revealing—​is that it would not be able to capture the depth and persistence of our interest in literature, and the breadth of its achievements. Surely literature draws from, transforms, and allows us to reflect on many things besides the medium of language itself. Written literature does not merely make writing or language strange, but restlessly redraws and reconfigures the world and life as we (think we) know it. Kafka offers a powerful example of such reconfiguration in The Metamorphosis. When Gregor Samsa woke up one morning from unsettling dreams, he found himself changed in his bed into a monstrous vermin. He was lying on his back as hard as armor plate, and when he lifted his head a little, he saw his brown vaulted belly, sectioned by arch-​shaped ribs, to whose dome the cover, about to slide off completely, could barely cling. His many legs, pitifully thin compared to the size of the rest of him, were waving helplessly before his eyes. ‘What’s happened to me?’ he thought. It was no dream.73

Kafka offers here a literalization of a metaphor—​though we are never exactly sure what Gregor is, he wakes up as an insect-​like parasite, a vermin. By the novella’s end, having sustained an injury at the hands of his father and lost his appetite, Gregor dies and is swept up and tossed away with the garbage. Perhaps the reader aware of metaphors such as that one can work like an insect, or live like a parasite, will use Kafka’s story to reflect on the nature of language.74 But Gregor has been working to pay off the debts of

73 Franz Kafka, The Metamorphosis, trans. and ed. Stanley Corngold (New York: Bantam, 1972). 74 See Corngold’s commentary in Kafka, The Metamorphosis, 63.

86  Jennifer Gosetti-Ferencei his father and support the family who then imprison, hide, and kill him. By literally confounding categories in the figure of Gregor-​turned-​vermin, the story makes strange the thematic subjects of the dissolved distinction:  humanness—​or the human worker and family member—​and more primitive forms of animal life. What is it about a human being that can be insect-​like, and why does Kafka situate such devolution in the domestic family? The thought-​experience of coming to understand Kafka’s work will urge this sort of interpretive question, by means of a literalized metaphor that prompts hypothetical imagining, where presumed divisions between evolved aspects of ourselves and a primordial substrate are dissolved. While many cognitivists regard literature as a tool or a technology,75 there are more biosensitive metaphors. Terence Cave proposes a cognitive criticism that would regard the literary work as part of our “cognitive ecology,” reflecting the “constantly shifting engagement between cognition and environment as things loom into focus, demand attention, fade, or are discarded as irrelevant.”76 Literature, although a special kind of language and an art, is not a world, or as Blanchot would say, a space, removed from the ways in which human beings livingly exist. To consider literature as an artifact of this ecology contextualizes it in the embodied, living human context in which we communicate. Why do we read poems or stories hundreds of years after they were written? Perhaps, Cave suggests, “not because they contain certain kinds of knowledge, or philosophical wisdom, but because we infer from them the continually remade presence of a living, embodied mind.”77 Literary evidence of mind, however, is not just of general and perhaps universal features of human cognition, as the literary cognitivist tends to describe, but almost always evidence of a unique and particularly situated human mind. Literary understanding, in addition to being experiential and generative, is also expressive, arising from a particular possible point of view, irrespective of any debates about access to an authorial source. It was mentioned earlier in this essay that among the kinds of knowledge even cautious cognitivists would claim literature provides, are knowledge of possibility and self-​knowledge, as the reader responds to literary imaginings.78 75 Swirski calls literature the “cognitive Swiss army knife” because it can perform lots of tasks, Of Literature and Knowledge, 10. 76 Cave, Thinking with Literature, 6, 5. 77 Cave, Thinking with Literature, 10. 78 Stecker, “Literature as Thought,” 22; Köppe, in “On Making and Understanding Imaginative Experiences,” 93.

On Literary Understanding  87 Conspicuously missing from the cognitive approach to literature thus far is any serious consideration of the fact that literature yields belief in, and justification for, the fact that there are other perspectives on the world which transcend our own situatedness. Our human interest in getting to know, from the inside, creative expressions of others may be a central motivation for literary reading. “Theory of mind” interpretations of literature are close to this insight, highlighting how literature engages skills of recognizing intentionality—​ thinking that others have thoughts, knowing that there are other minds. On this view we can find in complex literature levels of intentionality, such that we can imagine other characters imagining the thoughts or intentions of still others, thus exercising cognitive skills crucial to our social evolution.79 But literature also evokes concrete experiences that exceed the reader’s own life and thought, experiences that are intrinsically “other,” ones which by virtue of our own rootedness and specificity, we could not have in reality or generate from our own point of view. In reading Mrs. Dalloway, or The Metamorphosis, or the poetry of Dickinson, Rilke, or Rankine, we come to be intimate with the productions of other human imaginings. These works reflect both the repertoire in reality from which they draw, and the specific, even singular transcendence of that repertoire in the creative thought of the literary writer. Yet theories of literature over the last half-​century have neglected this possible knowledge in avoidance of what Monroe Beardsley called the “intentional fallacy”—​the mistake of confusing the meaning and content of the work for the author’s intention. New Criticism sought an objective knowledge of the literary work by dismissing the relevance of its personal and historical origins. Poststructuralism dismissed the author as the source of the literary work in favor of textual systems. Reader-​response theory validated the reading process as the source of meaning, since it is, unlike the authorial mind, accessible to analysis. But we need presume nothing about the author’s own intentions to respect, from the cognitive perspective, a literary work’s evidence of creative human thinking and the specific point of view formed by and shaped within the literary work. Coming to understand a literary work entails engaging this perspective, inevitably that of another point of view than our own.



79 Dunbar, “Theory of Mind”; Zunshine, Why We Read Fiction.

88  Jennifer Gosetti-Ferencei

5.  Conclusion We have characterized three major approaches to literature, our consideration of which has yielded important dimensions of literary understanding. Through examining the moral approach we have seen the experiential nature of literary understanding. Forms of involvement, such as empathy or taking up a particular point of view, are essential to grasping the work and its meaning. These activities relevant to moral life—​though more indirectly than advocates of the moral view tend to insist—​whether or not the content of a literary work can be said to accurately reflect reality. In advocating, with the aesthetic approach, that the generative nature of the literary work renders it irreducible to reproduced reality, we do not thereby abandon its relevance to real-​worldly experience. To understand a literary work is to grasp its transformation of any material it borrows from the world, and its creation of a new framework or perspective from which its subject matter can be regarded. Literary works can help us to reflect on reality by way of contrast to it. Cognitive approaches include the suggestion that literature reshapes and allows us to reflect on our forms of thinking. The literary work expresses a larger cognitive ecology, offering evidence, not of authorial intention, but of intentionality, or of the expressive nature of human creative thought. Literature’s expressiveness allows readers to revivify the thoughtful imaginings of the text, and to engage other ways of thinking than their own. These qualities of literary understanding, in their moral, aesthetic, and cognitive dimensions, are undoubtedly relevant beyond our understanding of any given work of literature. They may, further, attest to the significance of imaginative endeavors for our life more generally.80

References Appelfeld, Aharon. Beyond Despair: Three Lectures and a Conversation with Philip Roth. New York: Fromm Intl, 1993. Attridge, Dick. The Singularity of Literature. New York: Routledge, 2004. Berman, Russell A. Fiction Sets You Free: Literature, Liberty, and Western Culture. Des Moines: University of Iowa Press, 2007. Blanchot, Maurice. The Space of Literature. Translated by Ann Smock. Lincoln: University of Nebraska Press, 1982.

80 See Jennifer Anna Gosetti-​Ferencei, The Life of Imagination: Revealing and Making the World (New York: Columbia University Press, 2018).

On Literary Understanding  89 Bolens, Guillemette. The Style of Gestures:  Embodiment and Cognition in Literary Narrative. Baltimore: John Hopkins University Press, 2016. Booth, Wayne. The Company We Keep:  An Ethics of Fiction. Berkeley:  University of California Press, 1992. Boyd, Brian. On the Origin of Stories:  Evolution, Cognition, and Fiction. Cambridge, MA: Harvard University Press, 2010. Cave, Terence. Thinking with Literature:  Towards a Cognitive Literary Theory. Oxford:  Oxford University Press, 2016. Collins, Christopher. Neopoetics: On the Origins of the Literate Imagination. New York:  Columbia University Press, 2016. Currie, Gregory. “Creativity and the Insight That Literature Brings.” In The Philosophy of Creativity: New Essays, edited by Elliot Samuel Paul and Scott Barry Kaufman, 39–​61. Oxford: Oxford University Press, 2014. Derrida, Jacques. On the Name. Edited by Thomas Dutoit. Translated by David Wood, John P. Leavey Jr., and Ian McLeod. Stanford: Stanford University Press, 1995. Dickinson, Emily. The Poems of Emily Dickinson. Edited by Ralph W. Franklin. Cambridge, MA: Belknap Press, 1998. Dunbar, Robin. “Theory of Mind and the Evolution of Language.” In Approaches to the Evolution of Language: Social and Cognitive Bases, edited by James R. Hurford, Michael Studdert-​Kennedy, and Chris Knight, 92–​110. Cambridge:  Cambridge University Press, 1998. Elgin, Catherine Z. “Art in the Advancement of Understanding.” American Philosophical Quarterly 39, no. 1 (2002): 1–​12. Gadamer, Hans-​Georg. The Relevance of the Beautiful and Other Essays. Edited by Robert Bernasconi. Cambridge: Cambridge University Press, 1986. Gibson, John. Fiction and the Weave of Life. Oxford: Oxford University Press, 2007. Gosetti-​Ferencei, Jennifer Anna. The Ecstatic Quotidian: Phenomenological Sightings in Modern Art and Literature. University Park: Penn State University Press, 2007. Gosetti-​ Ferencei, Jennifer Anna. “The Mimetic Dimension:  Literature between Neuroscience and Phenomenology.” British Journal of Aesthetics 54, no. 4 (2014): 425–​48. Gosetti-​Ferencei, Jennifer Anna. The Life of Imagination: Revealing and Making the World. New York: Columbia University Press, 2018. Harris, Paul. The Work of Imagination:  Understanding Children’s Worlds. Oxford:  Blackwell, 2000. Heidegger, Martin. Poetry, Language, and Thought. Translated by Albert Hofstadter. New York: Harper & Row, 1971. Hirsch, E. D. Validity in Interpretation. New Haven: Yale University Press, 1967. Hirsch, E. D. The Aims of Interpretation. Chicago: University of Chicago Press, 1976. Huemer, Wolfgang. “Cognitive Dimensions of Achieving (and Failing) in Literature.” In Understanding Fiction: Knowledge and Meaning in Literature, edited by Jürgen Daiber, Eva-​Maria Konrad, Thomas Petraschka, and Hans Rott, 26–​44. Münster:  Mentis Verlag, 2012. Ingarden, Roman. The Cognition of the Literary Work of Art. Translated by Ruth Ann Crowley and Kenneth R. Olson. Evanston, IL: Northwestern University Press, 1973. Iser, Wolfgang. The Act of Reading:  A Theory of Aesthetic Response. Baltimore:  John Hopkins University Press, 1980. James, Henry. “The Art of Fiction.” Longman’s Magazine, September 1884.

90  Jennifer Gosetti-Ferencei Kafka, Franz. The Metamorphosis. Translated and edited by Stanley Corngold. New York:  Bantam, 1972. Kivy, Peter. The Performance of Reading. Hoboken, NJ: Wiley-​Blackwell, 2008. Kivy, Peter. “The Experience of Reading.” In A Companion to the Philosophy of Literature, edited by Garry L. Hagberg and Walter Jost, 106–​19. Oxford: Blackwell, 2010. Köppe, Tilman. “On Making and Understanding Imaginative Experiences in Our Engagement with Fictional Narratives.” In Understanding Fiction:  Knowledge and Meaning in Literature, edited by Jürgen Daiber, Eva-​Maria Konrad, Thomas Petraschka, and Hans Rott, 81–​95. Münster: Mentis Verlag, 2012. Lamarque, Peter. “Thought Theory and Literary Cognition.” In Understanding Fiction:  Knowledge and Meaning in Literature, edited by Jürgen Daiber, Eva-​Maria Konrad, Thomas Petraschka, and Hans Rott. Münster: Mentis Verlag, 2012. Lauer, Gerhard. “Going Empirical: Why We Need Cognitive Literary Studies.” Journal of Literary Theory 3, no. 1 (2009): 145–​54. Merleau-​Ponty, Maurice. Signs. Translated by R. C. McCleary. Evanston, IL: Northwestern University Press, 1964. Noë, Alva. Strange Tools: Art and Human Nature. New York: Hill and Wang, 2015. Nunberg, Geoffrey. The Way We Talk Now:  Commentaries on Language and Culture. New York: Houghton Mifflin Harcourt, 2001. Nussbaum, Martha. “Finely Aware and Richly Responsible:  Moral Attention and the Moral Task of Literature.” Journal of Philosophy 82, no. 10 (1985): 516–​29. Pippin, Robert. Henry James and Modern Moral Life. Cambridge: Cambridge University Press, 2001. Rankine, Claudia. Citizen: An American Lyric. New York: Penguin, 2015. Rexroth, Kenneth. “Henry James and H. G. Wells.” In Assays. Norfolk: New Directions, 1961, 114–​17. Robinson, Jenefer. “Emotion and the Understanding of Narrative.” In A Companion to the Philosophy of Literature, edited by Garry L. Hagberg and Walter Jost, 69–​92. Oxford: Blackwell, 2010. Russell, Bertrand. An Inquiry into Meaning and Truth. London: Allen and Unwin, 1962. Saegner, Paul. Space between Words: The Origins of Silent Reading. Stanford, CA: Stanford University Press, 1997. Sartre, Jean-​Paul. What is Literature? New York: Routledge, 2001. Scarry, Elaine. “Imagining Flowers:  Perceptual Mimesis (Particularly Delphinium).” Representations 57 (1997): 90–​115. Scruton, Roger. “Poetry and Truth.” In The Philosophy of Poetry, edited by John Gibson, 149–​61. Oxford: Oxford University Press, 2015. Shklovsky, Viktor. “Art as Technique.” In Russian Formalist Criticism:  Four Essays, translated by Lee T. Lemon and Marion J. Reis. Lincoln:  University of Nebraska Press, 1965. Starr, G. Gabrielle. Feeling Beauty: The Neuroscience of Aesthetic Experience. Cambridge, MA: MIT Press, 2013. Strecker, Robert. “Literature as Thought.” In Understanding Fiction:  Knowledge and Meaning in Literature, edited by Jürgen Daiber, Eva-​Maria Konrad, Thomas Petraschka, and Hans Rott, 22. Münster: Mentis Verlag, 2012. Swirski, Peter. Of Literature and Knowledge:  Explorations in Narrative Thought Experiments, Evolution, and Game Theory. New York: Routledge, 2006. Velleman, David. “Narrative Explanation.” Philosophical Review 112 (2003): 1–​25.

On Literary Understanding  91 Walsh, Dorothy. Literature and Knowledge. Middletown, CT:  Wesleyan University Press, 1969. Watt, Ian. The Rise of the Novel:  Studies in Defoe, Richardson and Fielding. New  York:  Penguin, 1963. Woolf, Virginia. The Essays of Virginia Woolf. Vol. 4: 1925 to 1928. Edited by Andrew McNeille. London: Hogarth Press, 1984. Woolf, Virginia. Mrs. Dalloway. London: Vintage, 2000. Zeki, Semir. “Neural Concept Formation and Art: Dante, Michelangelo, Wagner.” Journal of Consciousness Studies 9, no. 3 (2002): 53–​76. Zunshine, Lisa. Why We Read Fiction: Theory of Mind and the Novel. Columbus: Ohio State University Press, 2006.

5 Recasting the “Scientism” Debate Anthony Gottlieb

Here is a piece of table talk from an Oxford college in the early 1920s. To appreciate it, you need to know two items of Oxford terminology. “Greats” is a course comprising ancient Greek, Latin, history, and philosophy; and a “First” is the highest mark you can get in a British bachelor’s degree. The story goes that an eminent physicist, Frederick Lindemann, who had recently arrived in Oxford, was seated at dinner next to Margaret Pember, the wife of the head of the college. Lindemann rued the lack of satisfactory scientific education in England, whereupon Mrs. Pember is said to have remarked: “You needn’t worry. Any man who has got a First in Greats could get up science in a fortnight.”1 Perhaps Mrs. Pember learned to change her tune a few years later when she acquired a scientist (and grandson of Darwin) as a son-​in-​law. Be that as it may, you will not often hear science belittled in Mrs. Pember’s way now. It has in many places acquired the sort of prestige that classics once enjoyed. Indeed, some humanists feel that the tables have turned and that today’s scientists are prone to a prejudice which is the opposite of Mrs. Pember’s. Surely a first-​class physicist or neuroscientist could get up the humanities in a fortnight? It sometimes seems from popular-​science books and news reports that anyone who is au fait with the latest scientific work can get up philosophy in a fortnight. For example, one hears that after centuries of debate, a few laboratory experiments have shown that there is no such thing as free will. And we have it on the authority of Stephen Hawking that “philosophy is dead” because it “has not kept up with modern developments in science.”2 Some other recent American and British physicists have written similar things.

1 Roy Harrod, The Prof: A Personal Memoir of Lord Cherwell (London: Macmillan, 1959), 53. 2 Stephen Hawking and Leonard Mlodinow, The Grand Design (New  York:  Bantam Books, 2010), 5.

94  Anthony Gottlieb It is notable, though, that you do not find such attitudes in the writings of Einstein or Heisenberg, who were educated in a culture that encouraged one to spend more than a fortnight on philosophy. Every specialist will have a favorite gripe about an interloper who has blundered uncomprehendingly onto his or her turf. Academics seem to enjoy making such complaints and it is hard to imagine a world in which the cultivators of one field of knowledge do not engage in territorial disputes with those toiling in others. Such disputes break out within the sciences and within the humanities just as they do between these two families of disciplines. The sort of gripe I am going to consider here is a less excusable and also a more philosophically interesting one. It is a general complaint about the natural sciences—​I shall not be discussing social science—​and it arises from the idea that science is by its nature limited in scope. Precisely because it is science, it can tell us only so much and not more, or so the complaint goes. I want to suggest that such a stance makes less sense than one might think. This becomes evident, I believe, once we ask exactly what we mean by “science.” The trouble is that people tend not to ask that question, or do not press it hard enough. Arguments about the limitations of science tend to focus on particular shortcomings of present-​day work, and then jump to a conclusion about all possible science, as if today’s methods, concepts, and results define the enterprise for all time. Debates about the limits of science often involve charges of “scientism.” This term, in its current pejorative sense of an exaggerated reverence for science,3 or an unreasonable belief in its powers, is no older than the late nineteenth century. But the theme itself, or something very like it, goes back at least as far as Plato. In two of Plato’s dialogues, his mouthpiece in effect levels a charge of scientism against the investigators of nature. In the Phaedo, Socrates recounts how he was disappointed to find that Anaxagoras “adduced causes like air and aether and water and many other absurdities.”4 What was missing, according to Socrates, was an account that explained how some form of intelligence “sets everything in order and arranges each individual thing in the way that is best for it.”5 3 For an instance of this use of the term, see Tom Sorrell, Scientism (New York: Routledge, 1991), x. 4 Plato, “Phaedo,” in The Collected Dialogues of Plato, trans. Hugh Tredennick, ed. Edith Hamilton and Huntington Cairns (Princeton, NJ: Princeton University Press, 1961), 98c. 5 Plato, “Phaedo,” 97c (79).

Recasting the “Scientism” Debate  95 Similarly, in his last work, the Laws, Plato has the Athenian Stranger speak of “the unreason and error of all who have ever busied themselves with research into nature.”6 Their main mistake, according to the Stranger, was to focus on physical qualities such as “hard and soft, heavy and light.”7 This approach was allegedly atheistical, incomplete, and probably also immoral. Echoing Plato, conservative thinkers in the seventeenth century complained that the so-​called “mechanical philosophy” of Galileo did not pay enough heed to the immaterial realm or to the notion of purposeful order in the universe. Thus Leibniz wrote that he wanted to reconcile “the mechanical philosophy of the moderns with the caution of some intelligent and well-​ intentioned persons who fear . . . we are withdrawing too far from immaterial beings, to the disadvantage of piety.”8 Despite Leibniz’s mention of piety, and the fact that Plato associated natural science with atheism, we should beware of thinking that the charge of scientism in the old days was all about religion. That might get things the wrong way around, at least in some cases. Maybe some people thought: “Today’s physics is too simplistic. So perhaps we need more about God in the picture,” rather than: “We need more about God in the picture. So today’s physics is too simplistic.” Allegations of scientism have in more recent times certainly not been leveled only for religious reasons. In the late nineteenth century, when Matthew Arnold and T. H. Huxley debated the proper place of “physical science” in education, and in culture, Arnold invoked the venerable idea that there is something morally suspect in paying too much attention to the investigation of nature. Arnold’s suggestion was that “humane letters” served “the paramount desire in men that good should be forever present to them.” One limitation of “physical science,” according to him, is that it gives us facts that do not “put us into relation with our sense for conduct.”9 For this reason and others, Arnold concluded that an education focused on the physical sciences, rather than the humanities, would be incomplete. And if one had to make a choice between the two, it would be better for most people if they picked the humanities. (The exceptions were people who showed special

6 Plato, The Laws, trans. A. E. Taylor (London: J.M. Dent & Sons, 1934), 891c (1446–​47). 7 Plato, The Laws, 892b (1447). 8 G. W. Leibniz, “Discourse on Metaphysics” [1686], in Philosophical Essays, ed. and trans. Roger Ariew and Daniel Garber (Indianapolis: Hackett, 1989), 18 (52). 9 Matthew Arnold, “Literature and Science” (Rede Lecture at Cambridge University, 1882), Nineteenth Century, August 12, 1882.

96  Anthony Gottlieb talent in the natural sciences. In such cases, according to Arnold, it was fair enough to focus on nature.) Huxley’s position was that balance was needed. He aimed to undermine the idea that culture is nurtured only by a traditional schooling in the liberal arts. But he conceded that a purely scientific education would “bring about a mental twist as surely as an exclusive literary training.”10 Huxley’s notion of a scientific education, by the way, was broader than ours. He thought that a knowledge of French and German (especially German) was an essential part of it. The exchange between Arnold and Huxley was polite and respectful. The same cannot be said of the next notable skirmish on this topic in the English-​ speaking world. This was the debate that followed a famous lecture in 1959 about “The Two Cultures,” by C. P. Snow, a British champion of science and technology who was also a novelist. Snow described a gulf of incomprehension that separated “literary intellectuals” and “natural scientists.” Like Huxley, he was arguing for balance. Three decades after Lindemann’s encounter with Mrs. Pember, Snow felt that British society still needed to take science more seriously. For some literary intellectuals, that was asking too much. Snow suggested that literary intellectuals were “natural Luddites” in their attitudes to science,11 and one of them proved him right about this. The literary critic F. R. Leavis contemptuously dismissed Snow and his arguments. Snow, Leavis wrote, was as “intellectually undistinguished as it is possible to be.”12 Leavis also poured scorn on the idea that science can solve all our problems. Snow had not said that it could, but Leavis presumably felt that some such belief must lie behind any plea for equal treatment for the humanities and the sciences. A few years ago, the spat between Snow and Leavis was played out again in the pages of the New Republic. Leon Wieseltier, then the magazine’s literary editor, convincingly reprised the role of Leavis, and Steven Pinker largely took the part of Snow. Pinker’s thesis was that science has the potential “to enrich and diversify the intellectual tools of humanistic scholarship.”13 10 T. H. Huxley, “Science and Culture” (address at the opening of Mason College, Birmingham, 1880), in Essays: English and American (New York: P.F. Collier & Son), 1909–​14. 11 C. P. Snow, The Two Cultures:  A Second Look (Cambridge:  Cambridge University Press, 1993), 22. 12 See Stefan Collini’s introduction to Snow, The Two Cultures, xxxiii. 13 Steven Pinker, “Science Is Not Your Enemy,” New Republic, August 6, 2013:  https://​ newrepublic.com/​article/​114127/​science-​not-​enemy-​humanities.

Recasting the “Scientism” Debate  97 In fact, he claimed that it had already done so. He wrote that “Intellectual problems from antiquity are being illuminated by insights from the sciences of mind, brain, genes, and evolution.” Pinker conceded that many attempted applications of neuroscience and genetics to human affairs have been “glib or wrong, and . . . are fair game for criticism.” But, he wrote, “It is a mistake to use a few wrongheaded examples as an excuse to quarantine the sciences of human nature from our attempt to understand the human condition.” This last point is surely reasonable. Even if, say, every purported contribution of neuroscience to the understanding of the arts had been disappointing hitherto, it would be reckless to discount the idea that it might shed some light in the future. Enlightenment has a habit of emerging from unexpected places and of taking longer than expected to arrive. On the other hand, it is hard to agree with Pinker that the humanities have already been illuminated by studies of brains or genes. He does not seem to have any persuasive examples of this, though perhaps it all depends on what you mean by illumination. It is even harder to agree with the claim, made in a widely used textbook, that an evolutionary approach to psychology is already “beginning to transform” the study of the arts and religion.14 Premature pronouncements such as this may be prompted by what one might call the “stops-​at-​the-​neck” fallacy. Enthusiasts for evolutionary psychology sometimes diagnose resistance to their work as amounting to the idea that human evolution stops at the neck. Surely, the thinking goes, our minds are shaped by evolution just as our bodies are, so we should seek to understand our culture with the tools of evolutionary theory. But that inference is fallacious: particle physics does not stop at the neck any more than evolution does, yet nobody thinks that there must therefore be pertinent and informative explanations of culture or psychology that can be couched in terms of particle physics. Accounts of culture and psychology must, of course, not be inconsistent with evolutionary theory, just as they must not be inconsistent with particle physics. This does not entail that they can be expressed in terms of evolutionary theory or particle physics, or that these latter disciplines will necessarily illuminate them.15 To be clear:  I am not arguing that evolutionary or neurological work cannot shed any explanatory light on art or religion. I am suggesting that we

14 David Buss, Evolutionary Psychology (Boston: Allyn & Bacon, 2012), 428. 15 This point is well made by Elliott Sober in Philosophy of Biology, 2nd ed. (Boulder, CO: Westview, 2000), 189.

98  Anthony Gottlieb resolve to evaluate any such purported explanations on their merits, and not assume either that they will never be found or that they must by now be just around the corner. Some champions of science are too quick to credit today’s work on genes or brains with implications that it has not yet proved itself to have. In fairness, we should note that humanists have also been known to exaggerate the significance of their own work now and then. Wieseltier, in his responses to Pinker, maintained that the “swaggering scientists”16 are not merely too quick off the mark but are competing in the wrong stadium. He is, as he put it, “for a two-​state solution” to what he sees as a conflict between the sciences and the humanities. There is, he claimed, “a basis in reality” for the traditionally established borders between the two—​ a “momentous distinction between the study of the natural world and the study of the human world.”17 This gulf between the two worlds is apparently something to do with subjectivity. What the “swaggering scientists” supposedly cannot incorporate into their work is “the irreducible reality of inwardness, and its autonomy as a category of understanding.” Art and literature, on the other hand, can provide an “exploration of subjectivity and what is lived.” The idea here seems to be that there is a citadel of consciousness which is impregnable either to all science or, in a weaker form of the thesis, to present-​ day science. This theme appears in the influential work of the philosopher Thomas Nagel, though without the heated rhetoric of disciplinary border disputes. Nagel has argued that “The physical sciences . . . cannot describe the subjective experience of . . . organisms or how the world appears to their different particular points of view.” Purely physical descriptions of neurophysiology and behavior will, he says, “leave out the subjective essence of the experience—​how it is from the point of view of its subject—​without which it would not be a conscious experience at all.”18 Socrates wanted science to tell him more about purpose in nature. Arnold wanted to hear more about moral values. Today people want more about consciousness and subjective experience.

16 Steven Pinker and Leon Wieseltier, “Science vs. the Humanities, Round III,” New Republic, September 26, 2013: https://​newrepublic.com/​article/​114754/​steven-​pinker-​leon-​wieseltier-​debate​science-​vs-​humanities. 17 Leon Wieseltier, “Crimes against Humanities,” New Republic, September 3, 2013:  https://​ newrepublic.com/​article/​114548/​leon-​wieseltier-​responds-​steven-​pinkers-​scientism. 18 Tom Nagel, “The Core of ‘Mind and Cosmos,’” New  York Times, The Stone, August 18, 2013: http://​opinionator.blogs.nytimes.com/​2013/​08/​18/​the-​core-​of-​mind-​and-​cosmos/​. Reprinted in The Stone Reader: Modern Philosophy in 133 Arguments, ed. Peter Catapano and Simon Critchley (New York: Liveright, 2016), 233–​35.

Recasting the “Scientism” Debate  99 There is a difference, though, between Nagel’s position and Wieseltier’s. Nagel denies only that the methods of today’s sciences can capture the subjectivity of experience. He does not say that no science ever could, as Wieseltier appears to believe. In fact, Nagel suggests that science might eventually remedy this defect, by widening its horizons in radical ways: [A]‌scientific understanding of nature need not be limited to a physical theory of the objective spatiotemporal order. It makes sense to seek an expanded form of understanding that includes the mental but that is still scientific—​that is, is still a theory of the immanent order of nature.19

This invites the question:  what exactly makes a form of understanding “scientific?” I think Nagel is right to use the term in an elastic way—​that is, to envision a body of knowledge about nature that is significantly different from what we have today and yet might still qualify as science. This is because science seems to be an approbative rather than a purely descriptive concept. The term “science” is an honorific. When we look at the history of its employment, we see that what uses of the word and its ancestors have in common is not that they refer to a particular subject area or a particular set of methods. The term has in general been used to mark whatever was thought at the time to be the best sort of theoretical knowledge. It is fairly well-​known that “scientist” is a relatively new word. Although the English term was coined in the 1830s, many of the best-​known British “scientists,” as we now call them, still resisted it in the latter half of the nineteenth century.20 Kelvin and Faraday openly opposed the coinage and Darwin never used it. It did not become widely accepted in Britain until after the Second World War. The word “science” has been around for much longer, and I want to draw attention to the ways in which its meaning has evolved. Unlike “scientist,” which still means much the same as it did when it was coined, the meaning of “science” and its cognates is very different from what it was when the French translation of the Latin scientia first entered English in the Middle Ages.21 19 Nagel, “Core of Mind and Cosmos,” 235. 20 See Sydney Ross, “Scientist: The Story of a Word,” Annals of Science 18, no. 2 (1962), 65–​85. 21 I  am indebted to Robert Pasnau’s work on scientia. For one example, see “Science and Certainty,” in The Cambridge History of Medieval Philosophy, ed. R. Pasnau and Christina van Dyke (Cambridge: Cambridge University Press, 2014), 1:357–​69.

100  Anthony Gottlieb Most Latin authors, including Cicero, used scientia rather broadly, to cover understanding in general, or to refer to any corpus of knowledge. But philosophical writers in late antiquity, and in the Middle Ages, often used it in a more narrow and demanding sense, and it is this sense which is the root of our notion of science. This narrow sense derived from what Plato and Aristotle had said about the highest grade of epistēmē, or knowledge. Thus Augustine, in the late fourth century, wrote that I don’t call anything scientia where the person who professes it is sometimes mistaken. Scientia doesn’t consist merely in the matters that are apprehended. Instead, it consists in the fact that they are apprehended in such a way that nobody should be in error about it.22

Similarly, Aquinas—​ writing nearly 900  years later—​ held that a person who has scientia about something “knows that it is impossible for it to be otherwise.”23 Aquinas wrote those words in his commentary on Aristotle’s Posterior Analytics, in which Aristotle had given an account of the model or ideal form of organized knowledge. There is still some debate about what Aristotle was trying to do in his Posterior Analytics, and how best to translate some of his terminology. But it is safe to say that Aristotle’s model form of knowledge was characterized by mathematical-​style demonstrations proceeding from indubitable or self-​evident principles.24 Hence the notion that if you have scientia of something, you cannot be wrong about it. This idea was still around four centuries after Aquinas, when Descartes wrote that “no act of awareness that can be rendered doubtful seems fit to be called scientia.”25 Another medieval commentary on Aristotle went so far as to say that “only in mathematics is there scientia . . . in the strictest sense.”26 Biology was, in medieval times, usually held to fall far beneath the standards of scientia, 22 Augustine, Against the Academicians and The Teacher [Contra Academicos], trans. Peter King (Indianapolis: Hackett, 1995), I.7.19. Translation altered to include the original Latin scientia. 23 Aquinas, Expositio libri Posteriorum Analyticorum, trans. Fabian Larcher, II.20:  http://​ dhspriory.org/​thomas/​PostAnalytica.htm#220. Translation altered to include the original Latin scientia. 24 Aristotle, Posterior Analytics, Book 1, Ch. 2. 25 Descartes, “Meditations,” in Philosophical Writings of Descartes, trans. John Cottingham, Robert Stoothoff, and Dugald Murdoch (Cambridge: Cambridge University Press, 1984), 2:101. Translation altered to include the original Latin scientia. 26 Grosseteste, “Commentary on Posterior Analytics,” in The History of Science from Augustine to Galileo, trans. A. C. Crombie (New York: Dover, 1959), I.xi (31). Translation altered to include the original Latin scientia.

Recasting the “Scientism” Debate  101 because the truths it discovered were thought to hold only for the most part. It did not seem to yield exceptionless laws. Theology, on the other hand, was widely regarded as a genuine science. A few people worried that it did not quite qualify as scientia, because its first principles were accepted as articles of faith; but these people were in a minority. In the fourteenth century, some thinkers, including Ockham and Buridan, pointed out that weaker forms of knowledge were really quite useful. They meant the sort of knowledge that is gained by experience, rather than via mathematical-​style demonstration, and which is therefore merely probable rather than absolutely certain. But Ockham and Buridan were ahead of their time. Even in the seventeenth century, when such empirical knowledge was becoming all the rage, John Locke refused to call it “science.” Locke still used the term only in the old, technical sense of scientia. That is why he wrote that he suspected that “natural philosophy [by which he meant what we call physics and chemistry] is not capable of being made into a science.”27 However much progress we make in studying physical things, according to Locke, “scientifical” knowledge of them would still be “out of our reach.”28 Ethics, on the other hand, was capable of being turned into a science, because in Locke’s opinion, some ethical truths could be demonstrated deductively. It was for partly similar reasons that Hobbes had regarded politics as a perfect example of a science. In fact, politics was for Hobbes one of only two genuine examples of science, the other being geometry.29 The entry for science in the French Enlightenment’s Encyclopédie still defined it in terms of scientia: Science, as a philosophical concept, means the clear and certain knowledge of something, whether founded on self-​evident principles, or via systematic demonstration.30

But the term and its equivalents in other European languages were certainly also being used more broadly at the time. Samuel Johnson’s 27 Locke, Essay, IV.12.10. 28 Locke, Essay, IV.3.26. 29 Hobbes, “Six Lessons to the Professors of Mathematics” (1656), Epistle Dedicatory, in English Works of Thomas Hobbes of Malmesbury, ed. W. Molesworth (London: Longman, Brown, Green, and Longmans, 1845), 7:183–​84. 30 “Science” [1765], in The Encyclopedia of Diderot & d’Alembert Collaborative Translation Project, trans. Michele Pridmore-​Brown (Ann Arbor:  Michigan Publishing, 2003):  http://​hdl.handle.net/​ 2027/​spo.did2222.0000.195.

102  Anthony Gottlieb Dictionary, for example, listed five senses of “science,” including “Art attained by precepts, or built on principles” and “Any art or species of knowledge.”31 A simple way to describe what happened as the traditional concept of scientia became displaced by our modern notion of science is that indubitability and deductive demonstration were replaced as desiderata of the best sort of theoretical knowledge by the notions of empirical testing. Instead of saying, with Augustine and Descartes, that the best type of knowl­ edge is the sort that cannot possibly be wrong, we now say that the scientific type of knowledge is the sort that we have put most rigorously to the test. That is one key component of recent definitions of science. Another component, which is rather vaguer, is the idea that this knowledge is organized in a systematic way.32 The usage of the term “science” narrowed, and began to settle into more or less what we have today, only just after the middle of the nineteenth century. Up to that point, it had commonly been applied to all sorts of organized or well-​founded bodies of knowledge, or skills. The subject matter of this knowl­ edge was neither here nor there. Theology and rhetoric, for example, were still sciences. And in Pride and Prejudice, you may recall, Mr. Darcy was said to be adept at “the science of dancing.”33 We hear this as a joke, but it was much less of one in the time of Jane Austen. By the 1860s, “science” was largely reserved for what were described as “physical” and “experimental” investigations. Chemistry and physics were the paradigms of this, and mathematics was included, too. Matthew Arnold, John Ruskin, and others pleaded against this tide. Here is Ruskin, from a series of lectures given in Oxford in 1872: [I]‌t has become the permitted fashion among modern mathematicians, chemists, and apothecaries, to call themselves “scientific men,” as opposed to theologians, poets, and artists. They know their sphere to be a separate one; but their ridiculous notion of its being a peculiarly scientific one ought not to be allowed in our Universities.34 31 Samuel Johnson’s Dictionary, ed. Jack Lynch (Delray Beach: Levenger Press, 2004), 455. 32 See Peter Godfrey-​Smith, Theory and Reality (Chicago: University of Chicago Press, 2003), 71. 33 Jane Austen, Pride and Prejudice (1813), Ch. 6. 34 John Ruskin, Ariadne Florentina: Six Lectures on Wood and Metal Engraving, 1872, quoted in Ross, “Scientist,” 70.

Recasting the “Scientism” Debate  103 Ruskin complained that ethics, history, grammar, music, and painting were sciences just as much as chemistry was. Similarly, Matthew Arnold declared that “all learning is scientific which is systematically laid out and followed up to its original sources . . . a genuine humanism is scientific.”35 I am not going to suggest that we should belatedly follow Ruskin’s lead and lobby universities to move their artists and theologians into the science buildings. Arnold and Ruskin lost that battle and there is no going back. And I would not go so far as to call our present arrangements “ridiculous.” But perhaps reflecting on the way that the development of the concept of science has been intertwined with changing ideas about the best type of knowledge can free up our thinking about scientism. It may encourage us to look askance at the notion of intrinsic limits to the sciences. For if science is just what we call the best kind of theoretical knowledge, and physics, chemistry, and biology are paradigms of it because of their rigor, not because of their subject matter, then what sense is there in supposing that “scientific” methods can take us only so far? It may be said that what sets significant limits to science, given our current conception of it, is precisely its stress on rigorous testing, since some topics are not amenable to such treatment. For example, ethical considerations limit our experimentation on humans, so there may be some things we shall never discover about them in that way. There are also practical considerations that constrain the type and amount of testing we can do. It may well be that no researchers will ever replicate Irene Pepperberg’s work on the linguistic abilities of parrots, because they will not be willing to devote so much of their lives to talking with birds. If so, then some of her results will never enter the canon of scientific knowledge even if they are correct.36 And the complex social context of human action may well set de facto limits to what we can learn about it scientifically (and in particular to what we can learn about free will), because it may never be possible to construct sufficiently sensitive and controlled experiments in this area. If someone were to argue that topics in the humanities tend to be too complex to be subjected to the sort of testing and generalization to which today’s natural sciences aspire, I would have no quarrel with them. It is, though, perhaps worth pointing out that this line of thought cannot provide the basis 35 Arnold, “Literature and Science,” 22. 36 Irene Maxine Pepperberg, The Alex Studies:  Cognitive and Communicative Abilities of Grey Parrots (Cambridge, MA: Harvard University Press, 2002).

104  Anthony Gottlieb for a distinction between the sciences and the humanities, or between the natural and social sciences. For one thing, the natural sciences suffer from the same limitations, at least sometimes. The complexity of living bodies may, for instance, be such that there will always be large mysteries in medicine: perhaps we shall never be able to discover enough to predict and control our health in the way that medicine would like to be able to do. And Irene Pepperberg is, after all, a psychologist. But limitations imposed by the complexity of phenomena, and by the demands of gathering sufficient data, are merely contingent anyway. For any given difficulty of this sort, it is surely not impossible that ways to overcome them will eventually be found. So they do not seem capable of grounding Wieseltier’s “momentous distinction between the study of the natural world and the study of the human world.” Those, like Wieseltier, who believe in a “two-​state solution” evidently think that there is some deeper divide between the realm of the sciences and the realm of the humanities. As I noted, this divide is commonly thought to have something to do with subjectivity. In a recent book, Roger Scruton put the matter as follows: “the science of human biology . . . sees us as objects rather than subjects, and its descriptions of our responses are not descriptions of what we feel.”37 Expanding on this, Scruton puts weight on the notion of an individual perspective, which, it is alleged, science necessarily omits: As a self-​conscious subject I have a point of view on the world. The world seems a certain way to me, and this “seeming” defines my unique perspective. . . . When I give a scientific account of the world, however, I am describing objects only. I am describing the way things are and the causal laws that govern them. This description is given from no particular perspective. It does not contain words such as here, now, and I; and while it is meant to explain the way things seem, it does so by giving a theory of how they are. In short, the subject is in principle unobservable to science, not because it exists in another realm but because it is not part of the empirical world. It lies on the edge of things, like a horizon, and could never be grasped “from the other side,” the side of subjectivity itself.38

37 Roger Scruton, On Human Nature (Princeton, NJ: Princeton University Press, 2017), 46.

38 Scruton, On Human Nature, 32.

Recasting the “Scientism” Debate  105 But in what sense does biology not see people as subjects? Scruton must intend biology to include empirical psychology, and empirical psychology plainly does deal with persons as feeling and thinking beings—​i.e., as subjects of experience. It tries to explain why they think and feel as they do. In the course of their professional activities, doctors and psychologists have, after all, regularly been known to ask people how they feel. As for individual perspectives, scientific accounts do indeed aspire to explain things objectively—​i.e., “from no particular perspective.” But one thing they try to explain is perspective itself. They aim to account for the fact that what you see depends on where you stand, both literally and figuratively. So what, exactly, is supposed to be missing? A famous thought experiment by the philosopher Frank Jackson may help here.39 We are to imagine a scientist, Mary, whose field is color perception. Thanks to her assiduous research, she knows everything that science can currently teach us about colors and how we see them. Oddly, she has never left her windowless house, which does not have a television or computer and which contains only black, white, and gray objects. Even her body is painted in monochrome. Now suppose that one day she ventures outdoors and sees the colors of the rainbow for the first time. It seems natural to say that she thereby learns something about color that she did not know before, even though she was already an expert in the science of color. Some treatments of this thought experiment argue that what Mary learns when she steps outside is not a set of new facts but a set of new skills.40 She acquires, for example, the ability to distinguish the colors of objects just by looking at them. Leaving aside the question whether she learns new facts or new skills (or both), let us agree that the liberated Mary is newly acquainted with what Nagel calls “the subjective essence” of color. She now knows at firsthand what it feels like to see colors, and it seems that this is something which science could not tell her while she was confined to her bizarre house. Does the case of Mary exemplify the sort of limit to science that we have been looking for? 39 Frank Jackson, “Epiphenomenal Qualia,” Philosophical Quarterly 32 (1982), 127–​36. See also Jackson’s “What Mary Didn’t Know,” Journal of Philosophy 83 (1986), 291–​95. 40 See David Lewis, “What Experience Teaches” (1988), reprinted in Papers in Metaphysics and Epistemology (Cambridge:  Cambridge University Press, 1999), 262–​90; D. H. Mellor, “Nothing Like Experience” (1993), reprinted in Mind, Meaning and Reality (Oxford: Oxford University Press, 2012),  10–​21.

106  Anthony Gottlieb I think not. Consider some ways in which it might be said that science cannot tell you something: (1) Science cannot tell you how many pages there are in my edition of War and Peace. (2) Science cannot tell you how to ride a bicycle. (3) Science cannot tell you what it feels like to eat an egg sandwich. Nobody would take the first example to count as a substantive limitation of science. To ask a scientist for professional help with this question would not even be a case of using a sledgehammer to crack a nut: no corpus of scientific knowledge is even relevant to the matter at hand. All that is needed in order to answer the question is access to my copy of the book, plus the ability to read or count. The second example is roughly similar: scientific knowledge is not what is needed to tell you what you want to know. Any normal person who can ride a bicycle will be able to show and explain to a novice how to do it (though it is not impossible that research in physiology will one day lead to new tips for everyday riding, as it perhaps already does for racing). Now for the third example. What ought we to expect science to be able to tell us about the experience of eating an egg sandwich? Quite a lot. We would like physiology and psychology to explain why egg sandwiches do not, to a normal person, generally smell of fish or taste of cotton wool or feel like stones. That is, we expect an account of how and why eating an egg sandwich produces the sensations that it does and not other sensations. Ideally, we would also like scientists to understand the mechanisms of sensation well enough to let them help someone with a defective sensory apparatus to experience egg sandwiches in the way that others do, by mending that apparatus. And what if someone with normal taste buds wanted to be told what it feels like to eat an egg sandwich? A scientist could tell him a few things about what it is like and what it is not like, but the most effective answer would be to advise him to go to a sandwich shop. If he eats an egg sandwich, he will know, and the question of what it is like to eat one will have been resolved for him. Thus the third example is like the first one: getting the best answer is primarily a matter of access—​i.e., of being in the right position to learn what you want to know. Similarly, if Mary wonders what it feels like to see colors, all she has to do is step outside. She will then be acquainted with the “subjective essence” of color which Nagel believes that science leaves out.

Recasting the “Scientism” Debate  107 The complaint that scientific descriptions of experiences do not capture their subjective essences seems to amount to a demand that they produce some sort of verbal substitute for firsthand experience—​i.e., a description of an experience which is such that anyone who reads it will come to know what it is like to have the experience, in the way that the liberated Mary comes to know what it is like to see colors. And why should one expect science to be able to do that? Perhaps it will be said that literature can provide a sort of verbal substitute for firsthand experience, by stimulating us to imagine experiences which we have not had, so that the humanities may thus be said to do what science cannot. No doubt they can, but I am of course not out to establish that science can take the place of poetry, music, and the other arts. I am suggesting that it is no part of the remit of science to recreate subjective experiences on the page. I hope I have cast some doubt on the idea of a citadel of consciousness that cannot be breached by science. I suggest that we have not yet been given good reason to believe that it is impervious to the advances of present-​day science, as Nagel believes, let alone to those of any future science. And the overall contention of this chapter is that it is not clear how any topic in the domain of theoretical knowledge could be judged to be beyond the scope of scientific illumination. No doubt inflated claims will continue to be made about particular branches of science and about particular scientific results, as they will about much else. But that is another matter.

Acknowledgments This essay is an extended version of a talk read to the “Varieties of Understanding” conference at Fordham University on June 22, 2016.

References Aquinas. Expositio libri Posteriorum Analyticorum. Translated by Fabian Larcher, II.20. http://​dhspriory.org/​thomas/​PostAnalytica.htm#220 Arnold, Matthew. “Literature and Science.” Rede Lecture at Cambridge University, August 12, 1882. Augustine. Against the Academicians, and The Teacher [Contra Academicos]. Translated by Peter King. Indianapolis: Hackett, 1995. Buss, David. Evolutionary Psychology. Boston: Allyn & Bacon, 2012.

108  Anthony Gottlieb Descartes, René. “Meditations.” In Philosophical Writings of Descartes, vol. 2, translated by John Cottingham, Robert Stoothoff, and Dugald Murdoch. Cambridge: Cambridge University Press, 1984. Godfrey-​Smith, Peter. Theory and Reality. Chicago: University of Chicago Press, 2003. Grosseteste, Robert. “Commentary on Posterior Analytics.” In The History of Science from Augustine to Galileo, translated by A. C. Crombie. New York: Dover, 1959. Harrod, Roy. The Prof: A Personal Memoir of Lord Cherwell. London: Macmillan, 1959. Hawking, Stephen, and Leonard Mlodinow. The Grand Design. New  York:  Bantam Books, 2010. Hobbes, Thomas. “Six Lessons to the Professors of Mathematics” [1656]. In English Works of Thomas Hobbes of Malmesbury, vol. 7, edited by W. Molesworth. London: Longman, Brown, Green, and Longmans, 1845. Huxley, T. H. “Science and Culture.” Address at the opening of Mason College, Birmingham, 1880. In Essays:  English and American, 1909–​ 14. New  York:  P.F. Collier & Son. Jackson, Frank. “Epiphenomenal Qualia.” Philosophical Quarterly 32 (1982), 127–​36. Jackson, Frank. “What Mary Didn’t Know.” Journal of Philosophy 83 (1986), 291–​95. Lewis, David. “What Experience Teaches” [1988]. Reprinted in Papers in Metaphysics and Epistemology. Cambridge: Cambridge University Press, 1999, 262–​90. Leibniz, G. W. “Discourse on Metaphysics” [1686]. In Philosophical Essays, edited and translated by Roger Ariew and Daniel Garber. Indianapolis: Hackett, 1989. Mellor, D. H. “Nothing Like Experience” [1993]. Reprinted in Mind, Meaning and Reality. Oxford: Oxford University Press, 2012, 10–​21. Nagel, Thomas. “The Core of ‘Mind and Cosmos.’” The Stone, New York Times, August 18, 2013. http://​opinionator.blogs.nytimes.com/​2013/​08/​18/​the-​core-​of-​mind-​and-​ cosmos/​. Reprinted in The Stone Reader:  Modern Philosophy in 133 Arguments, ed. Peter Catapano and Simon Critchley (New York: Liveright, 2016), 233–​35. Pasnau, Robert. “Science and Certainty.” In The Cambridge History of Medieval Philosophy, vol. 1, edited by R. Pasnau and Christina van Dyke, 357–​69. Cambridge: Cambridge University Press, 2014. Pepperberg, Irene Maxine. The Alex Studies:  Cognitive and Communicative Abilities of Grey Parrots. Cambridge, MA: Harvard University Press, 2002. Pinker, Steven. “Science Is Not Your Enemy.” New Republic, August 6, 2013. Pinker, Steven, and Leon Wieseltier. “Science vs. the Humanities, Round III.” New Republic, September 26, 2013. Plato. “Phaedo.” In The Collected Dialogues of Plato, translated by Hugh Tredennick, edited by Edith Hamilton and Huntington Cairns. Princeton, NJ: Princeton University Press, 1961. Plato. The Laws. Translated by A. E. Taylor. London: J.M. Dent & Sons, 1934. Ross, Sydney. “Scientist: The Story of a Word.” Annals of Science 18, no. 2 (1962), 65–​85. Samuel Johnson’s Dictionary. Edited by Jack Lynch. Delray Beach: Levenger Press, 2004. “Science” [1765]. In The Encyclopedia of Diderot & d’Alembert Collaborative Translation Project, translated by Michele Pridmore-​Brown (Ann Arbor:  Michigan Publishing, 2003). http://​hdl.handle.net/​2027/​spo.did2222.0000.195 Scruton, Roger. On Human Nature. Princeton, NJ: Princeton University Press, 2017. Snow, C. P. The Two Cultures:  A Second Look. Cambridge:  Cambridge University Press, 1993. Sober, Elliott. Philosophy of Biology. 2nd ed. Boulder, CO: Westview, 2000. Sorrell, Tom. Scientism. New York: Routledge, 1991. Wieseltier, Leon. “Crimes against Humanities.” New Republic, September 3, 2013.

6 Firsthand Knowledge and Understanding Ernest Sosa

What follows aims to enhance our understanding of the notions of firsthand knowledge and of understanding, of how these are related, and of their importance in a flourishing human life. On certain questions of great human interest, firsthand knowledge and firsthand understanding are closely interrelated and have high priority. Such questions are often met in the humanities, broadly conceived to include not only appreciation of art, but also appreciation of sports, food, relationships, nature, and much more. And many such questions are to be found in philosophy. All such humanistic questions stand in contrast with practical questions where mere information suffices. What follows is devoted to explaining and defending these ideas. 1. Says Aristotle: It is possible to do something that is in accordance with the laws of grammar, either by chance or at the suggestion of another. A man will be a grammarian, then, only when he has both done something grammatical and done it grammatically; and this means doing it in accordance with the grammatical knowledge in himself. (Nicomachean Ethics II 4, 1105a22-​6)

An utterance can be in accordance with the laws of grammar in that the uttered sentence does not violate proper grammar. This is a success, if the speaker meant to speak grammatically. That success is “by chance,” however, if it fails to be properly in accordance with the speaker’s grammatical competence. In order to be “in accordance with” that competence, the utterance must not only accord with grammar, but must also do so under the guidance of the speaker’s competence, so that the success is not just “by chance.” A monkey at a keyboard may happen to produce a grammatical string, which would not be a success in accordance with the monkey’s (nonexistent)

110  Ernest Sosa grammatical knowledge. And even someone with such knowledge may succeed not because of it, but only because the grammatical string is whispered in their ear. Here again the success will not be due to the grammatical knowl­ edge of the speaker, even if it is not entirely “by chance.” How crucial such a concept of aptness is to Aristotle’s ethics may be seen in the following passage: [H]‌uman good turns out to be activity of soul in accordance with virtue, and if there are more than one virtue, in accordance with the best and most complete. (Nicomachean Ethics I 7, 1098a16–​17)

Just as the grammatical quality of an utterance can be in accordance with corresponding grammatical knowledge or competence in the agent, so the good quality of an action or an activity can be in accordance with corresponding virtue seated in the agent. In both cases, the performance falls short unless its good quality is sufficiently attributable to competence rather than chance. For a handy label, we can abbreviate such Aristotelian “success through competence rather than chance” as “apt success.” 2. Virtue epistemology lines up with this crucial component of Aristotelian virtue ethics, and is indeed a special case, as will now be argued. Virtue epistemology concerns performance that is epistemic rather than grammatical or ethical. Consider first affirmations. These can be either public, out loud, or to oneself in the privacy of one’s own mind. An affirmation could have any of various aims, and it could even have several at once. It could be aimed at misleading, or at showing off, or at propping someone up, or at boosting one’s own confidence as one enters athletic competition. An alethic affirmation is constitutively aimed at truth, at getting it right. So, the act of alethic affirming is the attempt—​the endeavoring—​to affirm a certain thing correctly, with truth, by intentionally affirming it. Humans perform acts of public affirmation in attempting to speak the truth, acts with crucial importance to a linguistic species. We need such affirmations for life in society: for collective deliberation and coordination, for example, and for the sharing of information. We need people to be willing to affirm things publicly. And we need them to be sincere (by and large) in doing so, by aligning public affirmation with private judgment.

Firsthand Knowledge and Understanding  111 3. Our virtue-​ theoretic framework is constituted by alethic affirmations, judgments, and judgmental beliefs (dispositions to judge). Since alethic affirmation constitutively aims at truth, this induces an epistemic normativity:  an alethic affirmation is Accurate if it succeeds in its aim, Adroit if it manifests enough relevant competence or skill, and Apt if its success also manifests such competence or skill. This is a telic triple-​A normativity of attempts as such, and of alethic affirmations (attempts to get it affirmatively right) as a special case. A judgment is a richer performance than a mere alethic affirmation, since (by definition) it is an affirmation that aims not only at getting it right, but also at doing so aptly, by manifesting epistemic competence. An alethic affirmation might be just a guess; but a judgment is epistemically more serious in aiming not just at truth but also at aptness. We have defined alethically affirming as attempting, by affirming, to affirm correctly. And we now define judgment as attempting, by affirming, to affirm aptly. 4. That is an initial sketch of the framework of virtue epistemology, which can now be enriched by distinguishing varieties of aptness. Thus, expert aptness goes beyond layman aptness. In domains of professional expertise, such as medicine and the law, the aptness attained through professionally expert judgment is more demanding than aptness through mere common-​ sense judgment. The stricter requirements of expertise are sometimes even codified in a rulebook, at least in part, as with law-​court procedure, admissible evidence, etc. 5. Expertise is thus a variety of competence. Please hold that thought for now, as we focus first on a further distinction in the philosophy of knowledge, between what one knows firsthand, and what one knows only through testimony accepted on sheer trust. We can distinguish firsthand vs. secondhand attainments of accuracy, adroitness, and aptness, and thereby distinguish firsthand vs. secondhand knowledge. In the history of epistemology, the secondhand is often dismissed as second rate or even worthless. Descartes and Locke are just two prominent examples.1 We can understand and make room for that attitude within virtue 1 Thus Locke: “The floating of other men’s opinions in our brains, makes us not one jot the more knowing, though they happen to be true. What in them was science, is in us but opiniatrety; whilst we give up our assent only to reverend names, and do not, as they did, employ our own reason to understand those truths which gave them reputation. Aristotle was certainly a knowing man, but nobody ever thought him so because he blindly embraced, and confidently vented the opinions of

112  Ernest Sosa epistemology. Indeed, we find the general sentiment enshrined already in our quotes from Aristotle. Query: When I see something on TV and have that kind of perceptual knowledge, is it secondhand? It’s not totally clear what makes testimonial knowledge “secondhand.” Reply: Interesting case! What is the epistemic role played by TV? Does it play a role more like that of a telescope or even like that of eyeglasses (for someone nearly blind), rather than like the role played by gauges? One can think of both sets of auxiliaries as perceptual aids. But they are importantly different. Gauges deliver propositional deliverances that we take as such, from which we then reason accordingly, to conclusions about the represented reality. Telescopes and eyeglasses make no such deliverances. And it seems plausible to classify TV projections, movies, and photographs similarly. We can hold all three to be cases in which, although we do perceive something (the screen or the photo) representing a reality beyond (let the TV program be a newscast, and the movie a documentary), still we do see the reality itself (with a delay for the movies and photographs, but the same goes for the stars we see at night). That seems plausible, as we move from the first to the third of these, even if our claim becomes steadily less plausible. With gauges, by contrast, we do not see the corresponding reality itself. So, testimonial reports are in that way more like gauge readings and less like screenings or photos. Based on your viewing a faithful photograph or a film, you can have firsthand knowledge of the aesthetically relevant qualities of a painting, nearly as well as you can have firsthand knowledge of the aesthetically relevant qualities of a film by viewing a meta-​film, a second film of a showing of that first film. There is a big difference between knowing the relevant aesthetic truths through these means, however indirectly, and knowing those truths only through deference to someone’s say-​so. another. And if the taking up of another’s principles, without examining them, made not him a phi­ losopher, I suppose it will hardly make anybody else so. In the sciences, every one has so much as he really knows and comprehends. What he believes only, and takes upon trust, are but shreds; which, however well in the whole piece, make no considerable addition to his stock who gathers them. Such borrowed wealth, like fairy money, though it were gold in the hand from which he received it, will be but leaves and dust when it comes to use.” Locke, An Essay Concerning Human Understanding, Book I.III.XXIV.

Firsthand Knowledge and Understanding  113 In any case, the distinction between hearsay and firsthand evidence seems clear and acceptable enough for our purposes, even without a sharp and distinct definition. 6. In recent years the worth of secondhand judgment based on testimony has been put in doubt within domains that host value judgments, whether aesthetic, moral, or philosophical. Compare our reliance on testimony when it comes to history or geography, or to traffic directions, or to random and isolated questions like When did Marco Polo arrive in China? On many and sundry subject matters, we happily accept brute testimony unaccompanied by any check of specific credentials. We then believe our informants without a second thought, as with directions from a stranger, encyclopedia entries, newspaper reports, etc. Often enough, we properly forgo any further inquiry. The contrast with moral or aesthetic issues is stark. Even when we grant that there is objective truth in morality and art, we are still properly reluctant to suppress our own critical faculties. And we are similarly reluctant in much academic judgment, where we aim to attain success firsthand, without essential reliance on sheer deference. 7. In a democracy, coordination must be based on equality and the vote. This is clear when the question is how we should coordinate our driving, whether “by driving on the right” or “by driving on the left.” Here democracy requires a collective choice by the subjects, in line with their preferred option. At least it requires agreement on how the decision is to be made, with technical details left to properly installed officials. But no antecedent objective fact can guide us when there is no antecedent advantage to everyone in the territory driving on the right, nor to everyone driving on the left, even if the following is antecedently best:  that either everyone drives on the right or everyone drives on the left. Democracy requires an informed citizenry willing to exercise their critical judgment, whether for the sake of a mere convention or for the sake of discerning together how to proceed on a matter of antecedent normative fact. 8. One can defer either on a question of convention or on a question of antecedent fact. Questions of convention concern no value facts antecedently in place. But not all questions of morality or art are to be settled by convention, with normative reality constructed from the ground up. Let us here focus

114  Ernest Sosa rather on questions with objective answers that correspond to antecedent facts (however grounded metaphysically). Even then we are reluctant to judge based just on hearsay. Although the pertinent evaluative domain may contain experts, we might properly refuse simply to defer. We prefer not to accept testimony uncritically, especially on general moral issues—​concerning abortion, for example, or homosexuality, or animal rights, etc. This surrender of our rational autonomy is repellent to most of us. 9. And the like is true of aesthetic judgment. We trust the newspaper on what film is showing where and when, and on plots, casts, and so on. These facts obviously bear on our decision. Based on a trusted critic’s rating, we may think we would enjoy a particular movie. Nevertheless, once we’ve seen it we refuse to let hearsay determine our aesthetic judgments. We insist on “making up our own minds.” True appreciation must be firsthand, as when we go to a theater or museum. It is not just that we want firsthand experience of the artistic object. Critical assessment requires more than just believing someone else, no matter how renowned a critic. That is not to say that we cannot gain knowledge by just trusting someone. We can surely gain apt judgment that way. Secondhand knowledge is still knowledge, which we need not reject even in evaluative domains. The point is not to reject secondhand knowledge, but to embrace firsthand knowledge. It is this desire for more that (in its scope and degree) distinguishes the relevant humanistic domains, as opposed to humdrum facts about weather reports, addresses, telephone numbers, and the vast amount of useful ordinary knowledge acquired through sheer trust of testimony. 10. Such secondhand knowledge does seem available in domains of morality, art, philosophy, etc. If there is objective truth in such domains, truth accessible to some of us, what could possibly preclude the testimonial “sharing” of such truth? If the distinctive yes/​no questions are answered aptly by an insightful judge, why deny ourselves secondhand knowledge through trust of such a source? At important moral junctures, we might even be morally negligent to act on our firsthand impression when others align massively against us. We might then act against what we know to be right. We would not know it through our firsthand judgment, but we might still know it testimonially nonetheless.

Firsthand Knowledge and Understanding  115 What is different about these domains is not that secondhand knowl­ edge is not available, but rather that much more is available and particularly desirable. Given our assumption of objectivity, what is further available and desirable in such domains is also knowledge, firsthand knowledge. And this is again a kind of apt judgment, where the thinker aims not just for alethic aptness but for alethic firsthand aptness. 11. Consider the issues distinctive of morality or art that we should address critically. Those issues overlap extensively with philosophical issues to be found in those domains. And philosophy seems particularly unreceptive to sheer testimonial deference. It seems different from the natural and social sciences in that regard, even if the difference is of degree rather than kind. That is why we so often react to a philosophical claim with “OK, but where’s the argument?” So, let us turn finally to philosophy itself.2 Those who aim to make original theoretical progress, scientific or otherwise, must limit sheer testimonial dependence on others. If you seek new, previously unappreciated facts in your domain of inquiry, you need to avoid direct reliance on others for your relevant beliefs. You may rely on others for data gathering, or even for helpful arguments, but the questions of distinctive interest to you must be attacked firsthand. In much analytic philosophy, as it has been generally practiced, we rely modestly on experimental data, and more on armchair thought. We rely even less than in normal science on the testimony of fellow inquirers and more on our own reasoning. More importantly, even when we do not aim to be original, we can properly avoid accepting mere hearsay, while still willing to be guided to helpful reasoning before making it our own. 12. Objection: Couldn’t the competence “seated within” a thinker be an unusually good ability to assess the testimony of experts? My knowledge that a restaurant will be good can then be firsthand—​the product of my competence in reading and interpreting reviews. Reply: Yes, the correct view must make room for this. Belief or knowledge that is essentially secondhand can still manifest more or less competence seated in the thinker. What makes it still essentially secondhand is the essential dependence of the rationale on the justificational input provided by the 2 More strictly, this concerns distinctively philosophical sides of philosophy, beyond overlaps with either science, or history, or scholarship more generally.

116  Ernest Sosa mere say-​so of a testifier. Thus, there is a difference between a great critic’s loud speech that someone speaks loudly and their speech that a certain movie is good. The first speech could give you firsthand knowledge that someone speaks loudly, but your knowledge that the movie is good based on the second speech would still be essentially secondhand even if it is dependent on your supreme competence to discriminate good critics. The content of the speech is justificationally incidental in the first case, but justificationally essential in the second. Rejoinder:  OK, but a more subtle problem must still be faced:  “When I want to know firsthand how subject S will answer a certain question, and I ask them, and they say, “Here’s how I answer that question: ‘Snow is white!,’ ” do I now know firsthand or second-​hand that they answer the question as follows, “Snow is white!”? Further Reply:  Intriguing example. I  know perceptually how S answers the question, and that is firsthand perceptual knowledge. Here, however, the content of the speech does not seem justificationally incidental as it does in the earlier objection! Further Rejoinder: Not in the same way, true enough, but that content is still relevantly incidental, in another way. My justification does depend on the content of S’s answer, but not on S’s reliable truthfulness in giving that answer, not on the truthfulness of that answer. Even if S is lying, or through some incredible glitch takes himself to be saying “Snow is black” rather than “Snow is white,” I can still know how he in fact answers the question, simply through my own (speech) perception of his actual answer, with no reliance on the epistemic status of that speech act. Moreover: When the testifier reveals a good argument, we might see the light thereby and make it our own. Here we’d rely causally on the testimony, without doing so epistemically. The validity and even soundness of the argument may be something we can appreciate fully on our own, with no need to rely epistemically on the endorsement of the testifier. This sort of knowledge can still be entirely firsthand even if causally it relies essentially on someone else’s say-​so. 13. Pessimism about moral or aesthetic testimony is defined by some as pessimism, concerning deferential judgment that p, in such domains. We are warned against judgment that is “deferential” by deriving simply from the say-​so of someone else, from their bare assertion that p. Some pessimists reject even the testimony of someone known to be much more reliable on the relevant subject matter and to share all of one’s pertinent information.

Firsthand Knowledge and Understanding  117 On the contrary, such deferential judgment can surely be appropriate or even knowledgeable in domains such as morality or art criticism. And moral action can surely be based properly on such deference. So, we might do best to defer in making a moral decision, while heeding the testimony of others that we recognize as more competent and better informed. However, it does not follow, from any such deference optimism, that we can lean back whenever a good enough testifier has had their say or is on call. Here we need distinctions. For the most part, we can rely on our tax preparers, attorneys, and medical doctors. Concerning those domains of expertise, we can appropriately defer even on important questions. Not so when we turn to moral, aesthetic, or political judgment, however, or to the humanistic questions generally, including philosophy. Here autonomous judgment, rational competence, and understanding come to the fore as desiderata absent from the domains of legal, medical, and financial advice. There is at least a substantial difference of degree. The humanities call for the exercise of one’s own critical faculties. You do not then function admirably by yielding your critical autonomy, so that your views forgo firsthand appreciation and understanding. This you can do with impunity in your financial, medical, or legal judgments, on issues of practical import for yourself and your family. Not so in what more intrinsically makes up your flourishing. Any proper human life will include a constellation of values, and a worldview integrally tied with such values. These values will be prudential, political, moral, and aesthetic, with the support of a humanly relevant worldview. Anyone wholly indifferent to such values and worldview yields the core of their humanity. To navigate uncritically, on mere instinct or tribal mores, is to neglect one’s rational nature. Prioritizing firsthand knowledge in those domains is integral to flourishing as a rational animal. 14. And so we come to the importance of understanding. Here we focus on a particular variety of understanding, namely understanding why, and argue for its importance as follows. a. To understand (outright) why p is to understand well enough why p. (Compare this: to be justified (outright) in believing that p is to be justified well enough in so believing.) b. To understand well enough why p is to know well enough why p.

118  Ernest Sosa c. On many questions of practical importance one can know well enough that p, and why p, even if one’s relevant knowledge is thoroughly deferential, secondhand, through trust in appropriately trusted advisors. Financial, moral, and legal advice tends to be thus merely informational. We very often need only the relevant information, with no need to understand why it is true. d. By contrast, on many aesthetic, ethical, or political questions, on many humanistic questions, we rational beings, including participants in a democracy, are right not to be satisfied with mere information. We seek deeper understanding. And, on many such questions, our understanding cannot reach proper depth if it does not go beyond sheer deference to the say-​so of others. On important life-​shaping and community-​shaping questions, we must aim beyond sheer deference toward firsthand understanding. e. And the same goes for appreciation of the artwork that suffuses our lives once we meet our bare necessities. Here again, we must go beyond sheer deference if we are to live a life that is full enough to fit our rational status. Our lives are not properly realized without enough of the firsthand appreciation and understanding distinctively proper for rational beings. 15. Let us consider next how understanding and inarticulacy are related in a flourishing human life. a. The bulk of what any one of us knows in an advanced industrial society is a river that flows on a riverbed of memory. Once beyond childhood, even once beyond infancy, the epistemic basis of one’s articulable knowledge at any given moment is to be found upstream, with multiply diverse tributaries—​testimonial, perceptual, or inferential, singly or in combination. b. The idea that at any given time your effective epistemic basis is articulable by reference to what is then available to you is a myth. Take at random nearly any common-​sense platitude and try to articulate your rational basis for that belief. Take “water tends to flow downhill,” or “day and night alternate regularly,” or “people normally have two hands,” and so on, for the massive articulable common sense that we live by. All of it is known perfectly well at indefinitely many times when one would be hard put to provide any adequate and then available rational basis. The

Firsthand Knowledge and Understanding  119 basis for such knowledge is to be found already in the diachronic experience of the growing child. At no time in that flow can one provide the full basis for one’s believed generalization, whose epistemic standing derives rather from long enough exposure to refutation or revision. Only the essentially diachronic spread of that exposure provides eventually the proper epistemic standing, when long, rich, and positive enough. Compare the knowledge captured when we prefix “all across the globe” to the earlier conjunction of platitudes. Attaining such knowledge required sophisticated scientific input to our received human wisdom. And how exactly does the child gain access to that more sophisticated knowledge? The child does so in very various ways, ranging from sheer testimony all the way to a deep and sophisticated understanding of astronomy. Even when the specifics of the relevant testimonial exchange recede from view, however, even when one forgets crucial details of the scientific reasoning, none of that need remove the knowledge lodged in one’s memory that there is such a thing as the globe, that there is a massive round earth on whose surface we stand, etc., even when we are forced to revise supposed platitudes like the one about alternating day and night. (Not at the poles!) Consider the epistemic standing of such platitudes, including those that reach the sophistication of astronomy. At any given time, the standing of such platitudes for a normal adolescent will derive not just from what they can cite as a rational basis at that time, but rather from how information has entered their diachronic flow of experience and belief. The quality of that inflow—​perceptual, testimonial, inferential—​has obvious bearing on the later epistemic standing of that belief, as does the quality of the mnemonic riverbed. One needs no present awareness of that erstwhile quality in order to properly retain the retained beliefs. Nor is one’s belief much impugned simply because one cannot presently adduce much by way of explicit rational basis. Take the contribution made by one’s present mnemonic seeming, by the fact that it seems to oneself that p, and that one seems to remember that p. This is at most a puny contribution to the epistemic standing of one’s belief, by comparison with the quality of the combined inflows that contributed diachronically to one’s holding it now. c. In addition, much knowledge is arguably built into our brains from birth, ready for triggering with normal experience, so that its epistemic basis derives from the evolution of our brains, and not just from the

120  Ernest Sosa individual’s river of experience, with its distinct tributaries, nor from enculturation. d. Consider finally the example of our mores knowledge (as opposed to our moral knowledge). This requires a long, sometimes painful learning experience, aided by reward and punishment, approval and disapproval. At first, the compliant child takes it on deference, in situation after particular situation, what they are to do in that situation. Their knowledge then falls far short of what they will eventually attain, as they increasingly put their conduct in line with community practice. Only after a long period of learning do they acquire a mastery of that practice. Importantly for us, what they learn need not be articulable. Such mastery often involves a depth beyond linguistic articulation, a depth revealed only in the systematic conduct itself. e. Why so? What suggests any such limits of linguistic formulation? One powerful hint points to words like “sufficient,” “enough,” “adequate,” and their negations. Our conduct must often be guided by attention to various dimensions within which lie our action and its consequences. These various costs and benefits must be assessed and a balance must be drawn as we face a question of the form “Will there be enough by way of benefits B1, . . . , Bm, in combination, to sufficiently outweigh the combination of costs C1, . . . , Cn?” f. Consider even cases where the action and its relevant consequences do not involve such multi-​factor complexity, cases where there is just B1 to consider, and there’s no cost worth considering. Even in such a simple case, we need to assess whether there is enough of B1 to render the conduct a must (according, again, just to the mores of the community, to the practice), or whether so acting is permissible but optional. Initially the child must be instructed in the ways of the community, case by case. Sheer deference is then required, and the child is right to comply deferentially. Fortunately, this deferential knowledge eventually graduates to something more general and ingrained. But it rarely if ever does so in a way that can be manifest in a sentence of English, except trivially, as “When there is enough of B1, then you must act accordingly.” Even the clueless can of course agree with that much! What distinguishes the adept, those who are now masters of the practice? It is not just their ability to mouth, or even seriously to believe, that when there is enough of B1, then (according to the practice) one must act

Firsthand Knowledge and Understanding  121 accordingly. No, what distinguishes the adept is their embodiment of the corresponding competence, their competence to tell systematically and well enough when the case does involve enough of B1, so that according to the practice they must issue the required action. They can of course manifest their mastery of the practice to this extent, without themselves complying systematically. They may remain rebelliously, deliberately noncompliant. All they need for the mastery is the ability to tell well enough what the practice calls for. 16. Certain questions in the humanities require a particularly large and salient element of direct rational appreciation. If so, why might this be so? Why might a question call for much more than deference? Because it might call for a kind of rational understanding.

Consider, for example, the aesthetic assessment of an artwork. If it is indeed successful, there will be reasons why that is so, reasons in virtue of which the work attains its success. And there is then a notable distinction between the following two cases: In the first case, one knows through sheer deference about the success of a certain artwork. In the second case, one spots at least implicitly the reasons why the work is successful, on which one’s knowledge of its success is based.

In the latter case, one has firsthand knowledge of the work’s success along with understanding its success by knowing why it is successful, through insight into the grounds for that success. Here firsthand humanistic knowledge comes with understanding attained through insightful rational explanation. One experiences the work in the relevant way—​be it a piece of music, a painting, or a novel—​and one discerns the reasons for the work’s success through firsthand experience. Only thus is there a proper appreciation of the work’s success. Otherwise there is only an ersatz grasp that enables words of assessment empty of real understanding. This paper has tried to enhance our understanding of the notions of firsthand knowledge and understanding, of how these are interrelated, and of their importance in a flourishing human life.

7 Toward a Theory of Understanding Linda Zagzebski

1.  Introduction One of the most important powers of the human mind is the ability to grasp simple structure in a complex array of phenomena. Some of the structures human beings see and hear occur in nature, such as the structure of a tree, the pattern of a birdsong, or the structure of the solar system, and some are created by us, such as the layout of a garden, or the structure of a fuel injection system. In the case of some structures, we may not know whether they are natural or whether they were humanly invented. I believe that to be the case with the structure of a proposition. Some structures are extended in time and some are not. If we take something that is not temporally extended and look at it over time, we might see a pattern—​e.g., the ideological leanings of Supreme Court justices over the last century, or the Dow Jones Industrial Average over the last ten years. Sometimes we see patterns when nothing is there to see, at least nothing interesting that will be repeated in the future. I have heard baseball statistics that are no doubt in that category, and the Dow Jones Industrial Average might be also. But whether structure is discovered or imposed, important or trivial, the urge to find structure and the ability to do so is a universal component of the human mind. Pythagoras produced some of the most impressive feats of the human ability to perceive structure. His discoveries in geometry led him to the view that the entire universe has a mathematical structure, and with that insight, he discovered the musical intervals, mapped the stars, and created a fascinating ethical system in which natural laws of harmony apply to the human soul and to the state. So the Pythagoreans had the ability to see recurring structures throughout the universe—​beyond the material world, and into the domains of musical harmony, morality, and human destiny. Few of the best philosophers today can boast of such an expansive ability to grasp repeating structures. But all of us perceive structure all the time, and in our

124  Linda Zagzebski more creative moments, we notice similarities of structure in one part of the world and another; for instance, the cross section of a tree trunk looks like expanding waves when a pebble is dropped into a pond. Neurons look like trees with main trunks (axons) and branches (dendrites). We can see the golden ratio in such widely different objects as the human face, the Parthenon, and Salvador Dalí’s Sacrament of the Last Supper. A structure is illuminating even when it does not repeat, but the ability to see the structure of one domain in an entirely different domain is an important extension of the human grasp of structure to which I will return in the last section of this paper. Very roughly, I think of a structure as what gives an object unity. We want to grasp structure because we see an object as an object when we grasp its structure. The structure of a thing is typically what we have in our minds when we think about it or remember it, and it is what we keep in our minds when we want to study it or affect it in some way through our actions. I do not mean that we are not capable of grasping anything but structure. We can have mental images or ideas that constitute a detailed grasp of physical or non-​ physical objects, but for cognitive economy, the structure is basic. We tend to put mental images or ideas into the structure we grasp, and the structure is what permits us to put the mental images or ideas in their proper place. The grasp of the structure of some phenomenon is the grasp of it as a whole. A human intentional act is a single act because it has a structure. World War II is a war, and not a series of random events, because it has structures, and the fact that it has multiple structures and to some extent, no structure at all, is perceived as a problem. It is a problem in understanding World War II. I propose, then, that understanding is the grasp of structure. When we grasp an object’s structure, we understand the object. The object of understanding can be anything that has structure: a living organism, an event, a narrative, a piece of music, a work of art, a metaphysical system, a philosophical argument, a causal relation, the stock market, human intentional action, a moral theory. I said that we sometimes impose structure on a phenomenon, and I think that we do that in order to understand it. In fact, that is almost always what moral theorists do. If we cannot discover structure in some phenomenon, we impose structure on it in order to make the pieces of it fit together. We do that because the pieces are too difficult to grasp together without a structure into which we can place them.1 1 I say explicitly that that is what I am doing in the moral theory I devise in Zagzebski (2017: Ch. 1).

Toward a Theory of Understanding  125 Some objects of understanding are components of other objects of understanding. An event is a component of a narrative. A particular causal relation is a component of an event. An intentional act is a component of an event, and it is a component of a person’s motivational system. Our planet is a component of the solar system, which is a component of larger celestial systems, the largest of which is the physical universe. Understanding must simplify what it grasps, and the larger and more complex the object of understanding, the more we must simplify and leave out of the phenomenon components that may be important at different times or for different purposes. A map of a geographical area is a good illustration of this point about understanding. A map leaves out a great deal of the physical region of the map. A map of a large region leaves out more detail than a map of a smaller region because the inclusion of too much detail would include more than the mind can grasp, and so it would not aid understanding. Simplification is therefore a cognitive advantage. We ignore parts of a phenomenon in order to grasp the whole. But sometimes we want to find something that is left off the map, and we will need to zoom into the map or find a more detailed map. What counts as a whole phenomenon will vary with our interests. I enjoy looking at a globe, but if all I want to do is to get from one building to another on my campus, I want a campus map, not a globe or a map of North America or even a city map. Simplifying also can distort what is there. For instance, the lines depicting roads on a street map may be straighter than the roads actually are. That usually does no harm, and can actually be beneficial. It would be distracting and perhaps confusing if the depiction of the layout of the region was more accurate. Similarly, as Catherine Elgin (1996) has argued, physical laws expressed in simple formulas are not quite accurate, but we get more understanding from the simpler formula that somewhat distorts what it is displaying. I have mentioned that structure can be depicted in various ways, and a map is one of them. Structure can also be depicted in sentences, diagrams, graphs, sketches, and mathematical models. Structures can be mentally grasped and depicted in a way that can be communicated to others. People who are unacquainted with some domain of the world or have trouble grasping its structure on their own can be led to see the structure of a thing through the help of others who are able to depict it. My idea of structure has historical roots in Aristotle’s idea that what we grasp when we know something is its form. Aristotle thought that in perception and in thought, the form of a thing is imprinted on our minds. The form

126  Linda Zagzebski of an object is what the mind grasps. What I mean by structure is related to what Aristotle means by form, only I think of structure as extending far beyond material substances and artifacts, and I am not endorsing the matter/​ form distinction as it functions in Aristotle’s theory of cognition. But I think that Aristotle is right that when we go beyond the simplest objects of perception or memory, what the mind grasps is the form of an object. What makes something a something is its form, or its structure. When we do not understand something, it is usually because either it has no structure, or we cannot figure out the structure. With this view of understanding, I will propose and briefly support four claims: 1. Understanding is the basic positive epistemic state. 2. True belief is the grasp of a propositional structure, and so it is a special case of understanding. Since knowledge is a special case of true belief, knowledge is a special case of understanding. 3. Knowledge and understanding of other structures in the same domain are checks on the veridicality of each other and show us that both states link us to the same world. 4. The grasp of a structure that repeats in different domains can lead to a strong form of discovery by analogy that is more powerful than its poor cousin, the inductive argument by analogy. It has been useful for innovation in physics, and is an important part of creativity that we should encourage in the education of students.

2.  True Belief and Understanding One kind of structure is the structure of a proposition. When we have a true belief, we grasp a propositional structure that some part of the world has. Propositional structure is fine-​grained, and the syntactical structure of language permits endless variations in the sentences we can produce with the same general structure, so the grasp of true propositions allows us to have an enormously powerful type of understanding. The ability to discover (or impose) propositional structure on the world is one of the greatest achievements of the human mind. A proposition conveys a particular kind of structure of some part of the world, but there is always more in the world than is conveyed by any set of

Toward a Theory of Understanding  127 propositions, no matter how long and complex. Compare a description of a room by even a highly talented novelist with seeing the room in person. At best the writer’s description works by permitting the reader to form her own picture of the room which is more or less accurate (usually less). At best the set of propositions permits the formulation of an imaginative substitute for seeing the room, sight being our most advanced sense, but even sight does not convey everything in the room. There is always much more in reality than what we can grasp—​certainly much more than we can grasp at any one time. Propositional structure aids the mind by expressing something we can grasp, remember, and communicate to others in a way that comes as close as human minds can get to an exhaustive grasp of some part of reality. The grasp of propositional structure when combined with a large vocabulary is therefore an incredibly powerful way to capture what the world is like for beings like us. The virtue of propositional structure is also its vice. Propositional structure is fine-​grained, but it sometimes is too fine-​grained to illuminate for us the domain it is depicting, and when that happens, understanding is jeopardized. Fortunately, the phenomena depicted in a proposition or a set of propositions often have more than one structure. In fact, they probably always do. There is no reason to think that propositional structure is all there is. Suppose, for example, I am interested in the layout of an ornamental garden. A sketch of the garden is far more useful to me than a list of propositions giving the appearance and dimensions of each plant in my garden and their relative positions. It is possible that a long list of propositions could convey an understanding of the garden’s design, but to read and grasp such a long set of propositions would be at best tedious, and it is unlikely to work anyway. For the same reason, if I wanted a friend to understand what the garden looks like, I would show her the sketch rather than to give her a verbal description. The same point applies to understanding any spatial structure. The layout of a city is best depicted on a map. The layout of the solar system is given on a drawing or a three-​dimensional model. Another kind of structure that is best depicted non-​propositionally is the pattern of a phenomenon that changes over time: monthly rainfall, interest rates, public opinion of the president, life expectancy, percentage of women in philosophy, yards per game of a running back—​each of which can be most easily depicted on a simple line graph. That explains why newspapers often include graphs along with an article verbally describing the same phenomenon. Graphs are also better than words at depicting phenomena in which

128  Linda Zagzebski one feature varies with a second feature, such as the sway of a tall building as a function of wind speed. More complex variables can also be depicted using two-​or three-​dimensional models such as climate change as a function of a set of variables. What this indicates is that any given part of the world has more than one structure. Propositional structure is a type of structure that every part of the world has, and we can grasp any part of the world we want by way of grasping it propositionally, but we can also grasp many parts of the world through the grasp of non-​propositional structures such as those depicted on maps, sketches, graphs, and three-​dimensional models. I  am proposing that the grasp of a structure that some part of the world has given us understanding of that part of the world. Since true belief is a grasp of propositional structure, true belief is a special case of understanding. Assuming that knowledge is a special case of true belief, it follows that knowledge is a special case of understanding. Both propositional knowledge and understanding of non-​propositional structures can be transferred to others. Whether the structure is depicted in language, in graphs, or in some other way, the receiver needs to be tutored in the art of interpreting the written or spoken word, the graph, or the model used. It is possible that there are structures that one person can apprehend without the ability to communicate the structure to others, but in general, understanding of non-​propositional structures can be acquired by a process parallel to propositional testimony. I don’t know if the word “testimony” is generally used when person A shows B a map or a sketch, but I think that it is the same phenomenon that occurs when A tells B that p. This way of approaching understanding gives us a different way to think of the issue of whether understanding includes knowledge. According to the view I am proposing, knowledge is a form of understanding, so it trivially follows that one kind of understanding entails knowledge. But most philosophers who are concerned with this question are thinking of a different issue, which in my terms would be formulated as the question whether the understanding one has when grasping a non-​propositional structure of some part of the world entails grasping propositional structure as well. The object of knowledge is different from the object of understanding of non-​propositional structure, but the domain might be the same. Suppose you understand the layout of a university, or the motives of another person, or the way a fuel injection system works. In each case, if you have understanding of the object, you will normally believe a lot of true

Toward a Theory of Understanding  129 propositions about it, but you might believe some falsehoods without much damage to your understanding of the domain. So if I understand the layout of the University of Oklahoma, I will truly believe that the library is located between the North and South Ovals, that the administration building is on the North Oval, and much else. I might not know the location of certain buildings or the distance between one and another, and if so, my understanding of the layout of the campus is diminished. Greater understanding of the layout of the campus is normally accompanied by more true beliefs about the location of various buildings and the way they are related to each other spatially. But I can have some false beliefs without significantly diminishing my understanding. I may misremember the number of seats in the stadium or the exact location of many buildings, and some buildings I may not remember at all, yet I grasp the layout of the campus quite well. I grasp it well because I know or at least truly believe key features of the layout of the campus, but some features are less important than others, and that is no doubt why it sometimes appears that understanding is relative to our interests. It is important to me to know where Dale Hall Tower is because that is the location of the Philosophy Department, but it may be more important to you to know where the stadium is. That is obvious, but it is not obvious that what counts as success in grasping structure is relative to our interests. What is relative to our interests is the degree of detail we need to grasp, and the degree of distortion we can safely accept in the structure we grasp, but not the structure itself. Given this view of understanding, it follows that understanding is answerable to the understanding of a different structure in the same domain. I don’t understand the layout of the city of Norman without some true beliefs such as the belief that Main Street runs roughly east/​west, and I-​35 runs roughly north/​south through the west side of town. So the understanding I have non-​ propositionally is answerable to certain propositional facts. But my grasp of propositions about the layout of Norman is also answerable to the understanding I have from a map or from the image I have in my head that I get from driving around. When my map or the image in my head conflicts with my beliefs, I know that one of them is mistaken. A map can be a way for me to tell whether the propositions I believe about the domain of the map are true rather than false. So putative knowledge can be corrected by understanding of a different structure of the same thing. Conversely, the map may be the one that is mistaken. In some cases (but not all) I can refer to a different and more reliable method of grasping structure to correct one that is in dispute. Both

130  Linda Zagzebski the propositions I believe and my grasp of non-​propositional structures are correctable by perceptual experience. The same point applies in other domains. Historians have a particular interest in causal structures, and one of the most hotly debated series of events in Western history is the decline and fall of the Roman Empire, the causes of which continue to fascinate historians as well as ordinary people. Kyle Harper (2017) has argued that there is a causal connection between climate change and disease and the fall of the empire. He says that most histories of Rome’s fall have been built on the assumption that the environment was a stable, invisible backdrop to the story. But given the advances in our ability to retrieve data about the paleoclimate and genomic history, this assumption has been proven wrong. To explain how the Roman Empire went from territorially dominant, populous, and prosperous to conquered, thinly populated, and impoverished requires the recognition of both human and environmental causes. Barbarians and social conflict were among the causes of decline, but climate and disease were causally important also, and climate change and disease were connected because change in climate stirred pathogen evolution. Two catastrophic epidemics stand out. In the third century, the empire was swept by the Plague of Cyprian, caused by a pathogen that is currently unknown or at least uncertain. Even more devastating was the Justinianic Plague, which DNA evidence now identifies as Yersinia pestis, the same bacterium responsible for the Black Death many centuries later (Harper 2017: 18). Histories often focus on military battles because they can be precisely dated and many of their effects are immediate and obvious. So the Battle of Adrianople in the late fourth century is sometimes credited with beginning the process that led to the fall of the Western Empire in the fifth century. That battle is reputedly the bloodiest in imperial history, when a desperate force of Gothic invaders overran the main body of the eastern army. But as Harper notes, at most twenty thousand Roman lives were lost that day, and while the fact that they were soldiers magnified the problem, his lesson is this: “germs are far deadlier than Germans” (2017: 18). The grasp of a causal structure explaining the decline of the Roman Empire is different from believing a series of true propositions about climatological and historical events. I am arguing that both kinds of grasp are forms of understanding, and each kind is answerable to the other. The grasp of causal relationships and interlocking structures is answerable to certain propositional facts, such as the date of the appearance of certain pathogens, the rate at which they spread, death rates from disease, facts about the economy

Toward a Theory of Understanding  131 in different parts of the empire, the removal of Roman authority from certain areas, and so on. At the same time, historians devise non-​propositional models to explain a host of facts, and some of these putative facts need to be revised in the light of a causal model. To some extent, Harper argues, historians have not been looking in the right places for the relevant facts. When it is assumed that human conflict in the form of invasions and social strife are the principal causes of the disintegration of the empire, only facts of that kind are considered, and the importance of some of them is exaggerated. But now climate science and the genomic revolution are making us aware that climate change and emerging infectious diseases have been an integral part of the human story all along. The causal narrative of the decline of Roman civilization is beginning to take a different form. The emerging pattern is quite different from what it once was. The desire to grasp the causal structure of the decline of the Roman Empire illustrates the fact that there is a human impulse to perceive structures that are larger than the structure normally depicted in a single proposition. Knowledge of a single proposition gives us a grasp of some part of reality, but knowledge of a string of propositions does not add up to a grasp of a larger part of reality unless each part of reality known in a proposition has a place in a larger non-​propositional structure. The causal structure of historical events is a large structure in the sense I mean. The propositional facts that are the object of knowledge and the non-​propositional causal structure that gives us a different kind of understanding are correctable by each other. That gives us reason to think that understanding and knowledge are connected to the same world.

3.  Repeating Structures and Analogical Discovery I am proposing that understanding is the grasp of structures of reality. In understanding we are able to see unity in complex phenomena, and that enables us to see some part of the world as a single object. We see a geographical area as a unit because we see a patterned structure in it that we can hold in our minds and possibly remember. We can see the Roman Empire as an object to the extent to which we see in it an evolving structure that eventually disintegrated. The human impulse to see structure makes us look for structure in events and in sequences of events such as the sequences that constitute a human life or a human society.

132  Linda Zagzebski The impulse to see structure leads to another kind of understanding, one that we get from seeing the same structure repeated in multiple domains. At the beginning of this paper I mentioned the golden ratio, which appears in many parts of nature as well as in art and architecture. What have we discovered when we see the same mathematical ratio in the growth of leaves and petals, in the Milky Way galaxy, in Leonardo’s Vitruvian man, and in the Fibonacci sequence? When a structure reappears in different parts of nature, that suggests something about the organizational structure of all of nature. Nature has a pattern. I find it fascinating to notice that and to realize that the ability to see the pattern can lead to some of the most impressive intellectual achievements in history. It led the Pythagoreans to believe that the basic structure of the universe is numerical. With that insight they were able to link together mathematics, music, physics, and an ethical system in which natural laws of harmony apply to the human soul. Dmitri Krioukov et al. (2012) argue that the universe may be growing in the same way as a giant brain, where the electrical firing between brain cells mirrors the shape of expanding galaxies. Physicist Stephon Alexander (2016) has argued that repeating features of nature validates a form of argument from analogy that, in my opinion, is much stronger than the standard inductive argument from analogy. If we see part of a structure repeated elsewhere, we have reason to expect the rest of the structure to be there also. This has led to important scientific discoveries. As Alexander points out, Kepler’s ability to connect geometry and musical intervals led him to discover his three laws of planetary motion. Kepler correlated the planets’ placement in the heavens and the speed by which the planets made their orbits with an accurate and measured complete musical scale. By 1605 he had discovered that planets move in an elliptical orbit and that a line joining them to the sun would sweep out equal areas of space in equal periods of time (Kepler’s second law). He calculated the astronomical motion of the heavenly bodies and described them as musical harmonics or note sequences. For each planet the lowest note corresponds to the largest distance from the sun (the lowest orbital velocity), and the highest note corresponds to the shortest distance from the sun (the highest orbital velocity). Each planet creates a harmony as it orbits. So the planet Saturn plays a major third (a pitch of 5:4), Jupiter plays a minor third (6:5), and Mars a perfect fifth (3:2). (Alexander 2016: 82). So Kepler offered the hypothesis that the planets make music as they orbit.2 This is fascinating 2 Kepler’s Music of the Spheres can actually be performed in real time, using Kepler’s formulas, but the entire composition would take 264 earth years to perform.

Toward a Theory of Understanding  133 in itself, but the point that Alexander stresses is that Kepler was not able to demonstrate his third law and show that his theory applied to all of the planets until fifteen years after he had used the analogy between music and geometry to drive his second law (2016: 81). His understanding of the connection between musical harmonies and the geometry of planetary motion led him to his astronomical discoveries. Alexander argues that the key to innovation in theoretical physics today is the same kind of analogical reasoning used by Kepler (2016: 2). He says that through his understanding of jazz he was able to quickly intuit string theory, and he proposes that the universe is composed of vibrating strings. Alexander’s works on the connection between the smallest and largest entities in the universe applies the physics of superstrings to address long-​ standing questions in cosmology. In 2001, he co-​invented the model of inflation based on higher dimensional hypersurfaces in string theory called D-​Branes. His fascination with the analogy between music and physics has led him to propose that the universe is one big vibrating string rhythmically moving from big bang to big crunch and back again (2016: 209). I am not in a position to have an opinion on whether the universe is vibrating strings on the largest scale as well as the smallest, but I think that the kind of analogical reasoning described and used by Alexander in his field of string cosmology is an illustration of the importance of a kind of understanding that applies the grasp of structure in one domain to a completely different domain. It works if the universe is unified in such a way that results in repetitiveness in structure. If it is not an accident that the orbits of the planets correspond to pleasing musical intervals, then it is not an accident that when Kepler discovered that Mars “plays” a musical fifth in its orbit, the orbits of the other planets would have orbits corresponding to other harmonious intervals. Since Kepler understood musical intervals and recognized that one part of the planetary system had the structure with which he was familiar in music, he was able to predict the same musical structure in other parts of the planetary system in advance of the measurements that confirmed it. This is a form of argument by analogy that is much stronger than the familiar form of analogy in which similarities are used as a basis to infer some further similarity that has yet to be observed. More specifically, the logical form of argument by analogy is typically said to be as follows: “Two otherwise unrelated objects, A and B, share properties P1, P2, P3, etc. Object A also has property Q. Therefore, B probably has property Q.” It has frequently and understandably been pointed out that such an argument is weak. In the kind

134  Linda Zagzebski of analogy I have described, however, the analogy is not based on a series of shared but possibly unrelated properties, but on similarity of perceived structure. If speed is related to musical pitch in one planet, it is reasonable to expect the same relationship in other planetary bodies, not because it is reasonable to expect them to share one more property if they are known to share several properties, but because the repetition of structure upon which the analogy is based has already been perceived. In education we train students to be critical reasoners. That typically means training them not to make mistakes in inferences, raising critical questions about their own and other people’s claims, and knowing how to identify and to evaluate evidence for a belief. None of this encourages understanding. Can we train students to grasp structure? How can we help them to see the same structure in more than one domain, such as the domains of music and geometry and cosmology? It is unlikely that we can do that in a single course, but it is helpful for everyone to have practice in seeing non-​ propositional structures and reflecting on them. Some courses teach the grasp of musical structure; others illuminate structure in fiction or film. The social and natural sciences frequently make use of graphs and charts, the interpretation of which can be fostered with practice. Historical writings help readers notice complex causal networks such as those proposed to explain the decline of the Roman Empire that I discussed previously. Grasping structure is an important human ability, and the ability to grasp the same structure in widely differing domains is an important kind of creativity that can generate advances in knowledge by analogical reasoning, as Kepler did in the field of cosmology. This ability cannot be exercised except by individuals who are knowledgeable in more than one field, and so this is a reason to educate students in many fields at the same time even if they ultimately focus on a single field of study or domain of human practice. Why is it that the same structures repeat in many areas of nature? An obvious answer is that all of nature is unified. We know that the desire to grasp that unity drives people like Stephon Alexander in his work on string cosmology. It also seems likely that all of reality, both physical and non-​physical, is unified in some way. The grand metaphysical systems of the past attempted to explain that unity, but there have been few such systems since the nineteenth century. Still, the human desire to grasp the totality of reality is a deep part of our nature. One of the mysteries of the world is how it is that beings with intellects like ours can grasp large portions of reality and almost everyone thinks it is possible to grasp the whole of it.

Toward a Theory of Understanding  135 An ancient view going back at least to the Pythagoreans and expressed in many of the world’s religions holds that the human mind or soul can reach union with reality as a whole. According to Aquinas, that is because “intellectual natures have a closer relationship to a whole than do other natures; indeed, each intellectual substance is, in a way, all things. For it may comprehend the entirety of being through its intellect” (1956: III. Ch. 112.5). What might be more surprising is that some of the most skeptical philosophers believe the same thing. This is what Bertrand Russell says at the end of The Problems of Philosophy: “Philosophy is to be studied . . . above all because, through the greatness of the universe which philosophy contemplates, the mind also is rendered great, and becomes capable of that union with the universe which constitutes its highest good” (1968: 161). Understanding is essential to satisfy our cognitive desires because it is a way of grasping reality without dividing it into propositional bits. I have proposed that even the propositional bits give us a form of understanding because reality has propositional structure as well as structures of other kinds. But the atomistic approach to knowledge hides some of the powers of the human mind that are most important. The impulse to grasp larger and larger wholes pushes us inevitably to the desire for comprehensive understanding of the totality of what exists. The idea that the human mind is capable of grasping all of reality is the greatest idea our species has ever had.

References Alexander, Stephon. The Jazz of Physics. New York: Basic Books, 2016. Aquinas, St. Thomas. Summa Contra Gentiles. Translated by Vernon Bourke. Notre Dame: University of Notre Dame Press, 1956. Elgin, Catherine Z. Considered Judgment. Princeton, NJ: Princeton University Press, 1996. Harper, Kyle. The Fate of Rome: Climate, Disease, and the End of an Empire. Princeton, NJ: Princeton University Press, 2017. Krioukov, Dmitri, Maksim Kitsak, Robert S. Sinkovits, David Rideout, David Meyer, and Marián Boguñá. “Network Cosmology.” Scientific Reports, November 2012. Russell, Bertrand. The Problems of Philosophy [1912]. New  York:  Oxford University Press, 1968. Zagzebski, Linda. Exemplarist Moral Theory. New York: Oxford University Press, 2017.

II

PSYC HOLO G Y OF U N DE R STA NDI NG

8 Technology as Teacher How Children Learn from Social Robots Kimberly A. Brink and Henry M. Wellman

Humans live in a world increasingly filled with smart technology—​laptops, tablets, smartphones, and robots. We not only live with them, we and our children receive increasing amounts of information from them. We can uncover when dinosaurs went extinct, how far it is from Earth to the Moon, and what the weather is like at the North Pole all from our laptops, tablets, phones, and now even robots. These interactions require and reveal two varieties of understanding: an understanding of these devices and their abilities and an understanding of the kinds of information that these devices provide. We discuss these varieties of understanding in this chapter with a focus on social robots. We focus on social robots for both practical and substantive reasons. Practically, considering how children learn from all smart technology is beyond the scope of a single paper. Substantively, humanoid robots are a special, and specially revealing, form of smart technology worthy of attention in their own right. First, they are increasingly present in the lives of children. And second, social robots provide a unique means to investigate the ways in which children learn from others. Every year, more and more robots are designed to befriend, teach, and care for children. Jibo, iPal, and Zenbo (Figure 8.1), all Pixar-​like robots, are designed to play games, answer questions, read stories, and even watch children unsupervised (Glaser, 2016; Low, 2016; Wong, 2016). Moreover, several robots have been working with children in classrooms, daycares, clinics, and hospitals for years. Across the globe, robots are teaching children language skills (Movellan et al., 2009), mathematics (Wei et al., 2011), science (Hashimoto et al., 2013), physical exercises (Mejías et al., 2013), and even social skills (Ricks & Colton, 2010). More and more, we expect children to accept these technologies and the information that they provide. Yet there has

140  Kimberly A. Brink and Henry M. Wellman

Figure 8.1  The robots iPal (left panel), Jibo (center panel), and Zenbo (right panel) have features and behaviors designed specifically for interacting with children.

been limited focus on whether and how children will accept and learn from these devices or the conditions that would enhance or detract from such interactions. Due to the increasing number of robots playing with, teaching, and caring for children, it is important to explore the nature of children’s understanding of these devices and how that understanding impacts their interactions with them. Furthermore, studying children’s willingness to learn from social robots provides a unique perspective with which to investigate children’s learning more generally. Until recently, children have efficiently and effectively learned about their world through knowledge passed on by other people (e.g., parents, teachers, peers). This phenomenon, known as “trust in testimony” or “natural pedagogy,” shows how children are adapted to learn gen­ eral knowledge from human communication (Csibra & Gergely, 2009). For example, children often decide to accept information from an instructor based on the instructor’s social and language cues (Csibra & Gergely, 2011; Koenig & Harris, 2007). But now, children are receiving information about their world from social robots which, like humans, are similarly capable of conveying information using language and social behaviors, but are not humans. Robots, unlike other forms of technology, can imitate humans in a myriad of ways: they may behave, respond, or even look like humans. By exploiting these similarities, robots may be able to take advantage of this same system of social learning to effectively transmit information to children

Technology as Teacher  141 (and we argue that they do). As social robots increasingly act as teachers to children, it would be valuable to explore whether and how children learn from these devices and under what conditions. This focus on social robots allows us to consider questions such as these: To what extent do children perceive that robots are like people? Does an understanding of the shared similarities between robots and people affect children’s ability to understand and accept information from these devices? And, of crucial importance we will argue, how do answers to these questions vary with children’s age? To address these and similar questions, first, we present the broad research on children’s understanding of robots and how their expectations change with age. Second, we demonstrate how that understanding impacts children’s feelings toward robots, including our own research on children’s “uncanny valley” reactions. Third, we discuss how children’s understanding of robots impacts their willingness to learn from them. Here, we review the literature on robots as educational tools for children. Fourth, we describe our own research addressing children’s trust in testimony given by humanoid robots. Finally, in the last part of the chapter, we conclude with suggestions for further exploration of how children learn from, and how they come to understand, smart technology—​computers, smartphones, and especially humanoid robots. Throughout, we emphasize the need to consider these issues through a developmental lens because the answers to these questions are arguably different for children of different ages with different kinds of experiences. In this chapter, we highlight an often-​overlooked perspective within the literature on children’s interactions with robots. As we will outline, research on the effectiveness of robotic social companions has, until recently, focused primarily on features of the robot. We argue, however, that research is badly needed to assess how children’s cognitive abilities and their developmental trajectories, as well as the design of the robot, all work together to impact children’s learning and feelings toward robots. Thus, we propose possible answers as to how and why these child factors—​cognitive abilities that themselves change with age—​interact with the overt features of social robots to affect the quality of social robots and children’s trust of and learning from them. Specifically, we demonstrate how children’s understanding of robots directly impacts their relationships and interactions with them.

142  Kimberly A. Brink and Henry M. Wellman

1.  Children’s Developing Understanding of Robots It is relatively easy to imagine ways that children’s own understanding of robots could impact their feelings about them and relatedly their willingness to trust and learn from them. Children may recognize that there are similarities between robots and other familiar categories (i.e., people, animals, or artifacts) and feel comfortable with them. Contrastingly, children may perceive robots as altogether unfamiliar and unique categories that either require further exploration or elicit distrust. Furthermore, these expectations about robots could, in turn, affect children’s willingness to learn from and trust robots. If children perceive robots as similar to infallible tools, like calculators or dictionaries, they may indiscriminately accept information from all robots. Or if children perceive robots, especially humanoid robots, as similar to people with thoughts, emotions, and sometimes flawed beliefs about the world, they may differentially accept information from only trustworthy ones. Alternatively, if a child perceives robots as entirely different from humans and other common sources of information, they may ultimately consider robots to be suspect and never trust information from them. To embrace and learn from robots, children likely must first determine what they believe about robots: are they infallible tools, intelligent beings, or something else entirely? The nature of robots, however, is not easy to define. Robots are unique devices that can simultaneously share similarities with artifacts, animals, and even humans. They, like other artifacts, are designed and built by humans. They do not live, grow, breathe, or (as adults typically believe) experience feelings (Gray & Wegner, 2012). Therefore, children may perceive robots as artifacts or infallible tools similar to textbooks or other educational devices. Yet, unlike more common artifacts such as books, bikes, beds, and balls, social robots are not only human-​built but humanlike to varying degrees. They can look, behave, and at times even “think” like humans or animals. Hence, children may alternatively see them as intelligent beings that can think and reason. As such, robots, like humans, might be fallible and should be considered carefully before trusting them. Indeed, research on children’s developing understanding of robots reflects children’s attempts to rationally and emotionally embrace and resist these competing identities of robots. Research on early childhood understanding of robots suggests that children initially attribute a myriad of humanlike qualities to robots. For instance, they expect robots to have emotional, social, as well as perceptual

Technology as Teacher  143 abilities. Young children report that robots have perceptual abilities, like sight and touch: three-​year-​olds claimed that a robot dog could see and be tickled (Jipson & Gelman, 2007). Nine-​and 12-​year-​olds similarly reported that a three-​foot-​tall interactive robot, Robovie, could be intelligent, have interests, and experience emotions (Kahn et al., 2012). These children also believed that Robovie could be their friend and could comfort them if they were sad. Whereas young children appear to treat and think about robots like people or animals, they, however, do not equate robots with them. Children recognize that a robot dog is not identical to a real dog, as demonstrated by their claims that robot dogs do not have biological qualities (Melson et al., 2009): young children report that robotic dogs cannot grow or eat like real dogs for example (Jipson & Gelman, 2007). As children age, however, their beliefs about robots change—​expectations that robots have emotional, social, and perceptual capacities decrease. Older children are less likely to report that robots have emotions, desires, or are capable of autonomous action (Mikropoulos, Misailidi, & Bonoti, 2003). Five-​year-​olds were less likely to claim that a robot dog could think or feel happy compared to three-​year-​olds (Jipson & Gelman, 2007). Children older than seven spoke differently about robots, more often using language specific to man-​made machines, compared to children younger than seven (Okita, Ng-​Thow-​Hing, & Sarvadevabhatla, 2011). Fifteen-​year-​olds were also less likely to believe that Robovie could have interests, experience emotions, or be a friend compared to nine-​and 12-​year-​olds (Kahn et al., 2012). We similarly show this relation in our own research (described in the next section):  children’s reports of robots’ perceptual abilities and psychological abilities decline across three to 18 years of age. With age, children begin to dissociate psychological, emotional, social, and perceptual abilities from robots and recognize that robots are more similar to artifacts. Finally, by adulthood, beliefs about the capacities of robots have decreased to a substantially reduced set of expectations. Adults essentially expect that robots are only capable of some forms of thinking and decision-​making, and they deny robots the ability to feel pain or fear (Gray & Wegner, 2012). Given this sort of evidence, we argue that there is a transition between how children and adults think about robots. Young children attribute more social, psychological, and perceptual abilities to robots. Yet, with age and experience, children gradually adjust their beliefs about robots so that, by adulthood, robots possess only limited psychological agency, that is, the ability to think and make decisions. We argue that this change in understanding of

144  Kimberly A. Brink and Henry M. Wellman robots impacts children’s interactions with and feelings toward robots. Our research, described next, speaks to this transition and its impact on children’s feelings toward robots.

2. Social Robots: The New Childhood Companion Over the past few years, a number of in-​home robots have been released that are designed specifically to teach, supervise, and play with children in their homes. Robots have also been steadily entering classrooms and hospitals with the purpose of teaching and comforting children. One might expect that for robots to be effective in these roles they should, at a minimum, promote positive feelings and comfort children. In fact, several studies show that robots can make children feel safe and comfortable. Children have reported that some in-​home robots, like Aibo the robot dog, could make them feel safe if they were home alone with it (Weiss, Wurhofer, & Tscheligi, 2009). Other children have claimed that even a simple robotic arm could be a friend that they would feel comfortable sharing secrets with (Fior et al., 2010). When receiving injections at a hospital, children with severe anxiety were comforted more by a playful Nao robot than a human nurse (Beran et al., 2013). What remains unknown, however, are the specific features and actions of robots that encourage children to feel comfortable with robots and which features detract.

2.1.  The Uncanny Valley Research with adults may be able to shed some light on this question. Decades of research with robots reveals that adults prefer somewhat humanlike robots but find very-​humanlike robots unnerving—​a phenomenon called the “uncanny valley” (MacDorman et al., 2009; Mara & Appel, 2015; Mori, MacDorman, & Kageki, 2012). As a robot appears increasingly humanlike, it becomes more attractive until it reaches a threshold where the robot is considered too humanlike and becomes uncanny and creepy (see Figure 8.2). It is this dip in affinity for very humanlike robots that is called the uncanny valley. Closely humanlike robots are distinctly creepier than other robots and, in particular, creepier than even the more unsettling of machine-​ like robots. For example, humanlike robots such as Kaspar, a childlike

Technology as Teacher  145 + Healthy person

Affinity

Anthropomorphized robot

Machine-like robot Industrial robot

Human-likeness

Stuffed animal Doll

100%

50% Corpse

–

Prosthetic hand

Human-like robot Uncanny Valley Moving

Still

Figure 8.2  A schematic depiction of the theoretical/​hypothetical uncanny valley (figure closely derived from Figure 2 in Mori et al. 2012). The uncanny valley is defined as the precipitous dip in affinity for closely human-​like robots.

robot, and Telenoid, a realistic ghostlike robot, as depicted in Figure 8.3, are considered creepy and unsettling by adults (Gray & Wegner, 2012; Mara & Appel, 2015). Support for the uncanny valley comes from many studies in which adults report feeling greater unease when presented with extremely humanlike robots compared to other less humanlike robots (Gray & Wegner, 2012; MacDorman, 2006). These reports of creepiness have been linked to not only humanlikeness more generally, but more specifically to beliefs about the humanlike abilities of robots. One theory proposes that, for humans, the uncanniness of very humanlike robots results from violations of an acquired everyday understanding of what makes humans distinct from machines (MacDorman & Ishiguro, 2006). In the case of robots, for example, when a robot closely resembles a thinking and feeling human being, this would violate expectations that machines should be

146  Kimberly A. Brink and Henry M. Wellman

Figure 8.3  Kaspar (left panel) and Telenoid (right panel) are two robots with humanlike features that adults consider unsettling, unnerving, and uncanny according to empirical research.

inanimate and hence incapable of thought and experience. Specifically, a very humanlike appearance in a machine, like in Kaspar and Telenoid, can prompt attributions of a humanlike mind (Epley, Waytz, & Cacioppo, 2007), and as humanlike minds are seldom ascribed to robots (Gray, Gray, & Wegner, 2007), this mismatch causes feelings of uncanniness (Gray & Wegner, 2012). Indeed, research with adults reveals that the more robots are seen to have human feelings, the more unnerving they seem (Gray & Wegner, 2012). Violations of expectations about the behavior and appearance of machines and humans thus link to uncanny valley experiences in adults.

2.2.  Children’s Perceptions of Robots Impacts Feelings toward Robots What about children? Do they, or at what ages do they, have adult-​like feelings and reactions to robots? These questions are surprisingly unstudied.

Technology as Teacher  147 Considering that children’s expectations about robots can and do differ substantially from those of adults, we might infer that children’s feelings toward robots may also differ according to their expectations and beliefs about robots. Children start with the expectation that robots are capable of several humanlike abilities—​social, perceptual, and psychological capacities—​that adults typically do not attribute to robots. Therefore, unlike for adults, perhaps when children perceive a robot as similar to a person, which is consistent with their expectations about the humanlike nature of robots, they consider that robot more familiar and comforting. As children’s expectations about mental abilities of humanlike robots decline with age, however, we might then expect this comfort with humanlike robots to decline. To investigate this and other hypotheses, we interviewed 240 children from three to 18 about their perceptions of the psychological abilities and the uncanniness of two robots (one closely humanlike and one machine-​like; see Figure 8.4) (Brink, Gray, & Wellman, 2017). We used these specific stimuli because they have previously been validated with adults to capture the uncanny valley phenomenon (Gray & Wegner, 2012). Children rated their feelings toward one of the two robots (uncanniness) and also rated its mental capacities, including the ability to think (agency) and feel (experience), as in past work with adults (e.g., Gray & Wegner, 2012). First, we found that the uncanny valley develops: younger children found the closely humanlike and machine-​like robot equally not very creepy, whereas older children found the closely humanlike robot much creepier than the machine-​like robot—​similar to adults. Second, we identified the approximate age at which the uncanny valley emerges. Differences in feelings about the two robots emerged progressively over age, but it was not until middle childhood, starting at approximately nine years of age, that children

Figure 8.4  Still frames from videos used in our research assessing the uncanny valley phenomenon in children: Kaspar from the back (left panel), Kaspar from the front (center panel), and Nao (right panel)

148  Kimberly A. Brink and Henry M. Wellman had a greater uncanny response to a closely humanlike robot compared to a machine-​like  robot. And third, critically, we found that children’s perceptions of mind (an aggregate of both attributions of agency and experience) were correlated with this change in uncanny responses. For younger children, increasing perceptions of mind tended to predict decreased uncanniness. Young children preferred robots that they perceived to have humanlike mental capacities, i.e., robots that seemingly had the ability to think and feel. These data are consistent with a developmental extension on the theory that the uncanny valley only occurs when expectations about robots are violated. As young children seemingly expect some robots to have mental abilities (according to previous research, and now our own research as well), then the perception of a humanlike mind may be a welcome familiarity for them and therefore be considered not very uncanny. Young children in our research found robots to be more pleasing (less uncanny) when they perceived the robots to have more mental abilities. From younger children to older children, however, the association between mind and feelings of uncanniness begins to shift. For younger children, attributions of mind predicted decreased feelings of uncanniness, whereas, for older children, attributions of mind no longer predicted a decrease in uncanny feelings and instead trended to predict an increase in uncanny feelings. The broader literature indicates that this correlation between attributions of mind continues to increase and indeed becomes distinctively positive for adults (Gray & Wegner, 2012). We argue that this shift occurs when children develop richer understandings of folk biology and folk psychology, and begin to separate the concepts of minds, brains, bodies, and machines (Wellman, 2014). For example, it is only at about 9–​12 years that children truly understand differences between the mind (as more “mental”) and the brain (as more part of the biological body; Johnson & Wellman, 1982; Richert & Harris, 2006). This understanding that the mind stems from the biological brain (i.e., a neurophysiological “machine”) could support the development of the uncanny valley: the uncanny valley may result from the mismatch of perceiving a human mind as stemming from a machine brain. It may not be until this later age that children develop an understanding that robots, as machines, should not have minds as humans do, making them uncanny when they seem like they do. Whereas our study is the first to speak to the uncanny valley in childhood, our results do not speak to degrees of the uncanny valley. The classic uncanny

Technology as Teacher  149 valley proposal is that a preference for increasingly humanlike robots follows a nonlinear curve (Mori et al., 2012) as in Figure 8.2. We, however, used only a binary comparison of machine-​like to very humanlike robots (consistent with Gray & Wegner, 2012). Future research with children should explore the full range of the uncanny valley’s trajectory and investigate children’s responses to a broad range of robots that vary more systematically in their levels of humanlikeness. From our research, it is possible that a particular humanlike robot could exist that would be considered very creepy by young children. Or if such a humanlike robot in fact does not exist, it is possible that young children may still find some robot (with very different or nonhuman features) very creepy. This leads to two points: First, as we outline in our final section, future research should explore children’s feelings and perceptions about a variety of robots much more broadly and systematically. Second, as investigations into the uncanny valley center around the construct of “humanlikeness,” a construct which has never been fully defined in the literature, future research should operationalize “humanlikeness” with a broader spectrum of robots. The uncanny valley has historically been argued to result directly from a robot’s level of humanlikeness. Some writings imply a clear distinction between androids (synthetic robots designed to look and act human, especially, in science fiction, one with a body having a flesh-​like “skin”) versus humanoid robots (with humanlike shape and behavior, like Star Wars’ C-​3PO). Nevertheless, there is no agreed-​upon terminology or objective test for “degrees” of humanness. In most studies of the uncanny valley, a robot’s level of humanlikeness is defined a priori by an experimenter (e.g., Gray & Wegner, 2012; Mori et al., 2012) or measured by asking participants to directly report how humanlike a robot is (MacDorman, 2006). Rather than this one-​dimensional perspective, however, we believe that robots can be more or less humanlike in several (or many) different multidimensional ways. Robots could be closely humanlike in face, limbs, behavior, language, and more. Future research would do well to investigate which of these features of humanlikeness is considered creepy and at what ages. And given our data, an important question would be which features are tied to mind. Future research examining a larger variety of robots across childhood is clearly needed. Relatedly, in what follows, we describe other research that has used what their authors term “humanlike” robots. When describing this research, we attempt to define the specific features and dimensions of the robot (voice, appearance, behavior, etc.) that are considered distinctly humanlike. We

150  Kimberly A. Brink and Henry M. Wellman outline how such robots are often effective with children for some task or another. But, as of yet, we do not know from these studies the extent to which those results are specific to a small specially designed subset of robots, or applicable more widely. This is an obvious limitation of the existing research, and our terminology for what constitutes “humanlike.” Additionally, because our results demonstrate this shift in feelings toward and perceptions of robots occurs at an (inexact) boundary between early and middle childhood, we will frame our remaining developmental claims focusing on differences in behavior and understanding between three age groups: younger (up to seven or eight or so) and older (nine and 10 and above) children plus adults.

3.  Social Robots: The New Teacher Above and beyond their roles as companions and caretakers, where robots need to comfort and connect with children, a central focus of robot design has been to develop them as educational tools to address growing concerns in the field of education. In the United States, for example, schools are facing significant shortages in teachers. Enrollments in teacher-​preparation programs have been declining for over 10 years and are worsening (Sawchuk, 2014). Moreover, 46% of all new teachers leave the profession within only the first five years of teaching (Brill & McCartney, 2008). As one remedy to this problem, engineers are recommending the implementation of robots to assist overburdened teachers (Han & Kim, 2009). Classrooms around the world are progressively utilizing educational robots to support and even substitute for teachers (Han et al., 2005). Already schools are employing socially assistive robots as a low-​cost means to alleviate teacher workload. This trend is not limited to the United States, to schools for older children, or to educational endeavors for typically developing children. In Korea, Robosem teaches English to children and, in the United States, RUBI teaches Finnish to toddlers (Powell, 2014). Robots are teaching children not only language skills (Movellan et al., 2009), but also mathematics (Wei et al., 2011) and science (Hashimoto et al., 2013). They are even working with children with special needs:  Ursus, a large robotic bear, administers physical therapy to children with motor disorders like cerebral palsy and brachial plexus palsy (Mejías et al., 2013). And several child-​ and animal-​like robots are helping children with autism spectrum disorder

Technology as Teacher  151 (ASD) practice and engage in social interactions through imitation games, turn-​taking, and conversation (Ricks & Colton, 2010). Robots are increasingly becoming a part of the daily lives of children and expected to convey and pass on knowledge to them. As robots continue to enter the classroom, and other learning spaces, it is crucial to identify the factors that improve children’s willingness to trust and accept information from these robots appropriately. While children’s expectations and beliefs about robots impact their feelings toward these robots, we argue that these beliefs additionally impact their willingness to trust information that they receive from robots as well. To understand why this might be the case, we pull from the literature on the factors that affect children’s willingness to learn from human instructors.

3.1. Natural Pedagogy Children learn a great deal about their world through knowledge passed on by their parents, teachers, and peers. Children trust that 8 × 8 = 64, that the earth is round, and dinosaurs are extinct, not because they have uncovered these facts themselves, but because reliable sources have told them so. Those researching this phenomenon, from either “trust in testimony” or “natural pedagogy” perspectives, have proposed that children are adapted to learn gen­eral knowledge through human communication (Csibra & Gergely, 2009). When children learn from other people, at least by early preschool ages, they monitor cues such as access to information, expertise, and confidence: young children are sensitive to whether a person has access to the right kind of information (e.g., did she see the thing she is telling me about?); they attend to the person’s qualifications (e.g., is she a knowledgeable adult or a naive child?); and they monitor a person’s confidence (or uncertainty) in their answers (e.g., did she say she knows it’s an X or thinks it’s an X?) (Koenig & Harris, 2007). Even infants utilize social signals, like direct eye contact and contingent interactions, when determining whether they should accept the information they receive from an informant (Csibra & Gergely, 2011). In total, these cues suggest that children focus on the knowledge and social roles of their teachers when determining whether they should accept information from them. Children appear to trust teachers that have psychological agency: the ability to think, make decisions, possess knowledge, and respond interactively.

152  Kimberly A. Brink and Henry M. Wellman Children can attribute psychological agency to robots as well, and prior research demonstrates that features of the robot can encourage attributions of agency. Specifically, humanlikeness and contingent social behaviors have both been empirically linked to perceptions of agency. For instance, infants and young children associate humanlike features, such as faces and eyes, with agency (Hamlin, Wynn, & Bloom, 2007; Jipson & Gelman, 2007). Further, when entities respond contingently with social behaviors, young children are also more likely to attribute agency to that object (Johnson, 2000). For example, young children are more likely participate in social activities with a robot when it contingently rocks or moves at the same rhythm that they do (Carey & Markoff, 2010). And, indeed, young children also seem more willing to accept information from technology when it demonstrates behaviors consistent with psychological agency, as children do with human instructors. For example, children learn better and pay closer attention to technological agents that interact contingently. When learning about novel animals from two furry stuffed-​animal-​like robots, three-​to five-​year-​olds were more likely to ask for information and agree with a robot that behaved contingently with their own actions compared to a robot that did not (Breazeal et al., 2016). When learning new vocabulary words, even infants were more likely to attend to a contingent human instructor compared to an inanimate audio speaker (Koenig & Echols, 2003). Four-​to six-​year-​olds were also more accurate when a robot used an interactive and contingent teaching style that required the children to perform tasks as a team with the robot (Okita, Ng-​Thow-​ Hing, & Sarvadevabhatla, 2009). Further, young children are more likely to learn from robots that share similarities with humans. For instance, in a table-​setting task, four-​to 10-​ year-​old children learned best from a robot when that robot’s voice, appearance, or behavior resembled that of a human (Okita et al., 2009). When the robot taught children the functions and locations of different dining utensils, children who learned from a robot with a humanlike voice recalled more when tested later compared to children who learned from the exact same robot only with a machine-​like voice. Children have also demonstrated improved learning when interacting with robots that respond contingently and socially with them. Children showed improved English speaking skills after practicing a foreign language with a robot that could display emotions and make humanlike gestures, like winking, nodding, and yawning (e.g., Lee et al., 2011; Saerbeck et al., 2010). In another study, children learned English

Technology as Teacher  153 better from an interactive robot that could emote, sing, talk, and dance compared to a non-​interactive computer or textbook with identical instructional material (Han et al., 2008). Children appear to learn best from robots with characteristics that have been empirically linked to agency, including humanlikeness and contingent social behavior. Although research has demonstrated young children’s willingness to trust teachers who seemingly demonstrate psychological agency, the impact of psychological agency on children’s trust in testimony has not been investigated directly. This is probably because human teachers all clearly express very high amounts of psychological agency. Social robots therefore provide a needed opportunity to vary psychological agency in a clear and empirical fashion.

3.2.  Understanding of Robots Impacts Learning from Robots in Young Children To investigate children’s willingness to trust and mistrust information received from robots and, moreover, the effect of agency on children’s learning, we asked nearly 60 three-​year-​olds to listen to the testimony of two Nao robots (see Figure 8.5) (Brink & Wellman, submitted). We investigated children’s reactions to Nao because in our uncanny valley research, not only do young children find this robot substantially not-​creepy, but they also attribute a substantial amount of psychological agency to this robot. In our “trust in testimony” research, we investigated children’s willingness to accept information from this robot and at the same time measured how much agency children attributed to it. In our initial study, three-​year-​old children were taught the names of novel objects by two humanoid Nao robots: one that had been previously shown to be always accurate and the other always inaccurate. First an actress asked the two robots to name four familiar objects (e.g., a ball, a teddy bear, etc.) and one robot correctly named all four while the other incorrectly named all four. She then asked the robots for the names of four totally unfamiliar novel objects and each robot provided a different unfamiliar (made up) name for each object. We asked children (1) which robot was not very good at answering questions, (2) which robot they would like to ask for the name of the novel object, and (after both robots named the novel object) (3) what the child thought the correct name of the object was. By assessing

154  Kimberly A. Brink and Henry M. Wellman

Figure 8.5  Still frame of video stimuli from initial study investigating whether children appropriately trust (and mistrust) the testimony of social robots

whether children were willing to agree with and ask for information from an accurate robot but not an inaccurate robot, we could determine whether children appropriately trust (and mistrust) robots. Finally, immediately following the trust in testimony task, we asked children whether they believed the robots had psychological agency using the same interview validated in our uncanny valley research. By simultaneously measuring whether children appropriately trust robots and their attributions of psychological agency to robots (e.g., can the robot think for itself or decide what to do), we could assess whether children’s perceptions of agency related to their willingness to trust robots. We found that, in general, preschool-​aged children appropriately trusted the more accurate robot. But, importantly, we also found that children who attributed more psychological agency to the robots were also more likely to seek out new information and endorse the accurate robot. Three-​year-​olds were more likely to correctly identify which robot was not very good at answering the questions, ask the accurate robot for the name of the object, and endorse the name of the novel object given by the accurate robot when the robots were perceived to have psychological agency. Note that these results also demonstrate that these young children learned from the accurate robot, at least in the minimal sense of learning from it the name of a novel object.

Technology as Teacher  155 In a follow-​up study, we further investigated the impact of psychological agency on children’s willingness to trust technological devices by having 45 children learn from two machines that showed no signs of psychological agency. Unlike the humanoid, interactive Nao robots used in the first study, we showed children inanimate machines. These machines were amorphous blob shapes lacking any humanlike features such as faces or limbs (see Figure 8.6). They also did not respond contingently to the actress presenting the familiar and unfamiliar objects. The machines only played an audio recording naming the objects when a person off-​camera was signaled to activate the toys by pulling a string on each machine. When signs of agency were almost entirely eliminated from machines, in this second study, we found that children were at chance when deciding to ask the accurate machine for the name of the unfamiliar object and for selecting the name provided by the accurate machine. Children were accurate in identifying which machines emitted the correct and incorrect names for familiar objects but they did not utilize the machines’ accuracy to decide which machine to trust and learn novel names from. That is, children randomly selected which machine to ask for the novel names and selected randomly which name—​from the previously accurate or inaccurate device—​they thought applied to the novel objects. Arguably, for children, it

Figure 8.6  Still frame of video stimuli from second study investigating whether children appropriately trust (and mistrust) the testimony of two inanimate machines with no signs of psychological agency

156  Kimberly A. Brink and Henry M. Wellman is important that an informant demonstrate the ability to think and make decisions in addition to being accurate.

3.3.  Developmental Trajectory of Perceived Agency on Learning As we have demonstrated throughout this chapter, children’s changing expectations and beliefs about the mental abilities of robots can and do impact their feelings toward and interactions with robots. Our sets of studies demonstrate the importance of psychological agency in determining whether young children trust social robots. However, our research as well as the broader literature demonstrate that older children think and feel differently about the psychological agency of robots. Whereas younger children in our studies were more likely to prefer and learn from robots with psychological agency, older children did not prefer these robots. Therefore, it is reasonable to expect that older children learn differently from instructional robots with agency compared to younger children. Indeed, existing research on the effectiveness of robot teachers for older children suggests that robots with more signals of agency, although effective educators for young children, show little or no improvements in learning for older children. Several studies demonstrate that signs of agency in robots are less effective for encouraging learning in older age groups compared to younger age groups. In the table-​setting task described earlier, younger children learned better from a robot with a more socially interactive and communicative style of teaching than older children (Okita et al., 2009). Children four to six years old improved so dramatically when taught by a robot with an interactive teaching style that they performed on par with seven-​to 10-​ year-​olds. There was no improvement in learning, however, for interactive robot-​teaching styles for the seven-​to 10-​year-​olds compared to other less agent-​like teaching styles. Further, while a humanlike voice was better for accuracy in general, the effect was more pronounced for younger children. Older children did not show as much improvement as younger children when learning from a robot with a humanlike voice compared to a monotone voice. In another study, the robot Robovie traveled around a Japanese school to speak English with first-​graders (six-​to seven-​year-​olds) and sixth graders (11-​to 12-​year-​olds) (Kanda et al., 2004). This robot performed interactive behaviors like hugging, shaking hands, playing rock-​paper-​scissors, singing,

Technology as Teacher  157 briefly conversing, and pointing at nearby objects. First graders spent significantly more time interacting with the robot than sixth graders did. Features of agency appear to be a more important factor for younger children when interacting with and learning from robots. Again, these studies with older children, while informative, have not directly measured the impact of attributions of agency on their learning. More research is needed to evaluate how robots’ features and children’s developing cognitive abilities interact to produce improved learning in children from robots. We have only begun to conduct such research. Some of that research evaluates the relationship between certain features of robots, and children’s perceived overall sense of the robot as well as their sense of the agency of the robot. Ultimately, beyond these initial studies, future studies should investigate how children’s understanding of robots impacts their perceptions of agency, the “worthiness” of robots as sources of information, and consequently the quality of children’s interactions with and learning from robots as children grow older.

4.  General Conclusions and Further Considerations Children increasingly encounter robots designed to comfort, teach, and play with them. It is therefore imperative to learn which robot features and child-​robot interactions impact these interactions and outcomes. And, importantly, it is imperative to identify how and when these interactions and outcomes change throughout childhood. For these and many other reasons, developmental research is badly needed to assess how children’s cognitive abilities, their developmental trajectories, and the design of robots all work together to impact children’s learning and feelings toward robots. Social robots are only one form of interactional, educational smart technology that may (or may not) promote these desired outcomes, but their special features and their increasing presence in children’s lives make them worthy of study in their own right. Moreover, they allow for larger lessons on children’s increasingly voluminous interactions with smart technological devices more generally. Clearly, from our review and research, children’s understanding of the mental capacities of robots appears as an important factor in their feelings toward and willingness to learn from robots. We found that as younger children perceived robots to have more mental abilities, they experienced decreased

158  Kimberly A. Brink and Henry M. Wellman uncanny feelings toward those robots. However, as children age and gain more experience with technology, their feelings about and perceptions of robots change and their perceptions of the mental capacities of robots differentially impact their feelings toward them—​older children do not prefer robots that appear to have mental abilities and adults find machines with closely humanlike minds and appearances to be distinctively unsettling. Further, according to the broader literature, whereas signals of psychological agency in robots impact younger children’s willingness to learn, they do not appear to impact older children’s willingness to accept information from those devices. In total, our research as well as others’ has only just begun to demonstrate how children’s beliefs and expectations about robots, their developmental trajectory with respect to this understanding, and the robot’s design together impact how children will feel toward and learn from robots. Future research should continue to explore each of these components of child-​robot interactions with more precision and within different contexts. For instance, research should more fully identify the fine-​grained developmental changes that occur in children’s conceptual understanding of and interactions with robots. Our research demonstrates that not only do children’s beliefs about technology change with age but also that this changing understanding directly impacts children’s interactions with this technology. For young children, when they perceived robots to have psychological agency, they used the robots’ accuracy to determine which robot to trust for new information about novel objects. Young children expect and even prefer robot instructors that appear to have minds of their own, whereas older children do not expect robots to have minds and are indifferent to interactive and agent-​like robot instructors. The research thus far, however, only captures the emergence of these developmental changes on a very crude timeline, before and after approximately nine to 12 years of age. One thing needed is more extended data (as in our study of uncanniness) of a range of ages that help us and the field to move beyond the description of “younger” and “older” age groups to a more fine-​grained set of developmental findings and hypotheses. In line with this more comprehensive approach for exploring developmental changes in children’s interactions with robots, future research should also investigate and clearly define the mechanisms that produce the child-​ adult differences that we identify here. The child-​adult differences we have uncovered may very well be the result of developments in basic aspects of cognition as we have discussed previously. Or they could result from younger

Technology as Teacher  159 children’s increased interactions with more sophisticated and social technology on a regular basis compared to older children and adults (Bernstein & Crowley, 2008). That is, today’s young children may grow up to look quite different from current adults because of different experiences with increasingly different technological devices, including robots themselves, from adults and older children. It remains unclear how children’s cognitive development in addition to their experiences with robots and smart technologies may interact to influence their interactions with these devices. It would be well worth following children (for example those in our sample) longitudinally to examine such intriguing possibilities. As we continue to more precisely explore children’s developmental trajectories and the mechanisms that determine how children will interact with technology, we can and should also consider child-​robot interactions in special populations. Not only are typically developing children encountering new and different technology daily, but so too are children with special needs. Children in these populations bring their own set of unique experiences and cognitive abilities to their interactions with robots that may produce different outcomes from typically developing children. Indeed, research with children with ASD shows that they prefer robots that are the least humanlike (Robins, Dautenhahn, & Dubowski, 2006), unlike the children in our research, who preferred the humanoid Nao robot over the mostly machine-​like robot shown in Figure 8.4. Given that robots are participating in educational endeavors with children with special needs, future studies should investigate the impact of child-​specific abilities and experiences on child-​robot interactions in these special populations. In addition to more fully exploring the impact and developmental trajectory of child-​specific features on child-​robot interactions, future research should also more extensively assess how these interactions are affected by different robot designs and features. Our investigations into children’s interaction with robots indicate that features of the robot can inspire attributions of mind and thus impact children’s learning from those robots. Therefore, the field would be greatly improved by more extensive research exploring how a broad range of robots and their features impact children’s feelings and perceptions. The studies we have conducted and been able to assemble almost exclusively focus on a small subset of robots that vary primarily in their similarities to people. They all have similar body types to people, with legs and arms, and differ primarily in minor features of their exterior appearance, like whether they have a skin-​like or hard plastic coating, for example.

160  Kimberly A. Brink and Henry M. Wellman However, robots can differ substantially from this basic body structure and design and instead imitate animals, cars, books, etc. They can even appear as entirely new and unfamiliar entities. It will be important for future research to investigate how children respond to a broad range of robots that vary substantially in appearance and behavior and moreover how development impacts children’s interactions with this broad range of robots (not just a few). As one example of this need, our studies have only assessed the educational effectiveness of two robot/​machine devices, one with several cues of agent-​like abilities (i.e., contingent responses, referential speech, and abstractly humanlike features) and the other which evidenced none of these cues whatsoever. However, technology can and does vary drastically in the quality and types of agent-​like characteristics it embodies. Instructional iPads in classrooms can interactively display real teachers or animated avatars. Artificially intelligent assistants, like Apple’s Siri or Amazon’s Alexa, can respond contingently and interactively answer questions. Our studies, however, do not speak to the effectiveness of such devices with varying levels of agency. Future studies should examine the effectiveness of devices that more systematically vary in the quantity and quality of features that encourage attributions of agency. The importance of robot design in child-​robot interactions also highlights the informative potential of collaborative investigations between engineers, designers, and social scientists. Robot design currently, even for children, has primarily been in the hands of engineers and corporate design teams. These designs are inspired, and limited, by the intuitions of these individuals and what they assume would be child-​friendly and effective for children. This is similarly true for those robots increasingly found in movies for children. But there is currently little collaboration between developmental researchers, engineers, designers, and movie animators. However, collaborations could aid both groups. On one hand, developers and animators have their own marketing and research teams undoubtedly collecting audience reactions and product development data which have been used to shape and develop their ideas and designs. Access to such information could help psychological researchers better understand the favorable and unfavorable reactions to various robot designs, including possibly the reactions of children of different ages. On the other hand, researchers could aid designers by sharing work on the interactions between children’s developmental trajectories and

Technology as Teacher  161 robot design that would impact children’s social and educational outcomes. For example, from our research we would likely conclude that designers should consider the effect of utilizing a variety of agent-​like cues (e.g., contingent behaviors or abstractly humanlike features) when designing educational robots. A collaboration among these groups could drastically expand our understanding of children’s conceptual and cognitive development and simultaneously improve robot design for children by implementing more evidence-​based decisions. The rising number of robots interacting with children also has important implications for a variety of other developmental outcomes that we have not yet addressed in this chapter. Thus far, we have focused primarily on a small set of outcomes—​how children feel toward robots and how children learn from robots—​but social robots also have the potential to impact both children’s social and moral development. There is growing concern in the field that increased interactions with smart technologies, including social robots, might adversely impact children’s social and moral development. Smart technology and social robots are frequently blamed for decreasing in-​person social interactions, preventing children from learning how to effectively interact with others, and thus hindering children’s social development (Turkle, 2011). Researchers, parents, and teachers are particularly concerned that interactions with robots will promote the development of antisocial behaviors. In fact, there is some evidence for this. A hitchhiking robot that had successfully traveled around Germany, Canada, and the Netherlands taking pictures and carrying on conversations with other travelers was eventually vandalized and destroyed a few weeks into its US journey (Victor, 2015). A mall security robot designed to share information with customers routinely faced abuse from unsupervised children as they would often kick and push the robot (Brscić et al., 2015; Nomura et al., 2017). Nevertheless, empirical research suggests that these antisocial behaviors toward robots can be reduced and, moreover, that social robots can even be used to promote positive social development for certain populations. Researchers have found that, while abusive behavior toward robots does occur, a few behavioral modifications to robots can reduce this behav­ ior. Preschool children in a classroom comforted a robot with a hug and protected it from other aggressive children when it started to cry after being damaged or played with too roughly (Carey & Markoff, 2010). Other studies

162  Kimberly A. Brink and Henry M. Wellman show that children claim that a robot deserves to be treated fairly and not psychologically harmed after conversing and playing with the robot for 15 minutes (Kahn et al., 2012). Social robots have also been used to aid children with ASD in social development by practicing social behaviors, like conversation, with them and demonstrating common social cues and behaviors (Ricks & Colton, 2010). Future studies should continue to investigate the complex relationship between children’s perceptions of robots, how they treat them, and ultimately how these interactions impact children’s later social and moral development and their interactions with others. Every year robots increasingly become a part of our lives and the lives of children. The National Robotics Initiative foresees a future in which “robots are as commonplace as today’s automobiles, computers, and cell phones. Robots will be found in homes, offices, hospitals, factories, farms, and mines; and in the air, on land, under water, and in space (“National Robotics Initiative 2.0:  Ubiquitous Collaborative Robots (NRI-​2.0),” 2017). Investigations into children’s understanding of and interactions with these devices is therefore only becoming more important. The future of research into child-​robot interactions should more fully explore what children think and feel about a wider variety of robots, how they interact with them in a variety of contexts, and also expand those investigations to explore a wider variety of developmental outcomes for children, including educational, social, and moral outcomes.

Acknowledgments Preparation of this chapter was supported in part by a grant from the Varieties of Understanding Project at Fordham University and the John Templeton Foundation.

References Beran, T. N., Ramirez-​Serrano, A., Vanderkooi, O. G., & Kuhn, S. (2013). Reducing children’s pain and distress towards flu vaccinations: A novel and effective application of humanoid robotics. Vaccine, 31(25), 2772–​2777. doi:  http://​dx.doi.org/​10.1016/​ j.vaccine.2013.03.056. Bernstein, D., & Crowley, K. (2008). Searching for signs of intelligent life: An investigation of young children’s beliefs about robot intelligence. Journal of the Learning Sciences, 17(2), 225–​247. doi: http://​dx.doi.org/​10.1080/​10508400801986116.

Technology as Teacher  163 Breazeal, C., Harris, P. L., DeSteno, D., Westlund, K., Jacqueline, M., Dickens, L., & Jeong, S. (2016). Young children treat robots as informants. Topics in Cognitive Science, 8(2), 481–​491. Brill, S., & McCartney, A. (2008). Stopping the revolving door: Increasing teacher retention. Politics and Policy, 36(5), 750–​774. Brink, K. A., Gray, K., & Wellman, H. M. (2017). Creepiness creeps in: Uncanny valley feelings are acquired in childhood. Child Development. Brscić, D., Kidokoro, H., Suehiro, Y., & Kanda, T. (2015). Escaping from children’s abuse of social robots. Paper presented at the Tenth Annual ACM/​IEEE International Conference on Human-​Robot Interaction. Carey, B., & Markoff, J. (2010). Students, meet your new teacher, Mr. Robot. New York Times. Retrieved from http://​nyti.ms/​1H9yiEN. Csibra, G., & Gergely, G. (2009). Natural pedagogy. Trends in Cognitive Sciences, 13(4), 148–​153. Csibra, G., & Gergely, G. (2011). Natural pedagogy as evolutionary adaptation. Philo­ sophical Transactions of the Royal Society B: Biological Sciences, 366(1567), 1149–​1157. Epley, N., Waytz, A., & Cacioppo, J. T. (2007). On seeing human: A three-​factor theory of anthropomorphism. Psychological Review, 114(4), 864. doi: http://​dx.doi.org/​10.1037/​ 0033-​295X.114.4.864. Fior, M., Ramirez-​Serrano, A., Beran, T., Nugent, S., & Kuzyk, R. (2010). Children’s relationships with robots:  Robot is child’s new friend. Journal of Physical Agents, 4(3),  9–​17. Glaser, A. (2016). Jibo is like Alexa and a puppy inside one adorable robot. Retrieved from http://​www.wired.com/​2016/​06/​jibo-​like-​alexa-​puppy-​inside-​one-​adorable-​robot/​. Gray, H. M., Gray, K., & Wegner, D. M. (2007). Dimensions of mind perception. Science, 315(5812), 619–​619. doi: http://​dx.doi.org/​10.1126/​science.1134475. Gray, K., & Wegner, D. M. (2012). Feeling robots and human zombies: Mind perception and the uncanny valley. Cognition, 125(1), 125–​130. doi:http://​dx.doi.org/​10.1016/​ j.cognition.2012.06.007 Hamlin, J. K., Wynn, K., & Bloom, P. (2007). Social evaluation by preverbal infants. Nature, 450(7169), 557–​559. Han, J., Jo, M., Park, S., & Kim, S. (2005). The educational use of home robots for children. Paper presented at the IEEE International Workshop on Robot and Human Interactive Communication, 2005, RO-​MAN 2005. Han, J., Jo, M.-​H., Jones, V., & Jo, J.-​H. (2008). Comparative study on the educational use of home robots for children. Journal of Information Processing Systems, 4(4), 159–​168. Han, J., & Kim, D. (2009). r-​Learning services for elementary school students with a teaching assistant robot. Paper presented at the Proceedings of the Fourth ACM/​IEEE International Conference on Human-​Robot Interaction. Hashimoto, T., Kobayashi, H., Polishuk, A., & Verner, I. (2013). Elementary science lesson delivered by robot. Paper presented at the Eighth ACM/​IEEE International Conference on Human-​Robot Interaction, Tokyo, Japan. Jipson, J. L., & Gelman, S. A. (2007). Robots and rodents: Children’s inferences about living and nonliving kinds. Child Development, 78(6), 1675–​1688. doi:  http://​dx.doi.org/​ 10.1111/​j.1467-​8624.2007.01095.x. Johnson, C. N., & Wellman, H. M. (1982). Children’s developing conceptions of the mind and brain. Child Development, 53, 222–​234. doi: http://​dx.doi.org/​10.2307/​1129656.

164  Kimberly A. Brink and Henry M. Wellman Johnson, S. C. (2000). The recognition of mentalistic agents in infancy. Trends in Cognitive Sciences, 4(1),  22–​28. Kahn, P. H., Kanda, T., Ishiguro, H., Freier, N. G., Severson, R. L., Gill, B. T., . . . Shen, S. (2012). “Robovie, you’ll have to go into the closet now”: Children’s social and moral relationships with a humanoid robot. Developmental Psychology, 48(2), 303–​314. doi: http://​dx.doi.org/​10.1037/​a0027033 Kanda, T., Hirano, T., Eaton, D., & Ishiguro, H. (2004). Interactive robots as social partners and peer tutors for children: A field trial. Human-​Computer Interaction, 19(1),  61–​84. Koenig, M. A., & Echols, C. H. (2003). Infants’ understanding of false labeling events: The referential roles of words and the speakers who use them. Cognition, 87(3), 179–​208. Koenig, M. A., & Harris, P. L. (2007). The basis of epistemic trust: Reliable testimony or reliable sources? Episteme, 4(03), 264–​284. Lee, S., Noh, H., Lee, J., Lee, K., Lee, G. G., Sagong, S., & Kim, M. (2011). On the effectiveness of robot-​assisted language learning. ReCALL, 23(1),  25–​58. Low, A. (2016). Asus’ adorable Zenbo home robot will be released in 9 to 12 months. Retrieved from http://​www.cnet.com/​news/​asus-​zenbo-​will-​only-​be-​released-​in-​9-​ 12-​months/​. MacDorman, K. F. (2006). Subjective ratings of robot video clips for human likeness, familiarity, and eeriness: An exploration of the uncanny valley. Paper presented at the ICCS/​CogSci-​2006 Long Symposium: Toward Social Mechanisms of Android Science, Vancouver, Canada. MacDorman, K. F., Green, R. D., Ho, C.-​C., & Koch, C. T. (2009). Too real for comfort? Uncanny responses to computer generated faces. Computers in Human Behavior, 25(3), 695–​710. doi: http://​dx.doi.org/​10.1016/​j.chb.2008.12.026 MacDorman, K. F., & Ishiguro, H. (2006). The uncanny advantage of using androids in cognitive and social science research. Interaction Studies, 7(3), 297–​337. Mara, M., & Appel, M. (2015). Effects of lateral head tilt on user perceptions of humanoid and android robots. Computers in Human Behavior, 44, 326–​334. Mejías, C. S., Echevarría, C., Nuñez, P., Manso, L., Bustos, P., Leal, S., & Parra, C. (2013). Ursus: A robotic assistant for training of children with motor impairments. In J. L. Pons, D. Torricelli, & M. Pajaro (Eds.), Converging Clinical and Engineering Research on Neurorehabilitation (pp. 249–​253). Berlin, Heidelberg: Springer. Melson, G. F., Kahn, P. H., Beck, A., Friedman, B., Roberts, T., Garrett, E., & Gill, B. T. (2009). Children’s behavior toward and understanding of robotic and living dogs. Journal of Applied Developmental Psychology, 30(2), 92–​102. Mikropoulos, T., Misailidi, P., & Bonoti, F. (2003). Attributing human properties to computer artifacts:  Developmental changes in children’s understanding of the animate-​ inanimate distinction. Psychology:  The Journal of the Hellenic Psychological Society, 10,  53–​64. Mori, M., MacDorman, K. F., & Kageki, N. (2012). The uncanny valley [from the field]. IEEE Robotics and Automation Magazine, 19(2), 98–​ 100. doi:http://​dx.doi.org/​ 10.1109/​MRA.2012.2192811 Movellan, J., Eckhardt, M., Virnes, M., & Rodriguez, A. (2009). Sociable robot improves toddler vocabulary skills. Paper presented at the Fourth ACM/​IEEE International Conference on Human-​Robot Interaction. “National Robotics Initiative 2.0:  Ubiquitous Collaborative Robots (NRI-​2.0).” (2017). Retrieved from https://​www.nsf.gov/​pubs/​2017/​nsf17518/​nsf17518.htm.

Technology as Teacher  165 Nomura, T., Kanda, T., Kidokoro, H., Suehiro, Y., & Yamada, S. (2017). Why do children abuse robots? Interaction Studies, 17(3), 347–​369. Okita, S. Y., Ng-​Thow-​Hing, V., & Sarvadevabhatla, R. (2009). Learning together: ASIMO developing an interactive learning partnership with children. Paper presented at the 18th IEEE International Symposium on Robot and Human Interactive Communication, 2009, RO-​MAN 2009. . Okita, S. Y., Ng-​Thow-​Hing, V., & Sarvadevabhatla, R. K. (2011). Multimodal approach to affective human-​robot interaction design with children. ACM Transactions on Interactive Intelligent Systems (TiiS), 1(1), 5. Powell, M. (2014). Robot teachers in the classroom. Retrieved from http://​iq.intel.com/​ robot-​teachers-​in-​the-​classroom/​. Richert, R. A., & Harris, P. L. (2006). The ghost in my body: Children’s developing concept of the soul. Journal of Cognition and Culture, 6(3-​4), 409–​427. doi: http://​dx.doi.org/​ 10.1163/​156853706778554913. Ricks, D. J., & Colton, M. B. (2010). Trends and considerations in robot-​assisted autism therapy. Paper presented at the 2010 IEEE International Conference on Robotics and Automation (ICRA), Anchorage, AK. Robins, B., Dautenhahn, K., & Dubowski, J. (2006). Does appearance matter in the interaction of children with autism with a humanoid robot? Interaction Studies, 7(3), 509–​542. Saerbeck, M., Schut, T., Bartneck, C., & Janse, M. D. (2010). Expressive robots in education:  Varying the degree of social supportive behavior of a robotic tutor. Paper presented at the SIGCHI Conference on Human Factors in Computing Systems. Sawchuk, S. (2014). Standards pose teacher-​prep challenge. Education Week, 33(29), S14–​16. Turkle, S. (2011). Alone Together: Why We Expect More from Technology and Less from Each Other. New York: Basic Books. Victor, D. (2015). Hitchhiking robot, safe in several countries, meets its end in Philadelphia. New York Times. Wei, C.-​W., Hung, I.-​C., Lee, L., & Chen, N.-​S. (2011). A joyful classroom learning system with robot learning companion for children to learn mathematics multiplication. TOJET: The Turkish Online Journal of Educational Technology, 10(2),  11–​23. Weiss, A., Wurhofer, D., & Tscheligi, M. (2009). “I love this dog”—​children’s emotional attachment to the robotic dog AIBO. International Journal of Social Robotics, 1(3), 243–​248. Wellman, H. M. (2014). Making Minds: How Theory of Mind Develops. Oxford: Oxford University Press. Wong, J. C. (2016). “This is awful”: Robot can keep children occupied for hours without supervision. The Guardian. Retrieved from https://​www.theguardian.com/​technology/​ 2016/​sep/​29/​ipal-​robot-​childcare-​robobusiness-​san-​jose.

9 Understanding Others to Learn and Help Others Learn Inferences, Evaluation, and Communication in Early Childhood Hyowon Gweon

Think about all the things you know: the names of things and places, the planets in the solar system, your favorite pasta recipe, or the meaning of “democracy.” Now, think about how you’ve come to know all this. You might soon realize that much of your knowledge about the world comes from what you’ve learned from other people. Humans are powerful learners. Within just a few years of life, young human learners construct a coherent system of intuitive theories about how the world works (Carey, 2011; Gopnik & Wellman, 2012; Schulz, 2012). What is even more striking is that this remarkable feat seems to happen almost effortlessly; children learn as they naturally play and interact with their environment. One important thing to note, however, is that humans, especially young children, rarely experience the world in isolation. Even though we learn a lot from our firsthand experiences with the things around us, these experiences are often heavily intertwined with the people around us. We observe other people as they navigate our surroundings, and as they show, tell, or teach us about faraway places, planets in the solar system, and the concept of democracy. Through these interactions, we learn things that go far beyond what we can directly experience. A fundamental challenge for theories of human learning is to provide a unified account for how we combine our own experience with the world and information provided by others in order to learn so much, so quickly, and so accurately, in a complex, noisy environment. One distinctive feature of human social learning is that it involves more than simply adopting others’ behaviors through observation and imitation; it

168  Hyowon Gweon also involves adopting abstract knowledge that resides in others’ minds, such as their causal and ontological beliefs about how the world works. This process is much more complex than observing others’ actions and reproducing them. Because one’s knowledge cannot be directly copied from one mind to another, a “teacher” needs to generate a set of observable behaviors for a learner, and the “learner” needs to interpret what those behaviors mean to recover the teacher’s knowledge. Even though the learners can only see the concrete actions of the teacher, what is actually shared (by the teacher) and acquired (by the learner) is the message encoded in these actions. Critically, in order for such encoding and decoding to occur successfully, both the teacher and the learner must reason and act with the other person in mind. Thus, the basic human capacity to understand other minds—​in particular, the ability to interpret others’ actions in terms of their underlying mental states—​is a key part of cognitive foundations of distinctively human social learning. Humans naturally understand others’ behaviors in terms of their underlying mental states. When someone is running up the stairs in a train station, we readily reason about the actor’s goals (e.g., catching the train), beliefs (e.g., he thinks the train has arrived), and even his expected utilities of successfully achieving his goal (e.g., the benefit of catching the train probably outweighs the cost of running). Even infants can represent and reason about these mental states (Repacholi & Gopnik, 1997; Woodward, 1998; Wellman, Cross, & Watson, 2001) and expect others to act in ways that maximize their utilities (Gergely & Csibra, 2003; Jara-​Ettinger et al., 2016; Liu et al., 2017). Indeed, these social-​cognitive capacities are critical for many different kinds of social interactions; the goal of this chapter is to focus on a particular class of these interactions—​pedagogical interactions between a learner and a teacher—​as a way to highlight how these capacities can make human social learning so powerful, flexible, and effective. Yetsuccessful social learning involves more than the ability to represent others and their internal states. These representations must be incorporated in inferences that also involve representations of the physical world, such as objects, number, or space. Although the importance of domain-​general inferential abilities in early learning has been widely recognized in cognitive development literature (Gopnik, 2012; Schulz, 2012; Xu & Kushnir, 2013), this literature has remained relatively agnostic about the social context in which early learning most frequently occurs, or how social contexts can modulate the inferences learners draw from data. Instead, theories of

Understanding Others to Learn and Help Others Learn  169 social learning have proposed learning mechanisms that operate specifically in social contexts (e.g., imitation, natural pedagogy; Meltzoff, 2007; Csibra & Gergely, 2009; Rendell et al., 2010; Dean et al., 2012; Legare & Nielsen, 2015). These domain-​specific learning mechanisms are important, but insufficient to fully capture the active nature of early social learning; rather than passively accepting information provided by others, young children readily seek out and solicit information from others, combine it with their own prior knowledge, and even provide useful information for others to share what they know. The studies reviewed here are at the intersection of these two perspectives, examining how children draw inferences from available data in their social environment to support their decisions as learners and as teachers. While some focus on children as learners and others look at children as teachers, they are aimed at a larger common goal: To better understand how humans flexibly navigate these roles—​recipients and providers of information—​ depending on the social context. To this end, they take a similar approach to study these interactions: Examining children’s inferences and communicative behaviors in experimental settings, specifically in contexts in which one agent provides information, and the other learns from that information. These contexts range from casual interactions between a child and an adult to more explicitly pedagogical ones between a “teacher” and a “learner.” In what follows, I first explain why pedagogical interactions can provide a particularly useful window into the cognitive mechanisms that underlie human social learning. Then, across three sections, I review studies that show how young children (1) draw rational inferences from information provided by others, (2) use such information to evaluate others’ quality as teachers, and (3) provide information for others as teachers themselves. These studies together show how humans, starting early in life, actively capitalize on their understanding of others in order to learn and to help others learn.

1.  Why Study Social Learning in Communicative, Pedagogical Contexts? Social learning occurs in many forms. Sometimes we learn by simply observing others’ actions even when they have no intent to communicate with us. However, social learning can be more effective when someone has a pedagogical goal to inform or teach us. A large body of work has shown

170  Hyowon Gweon that humans are responsive to pedagogy from early in life; for instance, when information about a particular object is provided along with ostensive, communicative cues (e.g., pointing, eye-​gaze), even infants show a tendency to interpret that information as generalizable knowledge that applies not only to that particular object but also to a broader set of objects in the same category (Csibra & Gergely, 2009; 2011; Csibra & Shamsudheen, 2015). While such sensitivity might play an especially important role during the earliest months of learning, the studies reviewed here mostly focus on late infancy to preschool years. In these studies, the presence of ostensive cues is often held constant, rather than manipulated, because the goal is not to examine the influence of being in a pedagogical context where these cues are present; instead, the goal is to show how children’s understanding of other minds can modulate their inferences, evaluations, and communicative decisions in pedagogical contexts where such communicative cues are amply present.1 Studying children’s behaviors as learners and as teachers in these contexts can be particularly interesting and informative for a few reasons. First, teacher-​learner interactions are ubiquitous, especially in young children’s lives. The fact that much of early learning unfolds in pedagogical contexts makes these interactions worth studying in their own right. Granted, there is some cultural variation; there are small-​scale societies where formal schooling does not exist, and explicit, direct instruction from adults to children is relatively infrequent (Rogoff, 2003; MacDonald, 2007). Nevertheless, the presence of informal social learning opportunities that involve intentional communication of information (e.g., a parent pointing to an object and labeling it, or an older sibling demonstrating how to play a game) may be a universal feature of young children’s social environment (Hewlett et al., 2011). As we will see later in this chapter, the ability to interpret information in pedagogical contexts may arise from basic cognitive capacities that are assumed to be universal, rather than from frequent exposure to the cultural practice of teaching. Second, both learning and teaching require an integration of a host of cognitive faculties, which may all have different origins and developmental 1 Communicative contexts can vary with respect to the strength of cooperative expectations that interlocutors hold about each other. In this chapter, I consider pedagogical interactions as special cases of communication where such expectations are very strong: the learner expects the teacher to communicate cooperatively, and the teacher expects the learner to interpret the evidence with that expectation (Grice, 1975; see also Goodman & Frank, 2016). Naturally, these interactions often involve the communicator’s use of ostensive cues such as eye-​gaze or pointing (Csibra & Gergely, 2009; 2011), but the presence of ostensive cues per se does not determine whether an interaction is considered pedagogical.

Understanding Others to Learn and Help Others Learn  171 trajectories. Thus, studying how these complex behaviors emerge early in life can also inform theories of human cognition more generally. In particular, recent computational work on pedagogical reasoning and communication offers formal descriptions of the representations and inferences involved in these interactions (Shafto, Goodman, & Frank, 2012; Shafto, Goodman, & Griffiths, 2014; Goodman & Frank, 2016). This computational framework provides firm grounds for developing theory-​driven hypotheses about developmental change in children’s inferential and communicative abilities in social contexts (see Figure 9.1). Thus, researchers can go beyond testing qualitative predictions about children’s successes and failures; they can formulate hypotheses about how children succeed and why they fail in precise, quantitative terms, and generate graded predictions about children’s behaviors across tasks and across age groups. Designing model-​based experiments also forces researchers to be more explicit about the mapping between experimental designs and the computational models that predict the results. Third, pedagogical contexts can be used as an effective methodological tool for discovering early competences for social learning and communication. In everyday social interactions, people’s goals and epistemic states are often ambiguous or left unspecified. However, when two agents engage in a pedagogical interaction, their goals, motivations, and epistemic states are naturally made clear in the context. When children participate in pedagogical interactions as learners, they expect to learn from a knowledgeable person who intends to teach them something new; when children participate as

What should I learn? What is the hypothesis that best explains the data?

Learner

Information is interpreted based on learner’s prior beliefs and how the data are selected

PL(h|d) ∝PT(d|h)PL(h)

What does the learner need to know? What data would help the learner to infer the correct hypothesis?

Teacher Information is selected to maximize the probability that the learner infers the correct, useful hypothesis

PT(d|h) ∝PL(h|d)

Figure 9.1  Learning in the social context. Young children learn from their own interactions with the world. However, learning can be more effective when another agent (“teacher”) selects and provides data for the learner. Successful learning requires not only the teacher’s ability to select the right kind of data, but also the learner’s ability to use this sampling process (PT(d|h)) to recover the knowledge that the teacher intends to convey.

172  Hyowon Gweon teachers, they expect that learners need their help to order to learn something new, and they are motivated to offer help. Even though these expectations are certainly not unique to pedagogical interactions, they are strengthened and amplified in these contexts, providing a useful test-​bed for identifying early inferential and communicative competences that may fail to manifest in other kinds of social interactions where such expectations are weaker. In sum, young children’s inferences, evaluation, and communication in pedagogical contexts is both a theoretically interesting area of inquiry and a useful methodological approach that allows us to study how humans use their understanding of other minds to learn from others and help others learn.

2.  Rational Inference in the Social Context 2.1.  Young Children as Active Interpreters of Information Provided by Others Observing and imitating others’ behaviors is a simple yet useful way to learn from others (Dean et al., 2012; Legare & Nielsen, 2015; Meltzoff, 2007). For instance, if you see someone press a button to activate a toy, you would likely imitate that action and press the button when given a chance to play with the toy. However, if you fail to activate the toy, what would you do next? One study shows that early-​developing inferential abilities can support learning even when imitation goes awry; even 16-​month-​old infants rationally decide what to do next in the face of a failed action, based on the pattern of others’ actions and their outcomes (Gweon & Schulz, 2011). In this study, infants watched two adult experimenters taking turns to press a button to activate a toy. The critical manipulation was the statistical dependency between the experimenters’ actions and their outcomes during the demonstration. Some infants saw one experimenter succeed twice and the other experimenter fail twice (i.e., the outcomes of actions covaried with the agent), while others saw each experimenter succeed once and fail once (i.e., outcomes varied independent of the agent). Infants were then given a chance to try the toy, only to fail to activate the toy, and they could either ask their parent for help, or try a different toy that was similar in appearance. Note that all infants saw the same two agents and equal numbers of successful and unsuccessful actions, and readily imitated their actions to press the button on the toy. However, their subsequent actions reflected different inferences about

Understanding Others to Learn and Help Others Learn  173 the underlying cause of their failures. When outcomes covary with the agent (i.e., one agent fails twice and the other succeeds twice), the evidence is consistent with the possibility that the failures are due to an incompetent agent rather than a faulty toy. Faced with their own failure, infants in this condition approached their parent for help, suggesting that they attributed the cause to the self (i.e., “I probably can’t do it”). However, when outcomes vary independently of the agents (i.e., both agents fail once and succeed once), this pattern of evidence suggests that their failures are due to a faulty toy rather than something about the agents. Consistent with this, infants in this condition reached out for a different toy, suggesting that they attributed the cause of their own failure to the object (“this toy is faulty”). Beyond inferring what to do now (i.e., how to activate the toy), infants in this study readily interpreted the statistical patterns of others’ actions and outcomes to figure out what went wrong, and used this causal attribution to decide what to do next. When agents act on the physical world, their actions and their outcomes generate a pattern of data. In Gweon & Schulz (2011), these outcomes were clear successes and failures; however, the data may come in many other forms, such as objects that have been drawn from a box full of toys. There is now a body of work on preverbal infants’ ability to draw probabilistic inferences based on statistical information (e.g., Aslin, Saffran, & Newport, 1998; Téglás et al., 2011; Xu & Garcia, 2008), and in particular, agents’ goal-​ directed sampling of objects (e.g., Xu & Tenenbaum, 2007; Kushnir, Xu, & Wellman, 2010; Ma & Xu, 2013). For instance, when an agent selects a clearly non-​representative sample from a population (e.g., a few blue toys from a box that contains mostly yellow toys and just a few blue toys), infants understand that the agent must have selected them for a reason (i.e., she must like the blue toys). Importantly, when these actions are directed toward someone with a communicative purpose, they can give rise to a rich understanding of what the agent intends to communicate, given how the agent selected the information (see Figure 9.2, left). Imagine someone pulls out such a non-​representative sample (a few blue toys from a mostly yellow box) and squeezes each blue toy to show that they all squeak. This sample would be highly improbable under random sampling, and the data strongly suggest that she deliberately selected the blue toys. Why would she bother selecting just the blue toys to demonstrate this squeaking property? Presumably because the yellow toys do not squeak! This intuitive conclusion is, in fact, the result of a sophisticated set of inferences about both the sample and the sampling process by which the

174  Hyowon Gweon

Yellow toys do not squeak!

(only) blue toys squeak!

This toy squeaks, and it has no other functions!

This is how my toy works!

Figure 9.2  Inferences in communicative contexts. Left: When an adult demonstrates the hidden property of a few blue toys (i.e., pressing them generates squeaky noise), infants use the statistical dependence between the sample and the population to infer the property of undemonstrated objects (yellow toys). When the toys are drawn from a box that contains mostly blue toys (sampling process is ambiguous), infants readily generalize the property to the yellow toys; if the sample is drawn from a box of mostly yellow toys (suggests deliberate sampling of blue toys), infants do not expect the yellow toys to have the same squeaking property (Gweon, Tenenbaum, & Schulz, 2010). Right: These inferences are even stronger in pedagogical contexts where a knowledgeable teacher is expected to help the learner acquire useful knowledge. If the teacher demonstrates just one function of a toy, the learner not only learns about that function but also further infers that the toy has no other functions (if it did, the teacher would have shown them, too).

sample was generated. Even preverbal infants can engage in such reasoning, using the statistical dependence between a population and a sample to guide their inferences about object properties (Gweon, Tenenbaum, & Schulz, 2010). In this study, infants first watched an adult draw three blue toys just like in the preceding scenario. Given a chance to explore the yellow toy (which never squeaked), infants squeezed the yellow toy far less often when the sample was likely to have been selectively drawn (i.e., the box contained mostly yellow toys) than randomly drawn (i.e., the box contained mostly blue toys). Remarkably, the strength of these inferences reflected the strength of the data infants observed; the less likely the samples, the less frequent were their squeezing behaviors. Another notable aspect of these findings is that infants’ inferences did not require a lot of observations; they were based on minimal statistical data. Using the relative proportion of blue and yellow toys and just a few exemplars from the population, infants inferred properties of objects that were never sampled or demonstrated. Such sensitivity to sampling processes may emerge from a basic understanding of

Understanding Others to Learn and Help Others Learn  175 agents’ goal-​directed actions and their costs. When agents’ behaviors violate the assumption that they should act in ways that minimize the costs (Gergely & Csibra, 2003; Jara-​Ettinger et al., 2016), even infants seek rational explanations for such behaviors (i.e., “why did she go out of her way to deliberately select these objects?”), and use them to learn about the world. Pedagogical contexts are an especially interesting and important domain in which deliberate, selective sampling processes affect learning. When a teacher—​someone who is assumed to be knowledgeable—​communicates information for a learner, it is implied that the teacher is deliberately selecting information that would be most helpful for the learner. Computational models of pedagogical reasoning (Shafto et al., 2012; 2014) have formalized these intuitions, characterizing “what the teacher communicates” and “what the learner learns” as a set of mutually dependent inferences. The teacher’s selection of data depends on what the learner would infer given the data (i.e., pedagogical sampling), and the learner’s inferences from the data depend on how the learner thinks the data were selected by the teacher. These two mutually constraining inferences help the learner draw powerful inferences that go beyond the face value of the evidence (see Figure 9.1). A series of studies have used demonstrations with causal toys to show that even children as young as two years old expect teachers to engage in pedagogical sampling and interpret the evidence accordingly (Bonawitz et al., 2011; Shneidman et al., 2016; see also Butler & Markman, 2012). In these studies, an adult experimenter claims to know everything about a complex-​looking novel toy, and says, “Let me show you how it works!” She then demonstrates that pressing a lever on the toy makes an interesting sound. In this context, the observer (i.e., the learner) expects the demonstrator (i.e., the teacher) to be knowledgeable about the toy and to provide the best set of evidence about the toy. Therefore, the learner can safely assume that the teacher would provide information that is not only true of the world but also sufficient for the learner to draw the correct inference. As a result, the learner not only learns that the toy has an interesting function (i.e., pressing the lever generates a sound), but also infers that there are no other functions to learn; if there were more, the teacher would have demonstrated them, too (Figure 9.2, right). This example illustrates how a learner’s expectations about the properties of pedagogically communicated evidence can support inferences that go beyond the teacher’s demonstration. Critically, such inferences can be measured through children’s spontaneous exploration of the toy; when children are given a chance to explore the toy after observing an experimenter pedagogically demonstrate one

176  Hyowon Gweon of its functions, they show limited exploration of the toy and instead spend the majority of their time playing with the demonstrated function. However, when the teacher’s demonstration was non-​pedagogical (e.g., apparently interrupted after showing one function, suggesting that she intended to show more, or when the teacher accidentally discovered the function, suggesting that she was ignorant of the toy’s functions), children explore the toy more broadly, searching for potential additional functions. One might wonder whether these results suggest a serious problem in how we educate our young: Does teaching always hinder children’s learning by limiting spontaneous exploration and discovery? If the toy actually had multiple functions, learners who learned about one of its functions from a teacher in a pedagogical context would likely have failed to explore and discover the other functions. However, it is important to note that this study deliberately used a teacher who failed to provide all relevant information about the toy, as a way to show how such reduction in exploration can be a natural consequence of the inductive inferences that, ironically, make pedagogical learning so powerful and efficient in most typical pedagogical interactions. Indeed, if the toy actually had only one function, children would have learned this function accurately and efficiently while avoiding the costs of fruitless exploration. If the toy actually had other functions (as was the case in this study), a good teacher in a real pedagogical context would indeed have shown these functions or even have encouraged the learner to explore and learn about them. What this means is that pedagogical learning is beneficial insofar as the teacher is knowledgeable and helpful. One important challenge for researchers is to figure out how to apply these findings to real-​world contexts to help “teachers” (e.g., parents, caregivers, educators) understand how pedagogical contexts influence learners’ inferences, when its power can backfire, and how pedagogical instructions can be used in ways that maximize their benefits while minimizing their unintended side effects. These studies also raise questions about whether such inferences depend on formal schooling and repeated exposure to pedagogical contexts. Recent work suggests that these inferences are observed even in cultures where children rarely receive direct pedagogical instruction from adults (Shneidman et al., 2016). Toddlers in Yucatec Mayan communities show patterns of exploration in pedagogical versus non-​pedagogical contexts similar to those of their same-​aged peers in the United States, who presumably receive more frequent instruction from their caregivers. These results provide important empirical support for the idea that children’s inferential abilities emerge from

Understanding Others to Learn and Help Others Learn  177 a basic understanding of others’ knowledge, goals, and their actions, rather than from culturally specific teaching practices. Together, these studies suggest that young human learners are not just passively absorbing information from others; they actively interpret others’ actions and their outcomes based on a rich understanding of other minds. In particular, when someone deliberately selects evidence with a clear communicative goal (e.g., to teach), children draw powerful inferences that go beyond the evidence. Even though children’s observations of others’ behaviors are often noisy and sparse, these inferences support rapid, robust, yet accurate social learning even from small amounts of data.

2.2.  Young Children as Active Evaluators of Others’ Informativeness In pedagogical contexts, strong inferences are licensed because the learners expect their source of information to be someone who is knowledgeable and helpful. However, the power of pedagogical learning can turn into a hazard when such expectations are violated. After all, some people may be less knowledgeable or less helpful than others, and some may even intend to deliberately mislead the learner. Therefore, in order to learn from others, it is critical to learn about others’ informativeness to avoid learning from untrustworthy sources. Let’s go back to the example of a teacher who demonstrates one function of a toy. While this would be considered reasonable and fully informative if the toy had just one function, the same demonstration would be considered underinformative if the toy had three other functions that were left undemonstrated. In fact, by leaving out relevant information, this teacher would be committing a “sin of omission,” misguiding the learner to falsely believe that the toy has just one function when it in fact has more. Compared to telling a lie or providing false information (Birch, Vauthier, & Bloom, 2008; Koenig & Harris, 2005; Sabbagh & Baldwin, 2001), this is a rather subtle form of misinformation. There isn’t anything inaccurate or wrong about the teacher’s demonstration, yet it would still lead to inaccurate learning about the toy! Critically, recognizing and evaluating sins of omission requires both prior knowledge about the world and an understanding of pedagogical sampling. Without prior knowledge about the toy’s actual functions, one might not even recognize that the teacher omitted some functions of the toy. Without

178  Hyowon Gweon an understanding of pedagogical sampling, an omission might be detected but not necessarily evaluate it as a “sin”—​such evaluation requires the understanding that the teacher should have selected a better set of evidence for the learner (i.e., showing all of its functions) given her knowledge about the toy. Building on prior work on young children’s understanding of pedagogical sampling, a series of recent studies tested early school-​aged children’s evaluations of sins of omission by comparing children’s ratings of a fully informative versus an underinformative teacher (e.g., Gweon et al., 2014, see Figure 9.3). One group of six-​to seven-​year-​olds were shown a toy with one function, while another group of children were shown a nearly identical toy with four functions (including the same function as the one-​function toy). All children then rated a teacher who taught a naive learner how the toy works. Critically, the teacher always taught the function that was common to both toys. Even though the teacher’s demonstration was true, useful, and identical across conditions, children rated the teacher as less helpful when the toy had other undemonstrated functions (suggesting that the teacher was under-​informative, showing only some of its functions) than when the toy had only one function (suggesting the teacher was fully informative). Thus children are not just sensitive to the accuracy of information in pedagogical contexts; their evalutions are based on a more abstract notion of informativeness, which includes whether the information was sufficient for learning. Importantly, such evaluations are not just fleeting impressions; a follow-​ up study showed that six-​year-​old children remember others’ past informativeness and modulate their subsequent learning (Gweon et al., 2014; see Figure 9.3, right). As in the first study, children saw a teacher who provided underinformative or fully informative demonstration about a toy. However, instead of rating the teacher, children then saw the teacher demonstrate a second toy that they had never seen before. Given a chance to play with the toy, children explored the toy more broadly (in search of potential additional functions) if the teacher had been underinformative in the past than when he had been fully informative. In other words, children who saw the underinformative teacher continued to question his credibility, and engaged in compensatory exploration to see if the toy had additional functions the teacher did not demonstrate. These results suggest that children can interpret the same pedagogical demonstration differently depending on the teacher’s past informativeness, and modulate their subsequent inferences and exploration accordingly. Pedagogical sampling assumptions may be a particularly strong instantiation of Grice’s cooperative principles (Grice, 1975):  Speakers should

Understanding Others to Learn and Help Others Learn  179 Teach 1/1 (Informative)

20

Teacher Helpfulness Rating **

***

***

15

100 % 80

% Time spent exploring new toy **

*

60 10

40

5 0 Teach 1/4 (Underinformative)

20 Age 6 & 7

Age 6 & 7

Age 4 & 5

Gweon et al. Gweon & Asaba (2017) (2014) *, **, *** : p< 0.05, 0.005, 0.0001

0

Teach 1/1 Teach 1/4 Teach 4/4 This is how my toy works!

Figure 9.3  Evaluation of under-​informative teachers. Left: Children rated a teacher who taught the only function of a novel toy (Teach 1/​1) higher than a teacher who taught the same function while omitting 3 others (Teach 1/​4). Four-​and five-​year-​olds show this distinction only when they’ve seen a clear example of an informative teacher (Gweon & Asaba, 2017). Right: Children’s exploration of a new toy is modulated by the teacher’s prior informativeness (Gweon et al., 2014).

be truthful (Maxim of Quality) and as informative as required (Maxim of Quantity). In linguistic communication, these cooperative expectations about the interlocutor can support various pragmatic inferences, allowing listeners to go beyond the literal meaning of speakers’ utterances (Grice, 1975; Clark, 1996; Goodman & Frank, 2016). For instance, saying that “the boy ate some of the cookies” is underinformative when the boy, in fact, ate all of the cookies; although it is logically consistent with either possibility, it pragmatically misleads the listener to infer that the boy presumably ate some (and not all). A series of recent studies on pragmatic implicature have shown that even though preschool-​aged children typically fail to evaluate such underinformative speakers, they do succeed given enough contextual support (Barner, Brooks, & Bale, 2011; Skordos & Papafragou, 2016). More specifically, children succeed when they understand that the speaker could have used the word “all” instead of “some.” Children’s surprising yet limited competence in pragmatic implicatures supports an interesting hypothesis:  If children’s evaluations of underinformative teachers are also based on a similar reasoning about what a teacher could have shown about a toy, we might see parallel limitations in younger children’s evaluations of teachers who communicate via goal-​directed actions (Baldwin, Loucks, & Sabbagh, 2008).

180  Hyowon Gweon A recent study provides empirical support for this hypothesis (Gweon & Asaba, 2017). Using a similar setup as in Gweon et al. (2014), children were shown two teachers who both demonstrated one interesting function of a toy; one was fully informative (because the toy had a single function), and the other was underinformative (because the toy had three additional functions). The critical manipulation was the order in which the teachers were evaluated: Some children saw the fully informative teacher first, while others saw the underinformative teacher first. If seeing and evaluating the fully informative teacher provides a useful reference point for evaluating the underinformative teacher, children younger than six years of age might successfully penalize the underinformative teacher, but only when the fully informative teacher is presented first. The results were consistent with these predictions. While six-​year-​olds penalized the underinformative teacher regardless of the order in which the teachers were presented (replicating Gweon et al., 2014), preschoolers (four-​and five-​year-​olds) did so only when they rated the fully informative teacher first. Importantly, seeing a teacher who was highly rated for a different reason (e.g., accurately naming an object) did not help; the effect was specific to seeing a fully informative teacher (e.g., providing all relevant functions of a toy). Furthermore, after seeing both teachers, even four-​year-​olds appropriately preferred the fully informative teacher, providing the earliest evidence for children’s ability to detect “sins of omission.” Even though these studies do not include computational models that generate specific predictions for children’s ratings, exploratory play, or teacher choice, the results are generally consistent with the computational account that formalizes pedagogical reasoning as a set of recursive mental-​state inferences between a teacher and a learner. The failure to provide all relevant information for a learner violates the assumptions of pedagogical sampling (i.e., a knowledgeable teacher selects data to increase the learner’s belief in the correct hypothesis; see Shafto et al., 2012; 2014), and supports the inference that the teacher is either ignorant or unhelpful (or both). Note, however, that these studies do not distinguish between these more nuanced attributions; they only show that children readily detect and evaluate sins of omission in pedagogical contexts. As adults, however, we do distinguish ignorant teachers from deliberately unhelpful teachers, and understand that not all omissions are equally blameworthy. For instance, failing to show one of four functions is less egregious than failing to show three of four, and omitting a useless function is better

Understanding Others to Learn and Help Others Learn  181 than omitting an important one. If the teacher only shows one function because he did not know about the other three, one might be more inclined to pardon this ignorant teacher than a knowledgeable teacher who knowingly omitted these functions. Furthermore, if a teacher shows only one of four functions because the learner already knows about the other three, omission could actually be deemed desirable for its efficiency. Thus, evaluation of omission depends on what and how much information was omitted, as well as the knowledge state of the learner and the teacher. Recent computational work has extended prior models of pedagogical reasoning to explain how people might make such nuanced evaluations of teachers, and showed that adults’ evaluations are consistent with the model predictions (Bass et al., 2015; see also Shafto et al., 2012). These models incorporate knowledge states of both the learner and the teacher, and predict graded judgments of teacher quality based on the amount, value, and cost of information taught by the teacher. Consistent with this work, an ongoing series of studies suggest that young children’s evaluations of teachers also reflect a rich, sophisticated understanding of what it means to be helpful and informative. For instance, children exonerate underinformative teaching if the teacher couldn’t be more informative due to her ignorance (Bass, Bonawitz, & Gweon, 2017), and understand that an exhaustive demonstration (e.g., showing every button on a multi-​button toy even though only a few generate an effect) can be necessary or overinformative depending on the learner’s prior knowledge (Gweon, Shafto, & Schulz, 2018). The fact that children penalize redundancy or overinformation (e.g., demonstrating functions that a learner already knows) reflects their understanding of the trade-​off between informativeness and efficiency. That is, a teacher who provides redundant information has failed to achieve efficiency at the expense of being maximally informative. These results suggest that the ability to reason about the costs and benefits of others’ actions (Jara-​Ettinger et  al., 2016)  may underlie these nuanced evaluations of teacher informativeness. In sum, children don’t simply avoid people who are wrong, nor do they indiscriminately prefer whoever provides more information. Instead, children reason about what the teacher knows and what the learner needs, and they evaluate whether the teacher selected information in ways that are beneficial for the learner. Collectively, this line of work shows how young children can protect themselves from the potential risks of social learning by learning about others; beyond detecting and recognizing misleading informants,

182  Hyowon Gweon they actively infer others’ qualities as teachers based on how the evidence is selected, and appropriately modulate their future learning and exploration.

2.3.  Young Children as Active Communicators of Information Children’s ability to evaluate others’ informativeness raises an interesting possibility:  The same cognitive mechanisms that underlie their ability as learners might underlie their abilities as informative teachers. Just as children’s understanding of pedagogical sampling supports their evaluations of other people’s teaching, it may also allow children to conform to these expectations as teachers themselves. Some recent and ongoing work begins to test this hypothesis by placing children in the role of a teacher and studying their ability to communicate information to learners. Recent work shows that around ages five and six, children readily infer the right amount of information to teach based on the learners’ knowl­ edge, goals, and competence, and tailor their demonstrations accordingly (Gweon, Shafto, and Schulz, 2018; Gweon & Schulz, 2019). In one study, given a chance to demonstrate a toy that had 20 identical buttons (only three of which played music, and the rest inert), children demonstrated different numbers of buttons depending on the prior knowledge of the learner (Gweon, Shafto, and Schulz, 2018). When the learner had already seen similar toys, and thus expected the toy to have just a few working buttons, children provided efficient demonstrations by showing just the three working buttons; however, when the learner was naive (thus expecting all buttons to work in the same way), children provided exhaustive demonstrations by pressing all 20 buttons which can help the learner acquire an accurate belief about how the toy works (i.e., only three buttons work). Another study suggests that children selectively provide costly causal demonstrations for a learner only when they are necessary given the learner’s goals (e.g., whether the learner wants to learn about the toy vs. observe the toy’s effect) or competence (e.g., whether the learner was introduced as an ordinary, naive learner vs. an exceptionally smart learner; Gweon & Schulz, 2019). Along with other recent studies on children’s ability to teach (e.g., Ronfard & Corriveau, 2016; Clegg & Legare, 2016; Ronfard, Was, & Harris, 2016; see also Strauss, Calero, & Sigman, 2012 for a review), these results provide compelling evidence that children’s teaching has the key properties of “good pedagogy”: informative,

Understanding Others to Learn and Help Others Learn  183 effective, and efficient. Children try to provide as much information as required given the learner’s mental states (e.g., goals, knowledge, competence) and avoid providing more than what is necessary. Children’s understanding of the trade-​off between informativeness and efficiency manifests not only in their inferences and social evaluations as learners but also in their decisions and behaviors as teachers, suggesting the importance of utility-​based social reasoning in all of these contexts (Jara-​ Ettinger et al., 2016). However, the studies reviewed previously do not distinguish whether children taught efficiently simply to reduce their own costs of teaching or whether they had a genuine concern for reducing others’ costs. One study shows that children can indeed represent others’ cost of learning and use it to decide what to teach, and what to let learners discover on their own (Bridgers, Jara-​Ettinger, & Gweon, 2016, in press). Although direct instruction allows learners to acquire useful knowledge without the cost of exploration, not everything can or needs to be taught. Therefore, it is important to prioritize teaching what is important yet costly to learn. This study shows that even young children can decide what to teach by reasoning about what others know or want, as well as what is easy or difficult for others to learn. Given a choice of one of two toys to teach to a naive learner (such that the learner has to learn about the other toy on her own), children choseto teach the toy that would (1) increase the learner’s expected rewards (i.e., the toy that generates a more enjoyable effect), and (2)  decrease the learner’s expected costs (i.e., the toy that is harder to figure out by oneself). To provide insights into how children make such decisions, this study compared their choices against the predictions of different computational models. Children’s responses were most consistent with the model that makes utility-​ maximizing decisions for the learner by considering the learner’s expected utilities of learning about one toy through social learning (i.e., being taught) and learning about the other through self-​guided exploration. In addition to finding early competence for making rational teaching decisions based on the principle of utility-​maximization in social reasoning, this study addresses an important question about human culture that has been sidestepped in prior literature. To explain how humans have developed cumulative cultural knowledge, it is critical to understand how humans decide what knowledge is worthy of sharing and teaching. Otherwise, our cultural knowledge would include lots of useless, trivial information as well as obsolete knowledge that is no longer useful. This study shows how basic social cognition can support selective transmission of information that prioritizes what is useful for others.

184  Hyowon Gweon Typically, children are characterized as learners who are on the receiving end of pedagogy. However, the remarkable success of human social learning depends just as much on the the inferences we make as teachers as it does on the inferences we make as learners. The studies reviewed here collectively suggest that the sophistication of adults’ intuitions as teachers may have early-​developing roots; young children not only learn from others, but also help others learn by reasoning about others’ goals, competence, and knowledge. Non-​human animals (e.g., meerkats, Thornton & McAuliffe, 2006) may show teaching-​like behaviors that have evolved to transmit particular kinds of information (see Kline, 2015, for a review). However, the ability to adjust and tailor the evidence based on an understanding of other minds is a feature that is distinctive of human social learning.

3.  Distinctively Human Social Learning: Looking Back and Looking Ahead The three lines of work reviewed in this chapter highlight an important feature of human social learning: It is a complex yet well-​coordinated process between two individuals who reason about each other’s mind. Each line of work focuses on different aspects of social learning, yet together they show that the inferences, evaluations, and pedagogical decisions humans make as learners and as teachers are rooted in early-​emerging cognitive capacities. First, children’s inferences reflect more than simple imitation; they actively use information from others to learn about the world, and they reason about what the teacher wants to communicate. Second, their evaluations of teachers go beyond mere attribution of knowledge for people who are nice, popular, or older (e.g., Lane, Wellman, & Gelman, 2013; Chudek et al., 2012; Wood, Kendal, & Flynn, 2012); children pay attention to the quality of information others provide to learn about their informativeness as teachers and modulate their future learning accordingly. Third, their communicative behaviors as teachers go beyond providing what is unknown (e.g., Liszkowski, Carpenter, & Tomasello, 2008); children actively consider learners’ mental states to teach in ways that increase the benefits of learning while reducing unnecessary costs. The richness and the power of these abilities may be grounded in two related ways of understanding others. First, they require an understanding of how observable evidence (e.g., a teacher’s behavior) influences the

Understanding Others to Learn and Help Others Learn  185 unobservable states of others’ minds (e.g., the knowledge state of the learner), a capacity has been traditionally studied in theory of mind research (Gopnik & Astington, 1988; Wellman et al., 2001). Second, they also involve an understanding of how agents’ actions are mediated by their goals (Gergely & Csibra, 2003) as well as the costs of achieving these goals (Jara-​ Ettinger et al., 2015; Liu & Spelke, 2016). The representations that support these understandings—​goals, beliefs, desires, as well as effort, competence, and costs—​are seamlessly integrated in our everyday social interactions, and manifest in ways we learn from others and help others learn. More specifically, children’s inferences, evaluations, and communicative behaviors are constrained by the expectation that other agents will act to pursue rewarding goals while minimizing the costs for achieving them. The idea that humans assume that other agents are utility-​maximizers (i.e., the Naïve Utility Calculus; see Jara-​Ettinger et al., 2016) provides a useful framework for understanding children’s learning and teaching in social contexts. Even early in life, children might engage in an intuitive cost-​benefit analysis of information transfer, and make effective decisions both as learners and as teachers. Pedagogical interactions are cooperative in nature; the teacher and the learner act together to achieve a joint goal (Tomasello, 2009). However, it is a rather peculiar kind of cooperation, as there is an apparent asymmetry in the division of costs and rewards between the teacher and the learner. The teacher incurs a significant cost (e.g., effort, time) to teach the learner even when there is little direct benefit to the self, while the learner reaps the rewards of the teacher’s knowledge while bypassing the costs of learning from directly exploring the world. How can such asymmetric relationships can be sustained, and even become so prevalent, in human societies? While teachers are typically considered to incur costs on behalf of the learner (Caro & Hauser, 1992), there may be sources of rewards that they reap from their act of teaching. This could come from the recursive reward a teacher gains through the learner’s reward (e.g., the joy of teaching), or other kinds of social rewards such as praise, prestige, or status. Furthermore, people sometimes provide information for others not to help others learn about the world but to deliberately show off something about ourselves. In such cases, they are still incurring costs as a “teacher” (i.e., provider of information) but the primary beneficiary is to benefit themselves rather than other learners. Thus the distribution of the rewards and costs between the teacher and the learner depends on the context. Future work might investigate how children

186  Hyowon Gweon reason about the cost-​reward distributions between agents in different social contexts, and how children modulate their learning or teaching accordingly. Indeed, there are many open questions about how young children navigate the complexities of learning in social contexts. Nevertheless, the studies reviewed here begin to paint a picture of young humans as active interpreters, evaluators, and communicators of information in social contexts. These inferential and communicative practices allow children to learn to learn about the world by making the best use of their rich social surroundings, and learn to help others learn by selecting the best set of evidence for others. These abilities may be rooted in basic social-​cognitive mechanisms for reasoning about the inner qualities of others, such as their goals, beliefs, preferences, and competences. So what makes human social learning so distinctive? The answers may lie in something we often take for granted: Our everyday common-​sense reasoning. The ability to understand other minds—​the basic component of human social cognition—​may give rise to the power and the richness of human social learning.

References Aslin, R. N., Saffran, J. R., & Newport, E. L. (1998). Computation of conditional probability statistics by 8-​month-​old infants. Psychological Science, 9(4), 321–​324. Baldwin, D., Loucks, J., & Sabbagh, M. (2008). Pragmatics of human action. In Shipley, T. F., & Zacks, J. M. (Eds.), Understanding Events: From Perception to Action (pp. 96–​129). New York: Oxford University Press. doi:10. 1093/​acprof:oso/​9780195188370.003.0006. Barner, D., Brooks, N., & Bale, A. (2011). Accessing the unsaid:  The role of scalar alternatives in children’s pragmatic inference. Cognition, 118(1),  84–​93. Bass, I., Bonawitz, L. B., & Gweon, H. (2017). Didn’t know, or didn’t show? Preschoolers consider epistemic state and degree of omission when evaluating teachers. In Gunzelmann, G., Howes, A., Tenbrink, T., & Davelaar, E. (Eds.), Proceedings of the 39th Annual Conference of the Cognitive Science Society (pp. 105–​110). Austin, TX: Cognitive Science Society. Bass, I., Hawthorne, D., Goodman, N. D., & Gweon, H. (2015). Not by number alone: The effect of teacher’s knowledge and its value in evaluating “sins of omission.” In Noelle, D. C., Dale, R., Warlaumont, A. S., Yoshimi, J., Matlock, T., Jennings, C. D., and Maglio, P. P. (Eds.), Proceedings of the 37th Annual Conference of the Cognitive Science Society (pp. 166–​171). Austin, TX: Cognitive Science Society. Birch, S. A. J., Vauthier, S. A., & Bloom, P. (2008). Three-​and four-​year-​olds spontaneously use others’ past performance to guide their learning. Cognition, 107(3), 1018–​ 1034. http://​doi.org/​10.1016/​j.cognition.2007.12.008. Bonawitz, E., Shafto, P., Gweon, H., Goodman, N. D., Spelke, E., & Schulz, L. (2011). The double-​edged sword of pedagogy: Instruction limits spontaneous exploration and discovery. Cognition, 120(3), 322–​330. http://​doi.org/​10.1016/​j.cognition.2010.10.001.

Understanding Others to Learn and Help Others Learn  187 Bridgers, S., Jara-​Ettinger, J., & Gweon, H. (2016). Children consider others’ expected costs and rewards when deciding what to teach. In Papafragou, A., Grodner, D., Mirman, D., & Trueswell, J. C. (Eds.), Proceedings of the 38th Annual Conference of the Cognitive Science Society (pp. 710–​715). Austin, TX: Cognitive Science Society. Bridgers, S., Jara-​Ettinger, J., & Gweon, H. (in press). Young children consider the expected utility of others' learning to decide what to teach. Nature Human Behaviour. Butler, L. P., & Markman, E. M. (2012). Preschoolers use intentional and pedagogical cues to guide inductive inferences and exploration. Child Development, 83(4), 1416–​1428. http://​doi.org/​10.1111/​j.1467-​8624.2012.01775.x. Carey, S. (2011). The Origin of Concepts. New York: Oxford University Press. Caro, T., & Hauser, M. (1992). Teaching in nonhuman animals. Quarterly Review of Biology, 67(151), 74. Chudek, M., Heller, S., Birch, S., & Behavior, J. H. E. A. H. (2012). Prestige-​biased cultural learning:  Bystander’s differential attention to potential models influences children’s learning. Evolution and Human Behavior, 33, 46–​56. http://​doi.org/​10.1016/​ j.evolhumbehav.2011.05.005. Clark, H. (1996) Using Language. New York: Cambridge University Press. Clegg, J. M., & Legare, C. H. (2016). Instrumental and conventional interpretations of behavior are associated with distinct outcomes in early childhood. Child Development, 87(2), 527–​542. http://​doi.org/​10.1111/​cdev.12472. Csibra, G., & Gergely, G. (2009). Natural pedagogy. Trends in Cognitive Sciences, 13(4), 148–​153. Csibra, G., & Gergely, G. (2011). Natural pedagogy as evolutionary adaptation. Philosophical Transactions of the Royal Society B: Biological Sciences, 366(1567), 1149–​ 1157. http://​doi.org/​10.1098/​rstb.2010.0319. Csibra, G., & Shamsudheen, R. (2015). Nonverbal generics:  Human infants interpret objects as symbols of object kinds. Annual Review of Psychology, 66(1), 689–​710. Dean, L. G., Kendal, R. L., Schapiro, S. J., Thierry, B., & Laland, K. N. (2012). Identification of the social and cognitive processes underlying human cumulative culture. Science, 335(6072), 1114–​1118. http://​doi.org/​10.1126/​science.1213969. Gergely, G., & Csibra, G. (2003). Teleological reasoning in infancy: The naive theory of rational action. Trends in Cognitive Sciences, 7(7), 287–​292. Goodman, N. D., & Frank, M. C. (2016). Pragmatic language interpretation as probabilistic inference. Trends in Cognitive Sciences, 20(11), 818–​829. http://​doi.org/​10.1016/​ j.tics.2016.08.005. Gopnik, A. (2012). Scientific thinking in young children: Theoretical advances, empirical research, and policy implications. Science, 337(6102), 1623–​1627. http://​doi.org/​ 10.1126/​science.1223416. Gopnik, A., & Astington, J. (1988). Children’s understanding of representational change and its relation to the understanding of false belief and the appearance-​reality distinction. Child Development, 59(1),  26–​37. Gopnik, A., & Wellman, H. M. (2012). Reconstructing constructivism: Causal models, Bayesian learning mechanisms, and the theory theory. Psychological Bulletin, 138(6), 1085–​1108. http://​doi.org/​10.1037/​a0028044. Grice, H. P. (1975). Logic and conversation. In Cole, P., & Morgan, J. (Eds.), Syntax and Semantics: vol. 3, Speech Acts (pp. 41–​58). New York: Academic Press. Gweon, H., & Asaba, M. (2017). Order matters: Children’s evaluation of underinformative teachers depends on context. Child Development. doi: 10.1111/​cdev.12825.

188  Hyowon Gweon Gweon, H., Pelton, H., Konopka, J. A., & Schulz, L. E. (2014). Sins of omission: Children selectively explore when teachers are under-​informative. Cognition, 132(3), 335–​341. http://​doi.org/​10.1016/​j.cognition.2014.04.013. Gweon, H., & Schulz, L. E. (2011). 16-​month-​olds rationally infer causes of failed actions. Science, 332(6037), 1524. Gweon, H., & Schulz, L. E. (2019). From exploration to instruction: Children learn from exploration and tailor their demonstrations to observers’ goals and competence. Child Development, 90(1), e148–​e164. doi:10.1111/​cdev.13059. Gweon, H., Shafto, P. & Schulz, L.E. (2018). Development of children's sensitivity to over-​ informativeness in learning and teaching. Developmental Psychology, 54(11), 2113–​ 2125. doi: http://​dx.doi.org/​10.1037/​dev00005802113 Gweon, H., Tenenbaum, J. B., & Schulz, L. E. (2010). Infants consider both the sample and the sampling process in inductive generalization. Proceedings of the National Academy of Sciences, 107(20), 9066–​9071. http://​doi.org/​10.1073/​pnas.1003095107. Hewlett, B. S., Fouts, H. N., Boyette, A. H., & Hewlett, B. L. (2011). Social learning among Congo Basin hunter-​gatherers. Philosophical Transactions of the Royal Society B: Biological Sciences, 366(1567), 1168–​1178. http://​doi.org/​10.1016/​j.socscimed. 2009.05.016. Jara-​Ettinger, J., Gweon, H., Schulz, L. E., & Tenenbaum, J. B. (2016). The Naïve Utility Calculus: Computational principles underlying commonsense psychology. Trends in Cognitive Sciences, 20(8), 589–​604. http://​doi.org/​10.1016/​j.tics.2016.05.011. Koenig, M. A., & Harris, P. L. (2005). Preschoolers mistrust ignorant and inaccurate speakers. Child Development, 76(6), 1261–​1277. http://​doi.org/​10.1111/​ j.1467-​8624.2005.00849.x. Kline, M. A. (2015). How to learn about teaching: An evolutionary framework for the study of teaching behavior in humans and other animals. Behavioral and Brain Sciences, 38, e31. http://​doi.org/​10.1017/​S0140525X14000090. Kushnir, T., Xu, F., & Wellman, H. (2010). Young children use statistical sampling to infer the preferences of other people. Psychological Science, 21(8), 1134. Lane, J. D., Wellman, H. M., & Gelman, S. A. (2013). Informants’ traits weigh heavily in young children’s trust in testimony and in their epistemic inferences. Child Development, 84(4), 1253–​1268. http://​doi.org/​10.1111/​cdev.12029. Legare, C. H., & Nielsen, M. (2015). Imitation and innovation: The dual engines of cultural learning. Trends in Cognitive Sciences, 19(11), 688–​699. http://​doi.org/​10.1016/​ j.tics.2015.08.005. Liszkowski, U., Carpenter, M., & Tomasello, M. (2008). Twelve-​month-​olds communicate helpfully and appropriately for knowledgeable and ignorant partners. Cognition, 108(3), 732–​739. http://​doi.org/​10.1016/​j.cognition.2008.06.013. Liu, S., Ullman, T. D., Tenenbaum, J. B., & Spelke, E. S. (2017). Ten-​month-​old infants infer the value of goals from the costs of actions. Science, 358(6366), 1038–​1041. Ma, L., & Xu, F. (2013). Preverbal infants infer intentional agents from the perception of regularity. Developmental Psychology, 49(7), 1330–​1337. http://​doi.org/​10.1037/​ a0029620. MacDonald, K. (2007). Cross-​cultural comparison of learning in human hunting. Human Nature, 18(4), 386–​402. Meltzoff, A. N. (2007). “ ‘Like me”: A foundation for social cognition. Developmental science, 10(1), 126–​134.

Understanding Others to Learn and Help Others Learn  189 Rendell, L., Boyd, R., Cownden, D., Enquist, M., Eriksson, K., Feldman, M. W., et  al. (2010). Why copy others? Insights from the social learning strategies tournament. Science, 328(5975), 208–​213. http://​doi.org/​10.1126/​science.1184719. Repacholi, B., & Gopnik, A. (1997). Early reasoning about desires: Evidence from 14-​and 18-​month-​olds. Developmental Psychology, 33(1),  12–​20. Rogoff, B. (2003). The Cultural Nature of Human Development. Oxford:  Oxford University Press. Ronfard, S., & Corriveau, K. H. (2016). Teaching and preschoolers’ ability to infer knowl­ edge from mistakes. Journal of Experimental Child Psychology, 150, 87–​98. http://​ doi.org/​10.1016/​j.jecp.2016.05.006. Ronfard, S., Was, A. M., & Harris, P. L. (2016). Children teach methods they could not discover for themselves. Journal of Experimental Child Psychology, 142, 107–​117. http://​ doi.org/​10.1016/​j.jecp.2015.09.032. Schulz, L. (2012). The origins of inquiry: Inductive inference and exploration in early childhood. Trends in Cognitive Sciences, 16(7), 382–​389. http://​doi.org/​10.1016/​ j.tics.2012.06.004. Sabbagh, M. A., & Baldwin, D. A. (2001). Learning words from knowledgeable versus ignorant speakers: Links between preschoolers’ theory of mind and semantic development. Child Development, 72(4), 1054–​1070. Shafto, P., Goodman, N. D., & Frank, M. C. (2012). Learning from others:  The consequences of psychological reasoning for human learning. Perspectives on Psychological Science, 7(4), 341–​351. http://​doi.org/​10.1177/​1745691612448481. Shafto, P., Goodman, N. D., & Griffiths, T. L. (2014). A rational account of pedagogical reasoning: Teaching by, and learning from, examples. Cognitive Psychology, 71, 55–​89. http://​doi.org/​10.1016/​j.cogpsych.2013.12.004. Shafto, P., Gweon, H., Fargen, C., & Schulz, L. (2012). Enough is enough: Inductive sufficiency guides learners’ ratings of informant helpfulness. In Miyake, N., Peebles, D., & Cooper, R. P. (Eds.), Proceedings of the 34th Annual Conference of the Cognitive Science Society (pp. 977–​982). Austin, TX: Cognitive Science Society. Shneidman, L., Gweon, H., Schulz, L. E., & Woodward, A. L. (2016). Learning from others and spontaneous exploration:  A cross-​ cultural investigation. Child Development, 87(3), 723–​735. http://​doi.org/​10.1111/​cdev.12502. Skordos, D., & Papafragou, A. (2016). Children’s derivation of scalar implicatures: Alternatives and relevance. Cognition, 153, 6–​18. http://​doi.org/​10.1016/​j.cognition. 2016.04.006. Strauss, S., Calero, C. I., & Sigman, M. (2014). Trends in neuroscience and education. Trends in Neuroscience and Education, 3(2), 38–​43. http://​doi.org/​10.1016/​j.tine.2014.05.001. Téglás, E., Vul, E., Girotto, V., Gonzalez, M., Tenenbaum, J. B., & Bonatti, L. L. (2011). Pure reasoning in 12-​month-​old infants as probabilistic inference. Science, 332(6033), 1054–​1059. http://​doi.org/​10.1126/​science.1196404. Thornton, A., & McAuliffe, K. (2006). Teaching in wild meerkats. Science, 313(5784), 227–​229. http://​doi.org/​10.1126/​science.1128727. Tomasello, M. (2009). Why We Cooperate. Cambridge, MA: MIT Press. Wellman, H. M., Cross, D., & Watson, J. (2001). Meta-​analysis of theory-​of-​mind development: The truth about false belief. Child Development, 72(3), 655–​684. Wood, L. A., Kendal, R. L., & Flynn, E. G. (2012). Context-​dependent model-​based biases in cultural transmission:  Children’s imitation is affected by model age over model

190  Hyowon Gweon knowledge state. Evolution and Human Behavior, 33(4), 387–​394. http://​doi.org/​ 10.1016/​j.evolhumbehav.2011.11.010. Xu, F., & Garcia, V. (2008). Intuitive statistics by 8-​month-​old infants. Proceedings of the National Academy of Sciences, 105(13), 5012–​5015. http://​doi.org/​0704450105 [pii] 10.1073/​pnas.0704450105. Xu, F., & Kushnir, T. (2013). Infants are rational constructivist learners. Current Directions in Psychological Science, 22(1), 28–​32. http://​doi.org/​10.1177/​0963721412469396. Xu, F., & Tenenbaum, J. B. (2007). Sensitivity to sampling in Bayesian word learning. Developmental Science, 10(3), 288–​297. http://​doi.org/​10.1111/​j.1467-​7687.2007.00590.x.

10 How Do Partial Understandings Work? Frank Keil

We know surprisingly little about how understandings can work when they are incomplete. It may seem obvious that full understandings of why phenomena occur or how things work are rare in the real world; but the extent to which our understandings are incomplete and the ways they are incomplete are only recently being explored. Here, I argue that a closer look at the degrees and types of incompleteness raises two questions: (1) Given how incomplete many understandings are how could they possibly ever be of any use? (2) Do the contents that do exist in partial understandings suggest a different account of what understandings are and how they are used not just by laypeople but even by scientists? For the purposes of this article, I define “full understanding” as knowing all the relevant components and their causal relations that explain how an entity or class of entities work or why they have the properties they do. The precise details can be difficult to fully specify given that causal relevance of properties and components can often be on a continuum and it may be unclear what one really needs to know to understand how something works as opposed to what may be causally necessary but does not provide distinctive insight. For example, while it is obvious that the compressor is a critical part of an understanding of how a refrigerator works and that the bolts that hold the parts of the refrigerator are not (even though they are causally necessary), it is less clear whether the radiator fins on the back are essential like the compressor or more supplemental. Nonetheless, for many systems, there is usually a rough consensus on what at least the minimal set of such elements and relations is. For a refrigerator, those components are of the sort found in the following description, where each component is boldfaced: The compressor takes a special gas that evaporates easily and compresses it into liquid form outside the refrigerator, the liquid then is pushed by the compressor through an expansion valve that pushes the liquid into the

192  Frank Keil refrigerator into tubes known as evaporators. As the liquid expands into gas inside the evaporators it absorbs heat from inside the refrigerator, thereby cooling the inside. The lower pressure gas then goes outside the refrigerator into the compressor which then compresses the gas and pushes the gas into tubes known as the condenser, where the gas releases its heat into the room with the help of radiator fins that dissipate the heat into the air. Because it is under higher pressure and losing heat, the gas turns into a liquid again and is pushed through the expansion valve for another cycle of cooling.

We normally think of this level of granularity as constituting a “mechanistic” understanding, in which a higher-​level function (e.g., cooling) is understood in terms of the causal interactions of all the key components (Craver & Bechtel, 2006). In addition, we will assume that mechanistic understanding in this sense can be “full” at each level immediately below the level being explained (Craver & Bechtel, 2006). Thus, if we are trying to understand how a refrigerator cools its contents, we would want to know all the causal interactions between its functional parts at the first coherent level immediately below the overall entity (e.g., the compressor, the gas, the condenser, the expansion valve, and the evaporator, but not the inner workings of each of those parts such as how a compressor works in terms of its motor, pump, and valves). When considering some systems a bit more closely, it quickly becomes apparent that many levels and details are often involved in any attempt to provide an “exhaustive explanation.” This proliferation of levels is especially common for natural kinds and the natural kind constituents of engineered devices. Thus, even a relatively simple device like a refrigerator unpacks fairly quickly into not yet fully understood areas of physical chemistry of gases and liquids, materials science and thermodynamics. The same explosion of complexity across multiple levels occurs even more dramatically for most biological systems, even at the cellular level. For example, an explanation of how a cell reproduces itself in mitosis may start in the seemingly straightforward level found in high school biology textbooks: prophase → metaphase → anaphase → telophase → cytokinesis. Each phase plays a sensible quasi-​mechanical role that sets up the next one. Chromosomes are formed and duplicated, then sorted into alignment, and then pulled along spindles into new clusters and eventually split into new cells. But each of these steps then unpacks into extraordinary levels

How Do Partial Understandings Work?  193 of complexity. Consider, for example, one small part of the metaphase: the alignment of the chromosomes. Even among those who are metaphase specialists, there are subspecialists who study the intricate problem of how the chromosomes come into such accurate alignment where even very small misalignments can result in lethal chromosomal anomalies. Major programs of active research are underway on the signaling sequences that regulate how the protein micro cables tug at the spindle in just the right ways to achieve perfect alignment (e.g., Kiyomitsu, 2016). As a consequence of these investigations there is a proliferation of highly focused areas of specialization and expertise. These divisions of labor in engineering and biology make two points: (1) mechanistic understandings can decompose again and again into other levels of equal or greater complexity, as why and how questions beget further why and how questions; and (2) even scientists can work successfully at one level while having major gaps in their own understandings at lower levels, gaps that may be understood by others or that may remain unknown to all. Sometimes, all one has at those lower levels are “black boxes” and a rough characterization of their inputs and outputs. Nonetheless, scientists at the upper levels often work successfully, making mechanistically motivated predictions and providing insightful explanations. Science makes progress because researchers are able to find levels where they can achieve relatively complete mechanistic understandings of the relations between various functional “black boxes” whose contents may be only partially understood. It therefore might seem that laypeople’s informal science works in the same manner, but just with somewhat larger black boxes whose contents are more complete mysteries. Laypeople might understand refrigerators as compressing special gases into liquids and then quickly expanding them so as to make them evaporate. That evaporation in turn enables them to have cooling effects in pipes that are configured so as to convey heat from the inside to the outside, as in the simple description offered earlier. Perhaps most people in developed societies could come up with a rough schematic drawing along these lines based on exposure to a few examples. If they could not generate a schematic, perhaps they could tell a functionally correct schematic from an incorrect non-​functional one. At the very minimum, they should be able to know the extent of their own understanding and be aware where their gaps are. In short, it seems plausible that, at sufficiently high levels of interactions of large functional units, people should have simple but workable mental models of how common devices work and how biological

194  Frank Keil systems function. To be sure, these are highly simplified schematic models that incorporate all sorts of idealizations (e.g., not assuming the effects of gas friction in compressor pipes in a refrigerator, etc.) and sometimes laypeople may have coherent but incorrect mental models such as a visualizing electrons as miniature “planet” particles orbiting nuclei rather than as probabilistic clouds. But, putting aside such simplifications and occasional errors, it seems plausible that laypeople grasp the world through such working mental models, which are used to guide their interventions and predictions. The reality, however, is very different. The vast majority of even the most educated people have hardly any such models at all. They fail dismally and catastrophically in attempts to provide such models. Even worse, they are usually blissfully unaware of the degree to which their understanding is fragmentary and incomplete. The problem has been documented in many ways. People cannot make accurate predictions of how simple clusters of gears will rotate when a specific gear is turned (Hegarty, 2004; Lehrer & Schauble,1998); readers of passages explaining phenomena miss blatant contradictions in sentences only a few lines apart (Schommer & Surber, 1986); and homeowners temporarily turn thermostats up to higher than desired levels to “heat their houses up faster” even though this completely misunderstands thermostat function (Peffer et al., 2011). A particularly egregious case was vividly shown in study with laypeople’s grasp of the simplest principles of bicycle mechanics (Lawson, 2006). When laypeople in Britain were shown schematic drawings of bicycles in which one functional diagram was contrasted with three non-​functional ones, they made frequent errors in selecting workable bicycles. For example, they often judged as acceptable a drawing in which the chain wrapped around the hub of the rear wheel and the hub of the front wheel, creating a “bicycle” whose front wheel could never be turned. Even avid members of cycling clubs sometimes made errors, albeit at lower rates. Moreover, all these errors were in recognition, not just in recall. Thus, the problem is not simply one of “dusting off ” hazy but intact memories.

1. Discounting Decay The surprising gaps in people’s understandings might be attributable to failures in education. Perhaps the problem is simple lack of exposure to mechanistic information and once people achieve mastery through correct

How Do Partial Understandings Work?  195 instruction, they can easily retain idealized mental models. Unfortunately, they cannot even when they reach high peaks of temporary mastery. We have informally shown this by assessing undergraduates who achieved near-​perfect scores on high school standardized tests of advanced biology curricula as little as one year earlier. In particular, we asked them to draw the Krebs cycle, a major component in standardized assessments that they had answered at near perfect levels on their tests. Despite remembering having learned the Krebs cycle in considerable detail, their recall was stunningly empty. Their drawings mostly consisted of two arrows indicating a cycle with large question marks and a stray comments such as “ATP becomes ADP or the other way around,” or “something is broken down into something different.” There was never even a hint of the components of the cycle, how they interacted, or the major inputs and outputs. To assess this kind of decline in knowledge more formally, we conducted a study on decline of knowledge among several popular college STEM (science, technology, engineering, and mathematics) majors (Fisher & Keil, 2016). We first asked college graduates to rate their abilities to give explanations about phenomena that had been central to their majors and which had been clearly covered by all common textbooks in their courses. For example we asked biology majors about the Krebs cycle, computer science majors about object-​ oriented programming, and chemistry majors about thermodynamics. We then asked them to provide the explanations and then to rerate their explanatory understanding. There was a huge disconnect in which they realized they had far less explanatory knowledge than they originally thought. Follow-​up studies revealed that they had originally thought their level of current understanding had only declined a bit from their original level of mastery when in fact it had been massive. More recently we have documented this decay-​ neglect phenomenon in detail for students taking a particular college-​level course with the same instructor in Introductory Psychology (Fisher & Keil, in preparation). We were able to ask students to rate their abilities to explain various phenomena covered in the course and then answer the precise exam questions they had taken months to years earlier. They rated their knowledge as barely declining at all when in fact it had declined massively and monotonically over time. This decay neglect is part of a broader phenomenon that has been extensively explored in our lab in detail (Fisher & Keil, 2016; Mills & Keil, 2004; Rozenblit & Keil, 2002)  and more recently in several other labs as well (Alter, Oppenheimer & Zemla, 2016; Fernbach et al., 2013). We call it the

196  Frank Keil “illusion of explanatory depth,” or IOED. The IOED is the tendency to think one understands the world in far more explanatory depth than one really does. As a result, we are often unaware of the gaps in our understanding until we are confronted with rare cases in which we are asked to provide full explanations. The most common way to demonstrate the IOED is to train people on a scale of explanation quality, illustrating a range of very weak to very strong explanations of how a device such as a crossbow works. They are then asked to rate their abilities to explain a long list of phenomena. We ask them to do so fairly quickly so as to ensure that they do not actually try to mentally flesh out a full explanation before giving an initial rating. We then ask them to write out in detail full explanations of a few of the items from the lists they just rated. We then ask them to rerate their understanding in light of the written explanations. In some studies, we go on to ask them critical questions and show them excellent explanations, each time asking them to rerate their understanding. In most cases we find the same result: ratings drop after they have to provide explanations themselves, answer questions, or see expert ratings. People are often genuinely surprised at just how incomplete their understandings are. Moreover the effect holds primarily for explanatory knowledge and not for knowledge of facts, procedures, or narratives. People are seduced into overestimating the richness of their explanatory knowledge in particular. There are several converging reasons for why the effect occurs more for explanations: we rarely attempt to provide full explanations and as a result have little data on failing, we often confuse having higher-​level functional or procedural knowledge with lower-​level mechanistic knowledge, and we confuse expertise in other minds with our own. These factors are revealing in their own right and are considered further in later sections of this chapter.

2.  Understanding and Explaining Understanding and explaining are intimately related but also clearly distinct, and the differences may be critical to an account of why we have partial understandings and how we cope with them. We can fully understand how a device works without being able to fully explain it to others because part of our understanding is nonverbal or even tacit. For example, I may understand why a particular knot works best for a rigging problem in a sailboat

How Do Partial Understandings Work?  197 for reasons having to do with how it is tied and untied under difficult-​to-​ describe adverse conditions in inclement weather. I  may understand perfectly why an animal engages in a peculiar foraging behavior but have great difficulty explaining the ecological and geospatial constraints to others not familiar with that specific niche. Perhaps less intuitively, the reverse situation may also occur, namely being able to explain a phenomenon completely to another without fully understanding it oneself. This may occur when explaining a procedure to a group that is more knowledgeable than oneself about a topic. For example, I might explain to a group of data scientists that I used a particular measure of data concordance over another because of the large sample size, as recommended by statistical expert sources. My audience may nod with a clear understanding of why that was the best choice while I do not fully grasp the underlying math and take it on faith. This can occur routinely in the sciences where one uses a particular procedure because experts fully understand why it is the best to use and one only partially grasps that it is better and is preferred. It is this reliance on networks of experts that may help us understand the role of partial mechanistic understanding.

3.  A Paradoxical Interest in Mechanism? Something odd seems to happen every day with young children. They seem to be preferentially fascinated with mechanism over other forms of information. Initially, this has been shown in studies of their spontaneous question asking. A child asks why a phenomenon occurs or why an entity has a particular property and then receives an answer from an adult. If the adult answers with a causal mechanistic response, the child is often satisfied with the answer and queries cease (although some children can persist with an endless regress of further mechanistic questions, this is less common). However, if an adult gives a non-​mechanistic response, the child is much less likely to be satisfied with the answer and will persist in asking for an answer to “why” or “how” until one is received (Callanan & Oakes, 1992; Chouinard, 2007; Frazier, Gelman & Wellman, 2009; 2016). For example, if in response to the question of why flashlights have batteries, an adult were to reply “They do have batteries . . . that’s right” or “Batteries are round and are shiny at each end,” the child would be likely to re-​ask the question, frown, or reject the response or even try to invent their own mechanistic reason. However, if

198  Frank Keil the adult were to reply in the following manner: “Batteries have energy that makes the bulb light up,” the child is much more likely to be content with the response. Even in cases where an adult explicitly uses “because” in an answer, preschoolers may reject the answer if it doesn’t include any real causal information and instead is empty, as illustrated by the following different responses (from Corriveau & Kurkul, 2014): C: Why does it rain? A1: “It rains because water falls from the sky and gets us wet.” A2: “It rains because the clouds fill with water and get too heavy.” Even three-​year-​olds are happier with A2’s response, presumably because it gives a real reason instead of simply restating the phenomenon. Children are particularly interesting with respect to partial understandings because their levels of forgetting of mechanistic details and corresponding illusions of deep understanding are even larger, often much worse than is found in adults (Mills & Keil, 2004). It is therefore especially important to know why they should be so interested in mechanisms if they quickly forget the details. To begin to answer this question we need to better understand the early preference for mechanism, as there are several limitations to the studies on spontaneous question asking. First, it is difficult to have tight control over the stimuli in a way that is consistent across children. Even when they are asking about the same thing, they may ask in subtly different ways that may have different optimal answers. Second, the reasons for their preferences are unclear. Do they prefer mechanistic responses because they merely give additional information that provides a legitimate reason even if it isn’t specifically mechanistic, or do they actually prefer mechanistic reasons over other reasons? Do they show a preference for a particular kind of mechanistic reason even before they know its content or fully understand it? Do they seek exposure to mechanistic information as a way of tracking access to the best sources of future information and help even if they cannot fully grasp the information when they initially receive it? To answer these questions, we have been conducting a series of studies looking at preferences for explanations to specific questions. This explanation choice paradigm allows for careful contrasts of explanatory responses to specific questions. Consider an example study (Lockhart, Kerr & Keil, 2017). We asked who knew more about a topic, someone who gave a reason that was mechanistic or someone who gave a reason that made sense but did not entail knowledge of mechanism.

How Do Partial Understandings Work?  199 Thus, we presented children aged five to seven years and adults with choices of the following sort: Two people are saying why they think a particular electric car is the right one to buy. They agree on which car is the right one, but they give different reasons. Which person do you think knows more about electric cars? Person 1: who says car X is the best because the colors to choose from are very pretty. When you turn it on, it makes a cool sound. The seats inside are very comfortable to sit in and the outside of the car is shiny because the builders use a new, special kind of paint. Person 2: who says that X is the best because the motor uses high-​quality wire that makes it run very smoothly. When you drive, you can speed up and stop quickly. Also, it has a very powerful battery and the battery can be recharged quickly.

Whether it was knowledge about the car or the best fruit to eat, the children strongly preferred the person who talked about the workings of the car, or the nutritional value of the fruit, as opposed to reasons that, while plausible, focused on surface properties unrelated to an understanding of the causal factors that explained the core functioning of the kind. This is a relatively subtle inference for such young children to make. It is equally plausible that someone might buy a car because of its colors and seating rather than its mechanics, indeed even more plausible in many cases. In addition, there is a wealth of non-​mechanistic trivia that one could know about cars or fruits, but children attributed more true knowledge to those who gave mechanistic responses. Some of their justifications made it clear that they sensed a more substantial, interconnected kind of knowledge. For example, one child, in rejecting the person who talked about car colors and seats, said that person didn’t know much “because that’s just fluff.” Clearly there must be limits to early preferences for mechanisms, as it is highly implausible that a young child would choose to continue to receive more and more reductionist elaborations on the mechanistic underpinnings of components whose interactions are fully described at a higher level. But even with these inevitable boundary conditions, it is remarkable that children seem so driven to acquire mechanistic details when they so quickly forget them. What possible utility could such exposure have? It is here that partial understanding may play its most important role.

200  Frank Keil

4.  Mechanism as a Critical Conduit Perhaps the child’s quest for mechanistic information is a quixotic one that is doomed to failure as the information quickly fades away; but an alternative account is now gaining considerable support. Exposure to mechanism is an essential way of gathering more abstract information about kinds and systems, information that is much more enduring and gives children and adults valuable competencies for navigating the world with partial understandings. One piece of evidence for such abstracted information comes from the phenomenon of “relearning,” namely the greater speed and efficiency with which students learn information the second time around even when they seem to have retained nothing from their first learning experience. This is a common lament by instructors of college courses that have introductory courses as prerequisites, whether it is a cellular biology course that requires introductory biology or a developmental psychology course that requires introductory psychology. For example, the cellular biology students may claim to remember nothing about the Krebs cycle, and indeed can fail dismally on a test of its particulars, but they reacquire the information much more quickly and efficiently. What is it that they bring to the relearning situation? What information persists in memory that facilitates learning? We have been investigating what is extracted from exposure to mechanism that endures in memory and supports relearning as well as a multitude of other tasks. Our inventory of what endures is still in progress, but some elements are clear. One major factor is extracting a sense of the causal complexity of a system. In a series of studies, we have shown that children and adults alike achieve strikingly consensual judgments about the relative causal complexity of devices and biological systems such as organs (Kominsky, Zamm, & Keil, in press). They do so despite having almost no insight into mechanistic details. Moreover, their intuitions are clearly distinct from those about visual complexity or about mere numbers of internal parts; instead they zero in on more subtle markers of causal mechanistic complexity. The task involves training children and adults how to use a scale for rating “how complicated these things are in terms of how they work.” They then rate devices ranging from flashlights to cell phones to submarines or they rate body parts ranging from the knee to the eye. Beginning at least by age seven, children tend to agree on what are mechanistically simple or complex items. Younger children do show an additional tendency to conflate ease of use with simpler mechanisms, but never entirely

How Do Partial Understandings Work?  201 so. It also seems that even young children are not fooled by the visual complexity but mechanistic simplicity of a box of jumbled parts or by entities with massive numbers of functional parts that are intrinsically simple, such as beanbag chairs. They know that true mechanistic complexity requires an orderly transmission of factors from one functional unit to another and that heterogeneously structured modules with clear reliable interactions are signs of mechanistic complexity. What good are such intuitions about mechanistic complexity? One use is to sense when one needs to defer to others. Thus, even young children link greater complexity to a greater perceived need to recruit others to help understand how entities work or can be fixed when broken. They also use such intuitions to guide hunches about what sorts of things one could learn about on one’s own as opposed to needing additional instructional support from others (Lockhart et al., 2016). In this way complexity serves as a guide for when to allocate cognitive resources for autonomous learning and discovery and when to allocate them for seeking out and evaluating informants. There is a great deal more to be said about emerging intuitions about mechanistic complexity. For example, during the early school years children learn to be more adept at inferring internal complexity. One heuristic they use involves tracking the number and diversity of behaviors or functions that an entity engages in. In one set of studies, we described two machines: one that washed pants and blue jeans and another that washed pants and drinking glasses (Ahl & Keil, 2017). A different condition contrasted a machine that did one thing (e.g., washed pants) with one that did two things (e.g., washed pants and glasses). Children were then shown images of the inner workings of two machines, one obviously more complex than the other (as depicted by more heterogeneous interconnected parts). They were asked which insides went with which device. A consistent developmental pattern emerged across several devices with different sets of functions/​behaviors. While younger children (under seven years) strongly preferred the more complex insides with the devices with more functions, when the numbers of functions were matched, they were unable to use greater diversity as a cue to greater internal complexity. This was despite their being easily able to judge the more diverse machine as doing “more different” things. They just didn’t connect diversity to greater mechanistic complexity. In contrast older children immediately inferred such a relationship. Apparently during the early school years, there is a major emerging insight into additional ways in which behavioral variation is revealing about

202  Frank Keil complexity. We are currently engaged in a series of studies showing that brief mechanistic explanations of how things such as door locks work result in shifts in judged complexity by children toward more consensual values and that such intuitions persist for weeks after the mechanistic details fade. We see inferences about mechanistic complexity as emerging through iterative interactions between exposure to mechanisms, information about functional and behavioral variation and even the existence of widely recognized experts. Each facet informs the other and motivates closer tracking of complexity.

5. Beyond Complexity Complexity is only one small facet of what emerges from exposure to mechanistic explanations. We are actively exploring other traces that endure after learning about how thing work in concrete detail. At a minimum the following all seem to be present: causal centrality, causal potency, causal relevance, temporal patterns, kinds of causality, large scale schematic patterns, and trends. Each is briefly considered here to illustrate the potential scope of the consequences of exposure to mechanism. Many of these may also be tacitly encoded. In general, they provide interpretative lenses, not tight predictions or specific mental models. What they do support is discussed in the final section of this chapter. Causal centrality is knowledge of what features of a system play the most critical causal roles, what ones have the biggest consequences if they are different or counterfactualized. These can be couched in terms of property types or functional components. Thus, one might believe that color is causally peripheral property type for bicycles but a causally central one for flowers. Conversely, size may be seen as more causally central to bicycles than to flowers. Although even young children have some intuitions about the most central property types for familiar domains (e.g., Keil et al., 1998), exposure to mechanism may greatly sharpen those intuitions. It may also provide insight into key functional components. Thus, upon exposure to the workings of a cylinder lock, a child may greatly increase their sense of the importance of small movable internal parts (i.e., the pins) even as he forgets the details of how they work. Similarly, after exposure to the heart, a child may place greater emphasis on the interactions among its chambers even if she cannot remember how they actually interact and control blood flow.

How Do Partial Understandings Work?  203 Causal potency and relevance are related but slightly weaker notions. A child might learn that, for most transportation devices, drive motors are nearly always relevant while surface accessories (such as cup holders) are not. They might also learn there is always a need for a steering mechanism even if it is not considered causally central. Thus, they might sense that while both motors and steering mechanisms are relevant, motors are more potent. Temporal patterns are a sense of the time courses associated with a bounded phenomenon. Is it glacial or rapid? Does it have smooth transitions or more abrupt staccato ones? Does it build to a crescendo or is it relatively steady state? All of these and more can emerge from exposure to a mechanism and sometimes they can be surprisingly informative. For example, simply knowing that human cellular mitosis never occurs faster than 20 hours for a complete division cycle, provides a sense of constrained chemical/​mechanical interactions that shift how one studies the process. Similarly; knowing that neurons transmit signals at roughly 100 meters per second, plus delays to cross synapses, radically changes one’s views of how event perception and perceptual motor responses must work. There may also be much more general heuristic insights such as that the smaller the scale of a system the faster the oscillations or cycles that typically occur, a striking observation made by Herb Simons in his classic Sciences of the Artificial (Simon, 1996). We are just beginning to understand the importance of sensing temporal patterns in our interpretations of causality (e.g., Rottman & Keil, 2012). Kinds of causality can range from very broad domain differences such as physical vs. psychological causality to physical contrasts such as direct contact force transmissions vs. action at a distance. There is ample evidence that even infants sense that social causation works differently from physical mechanical interactions (e.g., Gao, Newman & Scholl, 2009; Mascaro & Csibra, 2012; Newman et al., 2010; Rochat, Striano & Morgan, 2004). In terms of exposure to mechanism, these contrasts may be critical. Consider for example mechanistic descriptions for how a New Caledonian crow uses a tool to retrieve a desired food from a narrow container (McGrew, 2013). One description might focus on physical reflex loops and their tuning, while another may focus on problem-​solving and decision processes. These contrasting framings matter greatly for guiding further exploration and discovery. Kinds of causality fade into large-​scale schematic patterns. These involve senses of distinctive causal structures that are generally too high level to generate specific predictions but nonetheless constrain what kinds

204  Frank Keil of processes are expected and how systems are interpreted. To give one example, it appears that people have a strong tendency to interpret social events as having branching probabilistic causes and branching probabilistic effects but see physical events as more likely to have deterministic linear causes but branching effects (Strickland, Silver & Keil, 2017). It is too early to know the extent to which exposure to mechanisms influences such interpretative biases, but such influences seem likely. Other large-​scale causal schematic expectations can be whether a system is conceived as involving common cause vs. common effect patterns (Waldmann, Hagmayer, & Blaisdell (2006); cycles and positive or negative feedback loops (Kim et  al., 2009; Rehder, 2017), preventing vs. enabling forces (Wolff, 2007), tipping points, snowballing effects, serial vs. parallel processes, causal cascades, and causal homeostasis (Boyd, 2010). While this list is not exhaustive, it seems likely that the total set of such causal expectations may be quite modest, on the order of a dozen or two, and that people, at least laypeople, may typically select one or more of these patterns to understand virtually any phenomenon. Exposure to mechanism would tend to align certain patterns with certain broad areas such as understandings of infectious diseases, alarm systems, nutrition, and vehicles. Research designed to catalog such schemas has only just begun.

6.  How Cognitive Traces Are Used The argument so far is that cognitive traces from exposure to mechanism, such as complexity, causal relevance, and broad schematic patterns, persist in memory long after mechanistic details fade. While evidence is still accumulating, it is already clear that a diverse set of traces do indeed endure. What remains is an account of the utility of such traces. We’ve seen evidence for one role for complexity, namely guiding hunches about when one might most want to seek out help from experts and rely heavily on deference. However, when combined with other types of traces, the utility is far more than knowing when to defer. Traces also guide whom to defer to, evaluations of expert and explanation quality, real-​time entity exploration, manipulation and reverse engineering, diagnosis and repair of entities, creation of summary gists, and relearning. Consider a few examples of how these might occur and how they are all relevant to how we cope with problems of partial understanding. With respect to

How Do Partial Understandings Work?  205 domains of deference, factors such as causal relevance and abstract schematic patterns can guide us toward appropriate experts and classes of explanations even when our own understanding is quite modest. In one series of studies we showed how the expanding appreciation of abstract causal schemas (such as that a problem involves force exchanges between bounded objects) enabled children to make successively more refined judgments as they grew older about the divisions of cognitive labor around them (Keil et al., 2008). In another set of studies, we showed how intuitions about appropriate gists to generate from more complex detail explanations can be partial guides by providing views of causal relevance and abstract causal structure (Rottman & Keil, 2011). Those studies did not directly link the insights offered by factors such as causal relevance and abstract schema to prior exposure to mechanisms, but such an influence is strongly implied and an area of active research. Similarly, a sense of causally relevant prop­ erty types and schemas for a domain enable one to evaluate the quality of explanations or choose between competing experts. For example, if one is evaluating “experts” on the efficacy of vaccines, beliefs about relevant timing patterns and growth rates may rule out the explanations by an alleged expert who uses inappropriate timing and growth patterns in his explanation. Just how much leverage can be achieved by different degrees and kinds of partial understanding remain as important research topics for further investigation, but the potential seems very large. As a final example, consider once again relearning. Despite few studies on what is retained in memory, either explicitly or implicitly, that enables faster learning of material the second time around, there are hints of how exposure to mechanistic explanations might matter. Consider, for example, causal relevance and centrality. When reacquiring explanatory knowledge (in contrast to mere lists of facts), a student who has mastered the material once before is likely to have a lingering sense of what matters. This holds for a general sense of what kinds of causal patterns and elements play a role in the explanation as well as for a more refined sense of what are the most central elements and relations. A student relearning how a refrigerator works will immediately know to focus on the flow of the gas and to pay attention to where and how it transfers heat. That same student will be better equipped to construct more meaningful summary gists of complex and perhaps excessively detailed explanations and may also be better attuned to helpful analogies that highlight key relations. Finally, students relearning a topic should have a better sense of when and when not to defer to an instructor or to passages in a text. They should be

206  Frank Keil better able to take a more critical eye toward the content. Departures from core, well-​grounded causal powers and elements should be more detectable.

7.  Conclusions All of us, from scientists to young children, have partial understandings and usually fail to realize just how partial those understandings are. Yet those gaps may be highly adaptive, especially when one considers the early developing tools all humans have for coping with causal content. It simply is not possible for any one mind to store all the details necessary to completely understand how something works or why it occurs as it does, and often those details are not known to anyone. Our partial understandings gain their power through a variety of heuristics that enable us to use what we do know to appropriately defer and lock onto knowledge in other minds. Early exposure to mechanisms may provide an essential way for building up more abstract representations that help us navigate the division of cognitive labor even as mechanistic details fade from memory. Even the illusions of understanding may be useful, for they may stop us from trying to store details we don’t really need and may provide an appropriate sense of knowing through other minds.

References Ahl, R. E., & Keil, F. C. (2017). Diverse effects, complex causes: Children use information about machines’ functional diversity to infer internal complexity. Child Development, 88(3), 828–​845. Alter, A. L., Oppenheimer, D. M., & Zemla, J. C. (2010). Missing the trees for the forest: A construal level account of the illusion of explanatory depth. Journal of Personality and Social Psychology, 99(3), 436–​451. Boyd, R. (2010). Homeostasis, higher taxa, and monophyly. Philosophy of Science, 77(5), 686–​701. Callanan, M. A., & Oakes, L. M. (1992). Preschoolers’ questions and parents’ explanations:  Causal thinking in everyday activity. Cognitive Development, 7(2), 213–​233. Chouinard, M. M. (2007). Children’s questions: A mechanism for cognitive development. Monographs of the Society for Research in Child Development, 72(1), vii–​ix, 1–​112; discussion 113–​126. Corriveau, K. H., & Kurkul, K. E. (2014). “Why does rain fall?”: Children prefer to learn from an informant who uses noncircular explanations. Child Development, 85(5), 1827–​1835.

How Do Partial Understandings Work?  207 Craver, C., & Bechtel, W. (2006). Mechanism. In J. Pfeifer & S. Sarkar (eds.), The Philosophy of Science: An Encyclopedia. New York: Psychology Press, pp. 469–​478. Fernbach, P. M., Rogers, T., Fox, C. R., & Sloman, S. A. (2013). Political extremism is supported by an illusion of understanding. Psychological Science, 24(6), 939–​946. Fisher, M., & Keil, F. C. (2016). The curse of expertise: When more knowledge leads to miscalibrated explanatory insight. Cognitive Science, 40, 1251–​1269. Fisher, M., & Keil, F. C. (in preparation). Decay neglect: An illusion of knowledge persistence in students. Frazier, B. N., Gelman, S. A., & Wellman, H. M. (2009). Preschoolers’ search for explanatory information within adult-​child conversation. Child Development, 80 (6), 1592–​1611. Frazier, B. N., Gelman, S. A., & Wellman, H. M. (2016). Young children prefer and remember satisfying explanations. Journal of Cognition and Development, 17(5), 718–​736. Gao, T., Newman, G. E., & Scholl, B. J. (2009). The psychophysics of chasing: A case study in the perception of animacy. Cognitive Psychology, 59(2), 154–​179. Hegarty, M. (2004). Mechanical reasoning by mental simulation. Trends in Cognitive Sciences, 8(6), 280–​285. Keil, F. C., Smith, W. C., Simons, D. J., & Levin, D. T. (1998). Two dogmas of conceptual empiricism: Implications for hybrid models of the structure of knowledge. Cognition, 65(2), 103–​135. Keil, F. C., Stein, C., Webb, L., Billings, V. D., & Rozenblit, L. (2008). Discerning the division of cognitive labor: An emerging understanding of how knowledge is clustered in other minds. Cognitive Science, 32, 259–​300. Kim, N. S., Luhmann, C. C., Pierce, M. L., & Ryan, M. M. (2009). Causal cycles in categorization. Memory and Cognition, 37, 744–​758. Kiyomitsu T. (2016) Analyzing spindle positioning dynamics in cultured cells. In P. Chang and R. Ohi (eds.), The Mitotic Spindle. Methods in Molecular Biology, vol. 1413. New York: Humana Press, pp. 239–​252. Kominsky, J. F., Zamm, A. P., & Keil, F. C. (in press). Knowing when help is needed: A developing sense of causal complexity. Cognitive Science. Lawson, R. (2006). The science of cycology: Failures to understand how everyday objects work. Memory and Cognition, 34(8), 1667–​1675. Lehrer, R., & Schauble, L. (1998) Reasoning about structure and function:  Children’s conceptions of gears. Journal of Research in Science Teaching, 35(1), 3–​25. Lockhart, K. L., Goddu, M. K., Smith, E. D., & Keil, F. C. (2016). What could you really learn on your own? Understanding the epistemic limitations of knowledge acquisition. Child Development, 87(2), 477–​493. Lockhart, K. L., Kerr, S., & Keil, F. (2017) The privileged status of knowing mechanistic information: An early epistemic bias. Poster presented at 2017 SRCD meeting, Austin, TX. Mascaro, O., & Csibra, G. (2012). Representation of stable social dominance relations by human infants. Proceedings of the National Academy of Sciences, 109(18), 6862–​6867. McGrew, W. C. (2013). Is primate tool use special? Chimpanzee and New Caledonian crow compared. Philosophical Transactions of the Royal Society of London B: Biological Sciences, 368, 20120422. doi: http://​dx.doi.org/​10.1098/​rstb.2012.0422. Mills, C., & Keil, F. C. (2004). Knowing the limits of one’s understanding: The development of an awareness of an illusion of explanatory depth. Journal of Experimental Child Psychology, 87, 1–​32.

208  Frank Keil Newman, G. E., Keil, F. C., Kuhlmeier, V. A., & Wynn, K. (2010). Early understandings of the link between agents and order. Proceedings of the National Academy of Sciences, 107(40), 17140–​17145. Peffer, T., Pritoni, M., Meier, A., Aragon, C., & Perry, D. (2011). How people use thermostats in homes: A review. Building and Environment, 46(12), 2529–​2541. Rehder, B. (2017). Reasoning with causal cycles. Cognitive science, 41(S5), 944–​1002. Rochat, P., Striano, T., & Morgan, R. (2004). Who is doing what to whom? Young infants’ developing sense of social causality in animated displays. Perception, 33(3), 355–​369. Rottman, B. M., & Keil, F. C. (2011). What matters in scientific explanations: Effects of elaboration and content. Cognition, 121, 324–​37. Rottman, B. M., & Keil, F. C. (2012). Causal structure learning over time: Observations and interventions. Cognitive Psychology, 64(1), 93–​125. Rozenblit, L., & Keil, F. C. (2002). The misunderstood limits of folk science: An illusion of explanatory depth. Cognitive Science, 26(5), 521–​562. Schommer, M., & Surber, J. R. (1986). Comprehension-​monitoring failure in skilled adult readers. Journal of Educational Psychology, 78(5), 353. Simon, H. A. (1996). The Sciences of the Artificial. Cambridge, MA: MIT Press. Strickland, B., Silver, I., & Keil, F. C. (2017). The texture of causal construals: Domain specific biases shape causal inference from discourse. Memory and Cognition, 45, 442–​455. Waldmann, M. R., Hagmayer, Y., & Blaisdell, A. P. (2006). Beyond the information given: Causal models in learning and reasoning. Current Directions in Psychological Science, 15(6), 307–​311. Wolff, P. (2007). Representing causation. Journal of Experimental Psychology:  General, 136(1), 82.

11 Mechanistic versus Functional Understanding Tania Lombrozo and Daniel Wilkenfeld

A fawn is born with white spots. An alarm clock beeps unpleasantly. A professor decides to give a pop quiz. How can we understand why these events occurred as they did? Several psychologists have suggested that events like these support (at least) two forms of understanding—​what we will call mechanistic understanding and functional understanding. Mechanistic understanding relies on an appreciation of parts, processes, and proximate causal mechanisms. The fawn has white spots because of its genes and prenatal environment; the alarm clock beeps when the circuit connecting a power source to a buzzer is completed; the professor decides to give a pop quiz when she sees that her students have not been coming to class prepared. Functional understanding, by contrast, relies on an appreciation of functions, goals, and purpose. The fawn has white spots to hide from predators against sun-​flecked ground; the alarm clock beeps to wake its sleeping owner; the professor gives a pop quiz to assess and improve her students’ mastery of the course material. The distinction between mechanistic and functional understanding rests on substantive (if typically implicit) assumptions about what “understanding” amounts to, and about how understanding can be carved up into distinct forms. Our aim in this paper is to evaluate the evidence for mechanistic and functional forms of understanding through the lens of contemporary epistemology and philosophy of science, which offer valuable new tools for thinking about the nature and varieties of understanding. In particular, we evaluate two claims: the weak differentiation thesis, according to which mechanistic and functional understanding have importantly different objects, and the strong differentiation thesis, according to which mechanistic and functional understanding constitute qualitatively different kinds of understanding.

210  Tania Lombrozo and Daniel Wilkenfeld In Section 1, we briefly introduce a family of related accounts of understanding that have emerged from recent work in philosophy. While we don’t commit to a specific member of this family, we take on a shared commitment that guides our subsequent discussion:  that understanding is at least partially a matter of representing the right kinds of (explanatory) dependence relationships. In Section 2, we introduce the idea of “stances” or “modes of construal” as cognitive mechanisms that support the construction of mental representations that underwrite this notion of understanding. In Section 3, we review the empirical evidence for the psychological reality of mechanistic and functional modes of construal. In Section 4, we argue for the weak differentiation thesis. Finally, in Section 5, we offer more tentative arguments for the strong differentiation thesis.

1.  Understanding as Representing (Explanatory) Dependence Accounts of understanding within epistemology and philosophy of science differ along a variety of dimensions, including whether understanding is regarded as a type of knowledge (e.g., Grimm 2006), as an ability (e.g., Hills 2016), as possession of a mental model (e.g., Knuuttila and Merz 2009), or as some other form of epistemic relation (Wilkenfeld forthcoming). However, virtually all extant accounts share one thing in common:  because understanding is regarded as a fundamentally cognitive or epistemological relationship, it must be constituted at least in part by how we represent that which is understood.1 Typically, the contents of these representations are taken to be of particular kinds of dependence relations, be they causal, explanatory, or counterfactual. The most straightforward view of the representational content of under­ standing might be that understanding why P corresponds to having knowl­ edge of the causes of P.  This is the picture one gets from the influential 1 To our knowledge, the only philosophers who would take issue with this characterization are those who would eschew a representationalist framework on other grounds (e.g., Price 2011). Even on Hills’s (2016) ability-​centric account, part of understanding why is being able to put forward an explanation, which at least prima facie seems to require representing the understood as occupying a specific node in an explanatory nexus. Since our focus will be on how the representations that make up understanding are effected, we will concentrate mostly on the representational component—​ however, everything we say about generating such representations holds true whether they exhaust understanding or only complement some other property of the understander (e.g., an ability).

Mechanistic versus Functional Understanding  211 work of Woodward (2003), who has a well-​worked-​out account of the nature of causal information, causal understanding, and causal explanation. Alternatively, one could think that causation is just one species of a more general category of metaphysical dependence relation, knowledge of any of which might constitute understanding (Kim 1994). For instance, non-​causal dependence relations could include mereological or conceptual relations. Another line of thought is that the category of “dependence relations” might be too narrow (or at least not narrow in the correct way) to capture understanding (perhaps because it doesn’t capture explanatory connections between necessary truths), and that really the object of the knowledge that constitutes understanding is better thought of as explanation generally. The idea is that the representational content of understanding simply is knowl­ edge of an explanation. While this view has periodically arisen alongside theories of explanation (most famously Hempel 1965), it has not generally been favored by epistemologists of understanding itself.2 The best way to appreciate the source of skepticism regarding purely knowledge-​based accounts of understanding is to see what philosophers add to such accounts—​frequently a deeper epistemic relation often known as “grasping.” One example is what Strevens (2013) refers to as the “simple view,” which is the view that understanding is the state one is in precisely when one grasps a correct scientific explanation. What the grasping adds to knowledge is that it rules out what we might think of as inert knowledge—​ propositions that could be known without one being able to really see their inferential or practical implications. While these accounts are distinct, and their advocates propose their own variants, they share a core commitment that will guide our subsequent discussion of the empirical literature. Specifically, these views share the commitment that understanding is at least partially constituted by mental representations that encode the right kinds of dependence relations, where the right kinds of dependence relations are those that are causal, explanatory, or otherwise privileged, perhaps in terms of their functional or inferential roles. In the section that follows, we’ll suggest that “stances” or “modes of construal” support the creation of precisely these kinds of mental representations. 2 There are exceptions—​for example, Trout (2007) argues that understanding is redundant with explanation, because the only sort of worthwhile understanding is knowledge of an explanation. (Trout sometimes uses the language of “grasping,” but he also suggests [585] that it is really knowl­ edge that he has in mind.) For an extended reply to Trout, see de Regt 2009.

212  Tania Lombrozo and Daniel Wilkenfeld

2.  Stances or “Modes of Construal” as Paths to Understanding In several influential papers and books, Daniel Dennett introduced the idea of a stance: a strategy for interpreting the behavior of an entity (e.g., Dennett 1971; 1989; 2009). Most relevant for our purposes, Dennett differentiated between a physical stance, which involves predicting and explaining the entity through the application of physical laws, and a design stance, which involves predicting and explaining the entity on the basis of its design and proper functioning. For example, someone who predicts what will happen when pressing a button on an alarm clock by considering the underlying electronic components and the physical laws that govern them is applying the physical stance; someone who does so by thinking about how the alarm clock would be designed and assuming that it is functioning properly is applying the design stance. An idea akin to Dennett’s stances was introduced into the psychological literature by Frank Keil, who argued that even young children are equipped with multiple “modes of construal” that “frame” explanations by positing certain kinds of relations, properties, or arguments as central (Keil 2006). Like Dennett, Keil argued that these include a mechanical/​physical mode of construal and a teleological/​functional mode of construal. Keil argued for these modes of construal on the basis of children’s patterns of explanations and predictions across domains (Keil 1994; 1995). In particular, he distinguished between what we will call “mechanistic” explanations (involving parts or proximate causal mechanisms) and “functional” explanations (involving functions, purpose, or goals), where the former reflects the operation of a mechanistic mode of construal, and the latter the operation of a functional mode of construal. Importantly, modes of construal (or stances; we will use these terms interchangeably) are not themselves domain theories, such as (scientific or intuitive) physics, or (scientific or intuitive) psychology. Nonetheless, these domain theories may be prerequisites to the successful application of a mode of construal: it is these theories that supply the laws required to apply a physical stance, and that constrain inferences about what would constitute good design and proper functioning. Modes of construal, unlike domain theories themselves, provide a template or algorithm of sorts, determining the basis for a prediction or explanation, and accordingly constraining which domain theories will be consulted and how.

Mechanistic versus Functional Understanding  213 If modes of construal are strategies for interpreting entities and their behavior, they do not themselves constitute understanding. Nonetheless, we think there are two meaningful ways in which we might say that stances support or reflect understanding, corresponding to the output versus the input to the corresponding mode of construal. First, the representations that result from the application of a stance to a particular entity will include the representational bases for prediction and explanation—​typically the identification of causal and explanatory relationships that hold (or are believed to hold) for the case in question. For example, applying the physical stance to an alarm clock will involve representing the components of the alarm clock as instantiating more general causal relationships encoded in an intuitive physical theory. Applying the design stance to a fawn’s spots could involve drawing inferences about the function the spots might serve in a particular ecological context. In this way, the application of a stance will include the creation of representations that can constitute at least the representational component of understanding.3 They can constitute understanding because they satisfy the common requirements for understanding that we identified in Section 1: representing the right kinds of dependence relations. A second way in which modes of construal might relate to understanding is in the way they pick out aspects of intuitive theories. First, note that an intuitive theory could itself constitute some form of understanding. On most accounts, intuitive theories are defined in terms of the explanatory generalizations and causal relationships that they represent (e.g., Carey 1985; Gopnik, Maltzoff, and Bryant 1997; Gopnik and Wellman 2012; Murphy and Medin 1985), again providing a good match to the accounts of understanding that we identify in Section 1. On this view, the components of a theory that in a specific context are employed in applying a mechanistic mode of construal might be said to constitute “mechanistic” understanding (in that context), while those employed in applying a functional mode in a specific context might be said to constitute “functional” understanding (in that context). To illustrate these two ways of relating modes of construal to understanding, consider again our spotted fawn. Someone who has an intuitive theory of biology that includes resources for explaining biological adaptations might possess some “functional understanding” of biological adaptations in general. When applying this to the spotted fawn, she comes 3 For the remainder of the paper we will talk in terms of constituting understanding, leaving implied the caveat that there might be other necessary conditions on some accounts.

214  Tania Lombrozo and Daniel Wilkenfeld to possess some functional understanding of why fawns have spots. It is this latter form of understanding—​the understanding that results from the application of a mode of construal to a particular entity—​that the empirical evidence has most closely addressed, and that we turn to in Section 3.

3.  The Empirical Evidence for Mechanistic and Functional Modes of Construal The majority of research on mechanistic and functional modes of construal has focused on mechanistic and functional explanations, with the (often implicit) assumption that the endorsement or generation of each explanation type reflects the operation of its corresponding mode of construal. Accordingly, our review will focus on what we know about these two kinds of explanations. First, we begin with an important similarity: both mechanistic and functional explanations are understood as causal explanations. For mechanistic explanations this is an uncontroversial claim; they explicitly appeal to proximate causes and causal mechanisms. However, this isn’t self-​evidently the case for functional explanations—​after all, functional explanations explain current properties or events by appeal to potential future consequences, and thus seem to get the causal order wrong. When we explain that the teacher gave a pop quiz “to teach her students a lesson,” we seem to be explaining a current action by appeal to an anticipated but unrealized effect of that action. When we explain the fawn’s spots by appeal to camouflage, we seem to be explaining a current property by its potential future influence on predators. Several accounts of functional explanation offer ways to understanding future-​looking functional explanations in more standard backward-​looking causal terms (e.g., Allen 2009; Wright 1976). When we explain the teacher’s pop quiz by appeal to the goal of teaching the students a lesson, for example, we can take this as a shorthand for the beliefs and desires that were in fact the proximate causes of her action. She wanted to teach her students a lesson, and she believed that administering a pop quiz would accomplish her goal. These mental states preceded her behavior, and we can understand the functional explanation as a pointer to these antecedent causes. A  more complicated but similar move works for adaptationist explanations. When we explain the fawn’s spots by appeal to camouflage, this functional explanation

Mechanistic versus Functional Understanding  215 is underwritten by a particular set of causal commitments: that spots in fact support camouflage, and that the fact that they do so played a causal role in the maintenance and spread of spots in past fawns that causally led to the existence of the current spotted fawn. Wright (1976) provides a more gen­ eral formulation of the causal commitments that underwrite functional explanations in terms of what he calls a “consequence etiology.” Critically, empirical evidence supports this analysis of functional explanations as a descriptively adequate account of human cognition. Lombrozo and Carey (2006) presented adult participants with vignettes followed by why-​questions and candidate explanations, with the aim of identifying the conditions under which participants would find functional explanations acceptable. They found that Wright’s causal commitments were a necessary condition for acceptance, where the relevant causal commitments were both manipulated experimentally and assessed by having participants indicate their agreement with counterfactual claims. Roughly, for some prop­erty P to be explained by appeal to some function F, participants had to endorse the claim that had P not resulted in F, the entity with P probably wouldn’t have had P. Additional work supports the idea that functional explanations are tied to particular causal commitments. Kelemen and DiYanni (2005) found that children were more likely to accept a functional explanation for the origins of an entity or event (e.g., “the first ever thunderstorm occurred to give the earth water so everything would grow”) if they also believed that the entity or event was created by “someone or something.” Adults are also more likely to accept scientifically unwarranted teleological explanations (e.g., “water condenses to moisten the air”) if they endorse some Gaia-​like causal force (Kelemen and Rosset 2009; Kelemen, Rottman, and Seston 2013; see also ojalehto, Waxman, and Medin 2013). The evidence thus suggests that while functional explanations may differ from mechanistic explanations (as we’ll see subsequently), they should not be understood as non-​causal. If mechanistic and functional explanations are both causal explanations, we can already see why knowing or grasping them might constitute understanding on the sort of view sketched in Section 1. The next question, then, is how they differ from each other. One differentiating factor has already emerged: whereas mechanistic explanations invoke proximate causal processes directly, functional explanations do so indirectly; they don’t wear their causal commitments on their sleeves. But the literature provides two additional bases for differentiation that are worth reviewing in turn: functional

216  Tania Lombrozo and Daniel Wilkenfeld explanations are to some extent mechanism-​independent, and they have a distinct developmental and cognitive profile. First, consider the claim that functional explanations are mechanism-​ independent in the sense that they highlight dependence relations that can be multiply realized, and that their explanatory value is enhanced, rather than diminished, by the dissociation from particular mechanisms. The intuition behind these claims is nicely illustrated by William James’s description of the relationship between Romeo and Juliet (an intentional system) versus iron filings and a magnet (a physical system): Romeo wants Juliet as the filings want the magnet. And if no obstacles intervene, he moves toward her by as straight a line as they. But Romeo and Juliet, if a wall be built between them, do not remain idiotically pressing their faces against its opposite sides, as in fact the iron filings do, pursuing the magnet. Romeo soon finds a circuitous way, by scaling the wall or otherwise, of touching Juliet’s lips directly. With the filings the path is fixed; whether it reaches the end depends on accidents. With the lover it is the end which is fixed, the path may be modified indefinitely. (James 1890: 20)

Romeo, unlike the iron filings, will find an alternative way to reach Juliet. He’ll climb the wall; he’ll dig a tunnel. The relationship that’s stable is that between Romeo’s goal of reaching Juliet and his eventual arrival at her side; the means by which he accomplishes this might be variable and highly contingent on idiosyncratic features of the way things happened to unfold. It’s this sense in which reasoning about Romeo and Juliet in terms of functional relationships is mechanism (or means) independent. Correspondingly, we can explain Romeo’s actions with a functional explanation (“he went that way to reach Juliet”), and this might strike us as more appropriate than a mechanistic explanation (“he went that way because he moved his muscles in such and such a way”) precisely because it identifies the dependence relation that’s robust across irrelevant perturbations (see also Murray and Lombrozo 2017; Vasilyeva, Blanchard, and Lombrozo, 2018). Consider an example from Daniel Dennett, motivating the design stance: Suppose I categorize a novel object as an alarm clock: I can quickly reason that if I  depress a few buttons just so, then some hours later the alarm clock will make a loud noise. I don’t need to work out the specific physical laws that explain this marvelous regularity; I simply assume that it has a

Mechanistic versus Functional Understanding  217 particular design—​the design we call an alarm clock—​and that it will function properly, as designed. (Dennett 2009: 340)

In a case like this, the relationship between the buttons and the noises can be multiply realized; it’s the function or design of the clock that constrains their relationships. We don’t need to reason on the basis of physical laws and causal mechanisms because the explanatory and predictive relationships that we care about are mechanism independent. “The essential feature of the design stance,” Dennett writes, “is that we make predictions solely from knowl­ edge or assumptions about the system’s functional design, irrespective of the physical constitution or condition of the innards of the particular object” (Dennett 1971: 88). This is part of what makes the design stance so powerful: we can achieve some predictive and explanatory competence without detailed knowledge of general mechanisms or detailed knowledge of how particular causal processes unfolded in the past. Psychological evidence supports the idea that while functional explanations are understood as causal explanations, they are (at least somewhat) mechanism independent. One source of evidence comes from studies that have examined people’s patterns of generalization, where they could generalize on the basis of proximate mechanisms or on the basis of functions and design (Ahn 1998; Lombrozo 2009; Lombrozo and Gwynne 2014; see also Lombrozo and Rehder 2012). When participants were given or generated functional explanations, they were significantly less likely to generalize on the basis of proximate mechanisms (relative to functions). To illustrate, consider a study from Lombrozo and Gwynne (2014). In this study, participants learned about animals and artifacts, where each had a target property that could be explained either mechanistically or functionally. For example, some participants read about a plant called a narp with a speckled pattern. They learned that “biologists have discovered that in narps, the speckled pattern is caused by the XP2 gene.” This supported the mechanistic explanation that narps have a speckled pattern because of the gene. They also learned that “having a speckled pattern attracts butterflies, which play a role in pollination.” This supported the functional explanation that narps have a speckled pattern to attract butterflies for pollination. Participants were then asked to explain, in a sentence, why narps have a speckled pattern. This prompt was deliberately ambiguous:  it could be answered by providing a mechanistic explanation, a functional explanation, or both.

218  Tania Lombrozo and Daniel Wilkenfeld After responding to the ambiguous prompt, participants learned about novel items that shared either the proximate cause (e.g., another plant with the XP2 gene) or the function (e.g., another plant that attracts butterflies for pollination), and they were asked whether they would generalize properties from the initial item (the narp) to these new cases. For example, if participants were told that the speckled pattern on narps is high in contrast, would they be inclined to think that the speckled pattern of the other plant with the XP2 gene, or the other plant that attracts butterflies, was also high in contrast? A key finding was that for biological organisms, those participants who provided a functional explanation in response to the ambiguous prompt were less likely than those who did not do so to generalize on the basis of underlying causal mechanisms. Instead, for all types of items, participants who provided a functional explanation were more likely than those who did not do so to generalize on the basis of shared functions. A second source of evidence for the idea that some level of mechanism independence can be induced by a functional mode of construal comes from people’s causal ascriptions. Lombrozo (2010) presented participants with vignettes in which three causal factors interacted to bring about an effect. For example, in one vignette, participants read about a type of shrimp that eats three food sources, call them A, B, and C. They further learned that eating these three food sources results in the shrimp reflecting high frequencies of UV light. The causal relationships between these factors were designed to create a situation involving “double prevention”: A could cause the shrimp to reflect high frequencies of UV light on its own, if not prevented by B. But C prevented B from preventing A, thereby resulting in the effect. This causal structure was used to isolate a notion of causation based on counterfactual dependence from one based on a physical mechanism involving what philosophers often call production or transmission (e.g., Hall 2004). Specifically, while C “caused” the effect in the sense that the effect would not have occurred in its absence, C did not produce the effect through some spatio-​temporally continuous mechanism or direct transmission of force. The key experimental manipulation was whether participants were given additional information that would allow them to construe the relationship between C and the effect functionally. Half the participants were told that the effect (reflecting high frequencies of UV light) serves a biological function (temperature regulation), and that the shrimp evolved to eat A and C for this reason. The key finding was that participants were significantly more inclined to consider C a cause of the effect when this functional relationship

Mechanistic versus Functional Understanding  219 held, such that the difference in ratings between A (the productive cause) and C (the dependence cause) was decreased. This suggests that when construing a relationship functionally, participants’ judgments of whether some factor caused an effect were less sensitive to the nature of the mechanism mediating the counterfactual dependence between the effect and the candidate cause. A third source of evidence that functional thinking induces some insensitivity to mechanistic information comes from the illusion of explanatory depth, or IOED (Rozenblit and Keil 2002). The basic finding is that people tend to overestimate their mechanistic understanding of how devices such as helicopters or flush toilets work. One suggestion is that people mistake a functional understanding for how something was designed or operates for a mechanistic understanding of the actual causal processes involved. Consistent with this idea, Alter, Oppenheimer, and Zemla (2010) found that when participants adopted a more abstract mode of construal, which is itself associated with reasoning in terms of functions, they experienced a larger IOED. When reasoning functionally, it seems, they had less metacognitive access to their deficient mechanistic understanding. A fourth source of evidence for a relationship between a functional mode of construal and mechanism independence comes from looking-​time studies with infants. Woodward and her colleagues have shown that when infants construe an agent’s action as a goal-​directed reach, they are more likely to expect that the agent’s next action will preserve the same goal, even if it involves a departure in means, such as reaching left versus right (e.g., Cannon and Woodward 2012; Woodward 1998). Research by Gergely and colleagues illustrates that infants can also use variation in means as a basis for inferring that an agent is rational in its pursuit of goals: when an agent’s goal is preserved despite variation in means, 12-​month-​olds develop expectations that the agent will seek the goal, and will do so in the most rational (i.e., spatially efficient) way possible (e.g., Gergely et al. 1995). To sum up, these studies on generalization, causal ascription, metacognition, and infants’ perception of goal-​directed action support the idea that adopting a functional construal differs from a mechanistic construal in that the former allows for a more mechanism-​independent form of reasoning. More recent work by Liquin and Lombrozo (2018) sheds further light on why this might be. They find that when evaluating a functional explanation, judgments are largely (though not exclusively) driven by an assessment of how well a proposed feature (such as “reflecting high frequencies of UV light”) “fits” a given function (such as “thermal regulation”). This evaluation

220  Tania Lombrozo and Daniel Wilkenfeld of structure-​function fit may involve some mechanistic reasoning, but it crucially does not depend upon a detailed analysis of the feature’s etiology. Indeed, Liquin and Lombrozo find that when an explanation contains functional information, participants become less sensitive to etiological detail. A second factor that differentiates mechanistic and functional construals may or may not be related: there’s evidence that functional explanations may be psychologically privileged in the sense that they are often favored and seem to be less cognitively demanding. In particular, there’s evidence that children use them “promiscuously” (Kelemen 1999), and that adults will accept scientifically unwarranted functional explanations when cognitively impaired (Lombrozo, Kelemen, and Zaitchik 2007), when responding under speeded conditions (Kelemen and Rosset 2009; Kelemen, Rottman, and Seston 2013), or when they engage in less reflective thought (Steiner, Zemla, and Sloman 2016). Kelemen and colleagues have argued that a teleological mode may be a “cognitive default” that emerges early in development and remains throughout the lifespan, re-​emerging when alternative cognitive resources are taxed (Kelemen, Rottman, and Seston 2013; see also Shtulman and Lombrozo 2016). Liquin and Lombrozo (2018) argue that this is because structure-​function fit serves as an intuitive but defeasible cue to the acceptability and quality of a functional explanation (see also Lombrozo, Kelemen, and Zaitchik 2007). The evidence reviewed so far suggests that mechanistic explanations differ from functional explanations in their causal commitments, that mechanistic and functional explanations support different patterns of generalization and causal ascription, and that functional explanations may be cognitively privileged in some sense. These forms of differentiation suggest that mechanistic and functional modes of construal could be tuned to different inferential functions. Rather than adopting a mechanistic mode of construal for some kinds of objects and a functional mode of construal for others, people could flexibly adopt one mode or the other depending on the entity in question and their inferential aims. Indeed, there’s some evidence that this is the case. Not only do people spontaneously offer one kind of explanation or another in response to various features of the entity in question (e.g., Lombrozo and Carey 2006; Sanchez et al. 2016), they also adapt their evaluations to their inferential goals: they rate mechanistic explanations more highly when they anticipate making inferences on the basis of proximate mechanisms, and they rate functional explanations more highly when they anticipate making inferences on the basis of function (Vasilyeva, Wilkenfeld, and Lombrozo 2017).

Mechanistic versus Functional Understanding  221 In sum, there is good evidence for the claim that mechanistic and functional explanations are psychologically distinct. They not only differ in their causal commitments, but also in the extent to which they demand and depend upon an articulation or specification of mechanisms or particular causal processes. They also differ in the dependence relations that they privilege for the purpose of generalization. Perhaps for these reasons, functional explanations seem to have a special role in development and may be less cognitively demanding. On the view of understanding articulated in Section 1, representations of mechanistic and functional explanations are good candidates for understanding:  they encode causal and explanatory relationships that plausibly support an understanding of why some entity has a particular property or exhibited a particular behavior. But do they merely support understanding of different things? Or do they constitute different kinds of understanding? We turn to the weak and strong differentiation theses in Sections 4 and 5, respectively.

4.  The Case for Weak Differentiation In Section 1, we suggested that understanding involves some representation of dependence or explanatory relations. In Sections 2 and 3, we suggested that mechanistic and functional modes of construal support mechanistic and functional understanding, respectively. In this section, we consider whether this evidence supports the weak differentiation thesis, namely that mechanistic and functional understanding are different insofar as they involve different objects (whether or not they also involve different epistemic relations to those objects). We suggest that mechanistic and functional understanding indeed involve different content, support different functions, and have a distinctive phenomenology. However, we will also argue that each of these claims is insufficient to support the strong differentiation thesis that these constitute different kinds of understanding. The claim that mechanistic and functional understanding involve different content follows straightforwardly from the data presented in Section 3. We’ve seen that they involve mechanistic versus functional explanations, privilege production versus dependence notions of causation, and privilege different dependence relations as a basis for inference. Yet there are good reasons to doubt that understanding should simply inherit criteria for

222  Tania Lombrozo and Daniel Wilkenfeld individuation from explanation, causation, or some inferential role. If understanding is a relation between mind and world, it might be the same relation even when the world provides starkly different relata. As an analogy, knowl­ edge of mechanisms and knowledge of functions take very different objects, but we would not for that reason usually be inclined to say that they manifest more than one knowledge relation. There might be some objects so diverse that we have reason to posit multiple knowledge relations, but mechanisms and functions can still be known in approximately the same way. Mechanistic and functional understanding also differ with respect to their core functions. While both support prediction and explanation, mechanistic understanding is particularly useful for prediction and explanation in some domains, while functional understanding is more useful in others. Moreover, as shown in Vasilyeva, Wilkenfeld, and Lombrozo (2017), people privilege the explanations that support their current inferential goals. But once again, it’s not clear that supporting different kinds of inferences underwrites the stronger claim that an understanding of whatever-​supports-​mechanistic-​ inferences and an understanding of whatever-​ supports-​ functional-​ inferences are different kinds of understanding. Knowledge of statistics supports inductive inferences, whereas knowledge of geometry supports deductive inferences, but we would not on that basis typically be inclined to consider them different kinds of knowledge. Finally, consider the claim that mechanistic and functional understanding are distinct with regard to their phenomenology. This claim goes admittedly beyond the data, but it’s only a modest step from the claim that functional explanations are a cognitive default of some kind (a claim that may or may not be right) to the claim that they are satisfying in a more basic or intuitive way. However, there is reason to doubt that when two tokens of understanding feel different to their respective understanders, we have good grounds for saying that they belong to two different kinds of understanding. Knowledge that one is in danger might feel quite different from knowledge that one is safe; it doesn’t follow that the knowledge relation is different in kind. If differences in content, function, and phenomenology are insufficient to support the claim that mechanistic and functional understanding are different kinds of understanding, it might be tempting to reject their uniqueness entirely, and to instead consider the possibility that mechanistic and functional understanding are but two among a very large number of possible targets for understanding. On this licentious view, any strategy for privileging a subset of the enormously complex (explanatory) dependence relations

Mechanistic versus Functional Understanding  223 in the world offers a “mode of construal” and thus a possible target for understanding. Moreover, mechanistic and functional understanding have no special status with respect to these alternatives, and all of these alternatives support understanding in just the same way: by supporting representations of the dependence relations that constitute understanding. We think this possibility misses something important. It’s not a coincidence that mechanistic and functional stances or construals arise again and again in philosophy and in psychology, across disciplines and over time. These two construals—​ unlike an arbitrary subset of dependence relations—​seem to capture something important about the structure of the world and our goals within it. Proximate causes and goals, under the right circumstances, identify dependence relations that are particularly stable, or insensitive to perturbations in background conditions (Lombrozo 2010; see also Blanchard, Vasilyeva, and Lombrozo 2018; Woodward 2006). Given our goals, they might be particularly useful bases for prediction and intervention. For these reasons, it seems appropriate to recognize mechanistic and functional understanding as understanding of special kinds of targets, even if the understanding itself is not different in kind.4 It’s for this reason that we favor some form of differentiation between mechanistic and functional understanding, even if it’s only a weak form. In sum, we think there is good evidence for the weak differentiation thesis: mechanistic and functional understanding have objects that are both important and importantly different from each other. At the same time, we don’t think that the evidence just reviewed supports the stronger claim that mechanistic and functional understanding involve qualitatively different kinds of understanding. In Section 5, we evaluate this stronger claim.

5.  The Case for Strong Differentiation In this section, we consider two tentative arguments for the strong differentiation thesis: the claim that mechanistic and functional understanding reflect different epistemic relationships to the world. The two arguments that 4 We don’t mean to suggest here that functional and mechanistic understanding are the only targets of understanding that might have this special status. There’s already good evidence for the psychological reality of something like intentional (e.g., Bertram and Hodges 2005) and formal (e.g., Prasada 2016) modes of construal, which could either be special targets of understanding or perhaps constitute different kinds of understanding.

224  Tania Lombrozo and Daniel Wilkenfeld we consider are that mechanistic and functional understanding differ in their normative entailments and that they differ in their modal implications. These arguments are tentative in part because they stem from intuitive considerations rather than fully developed theoretical arguments, and in part because they have empirical commitments that have yet to be tested. Nonetheless, we think these possibilities merit further study, and so we sketch them here. First, mechanistic and functional understanding seem to differ with regard to normative considerations. When we learn that an alarm clock has the function of waking its owner, we’re in a position to evaluate whether it has done so well. When we learn that a fawn has spots for camouflage, we can evaluate how it might better accomplish this goal. A mere causal process, on the other hand, does not support normative evaluations, at least not on its own. Causal processes—​absent any reference to goals—​are simply facts about the world, with no standard available against which to compare them. In David Hume’s Treatise of Human Nature (Book III), he famously puts forward the claim that you cannot derive claims about what ought to be from any number of premises about what is. We cannot even rightfully say that a causal process is effective without making assumptions regarding what counts as the relevant effect.5 Functional explanations also differ from mechanistic explanations in that the normative evaluations supported by the former involve an implicit perspective or point of view (see also ojalehto, Waxman, and Medin 2013, for the idea that functional explanations are importantly perspectival). When we evaluate the alarm clock, we do so from the perspective of the designer or the user. When we evaluate the fawn’s spots, we do so from the perspective of the fawn. (Presumably, effective camouflage is not a desirable characteristic for a fawn from the perspective of a mountain lion.) Again, causal processes are simply features of the world—​they do not, on their own, offer a perspective or point of view for further evaluation. If functional understanding involves normativity and an implicit perspective, but mechanistic understanding does not, we have the first hints that the relationship between the mind and the world may differ across these cases of understanding. Whereas mechanistic understanding involves a 5 We might be able to evaluate which of two processes is more “efficient” in the purely technical sense of “using less energy to accomplish the same task.” However, to see that this still has no normative implications, we need only imagine a goal for which the less efficient process is better—​for example, in using up a department’s budget so that it is not slashed for the following fiscal year.

Mechanistic versus Functional Understanding  225 mind-​to-​world fit (like a belief), functional understanding additionally has elements of a world-​to-​mind fit (like a desire). Functional understanding involves a perspective from which one can appreciate how the world would be more desirable (from that perspective). This aspect of functional understanding parallels an aspect of first-​personal understanding as developed by Grimm (2016). Grimm’s aim is to articulate how understanding the action of another person differs from understanding the structure of the natural world. Grimm argues that in the former case, it’s not enough to identify some of the (causal) structure upon which an action depends; it’s also important to understand why the action was desirable, or the choice choiceworthy. He writes that “this sort of seeing [of a goal as desirable or choiceworthy] plausibly requires a different cognitive attitude—​ and hence, apparently, a different cognitive method—​that we need to draw upon when we try to understand other human beings” (Grimm 2016: 217). This is part of what motivates Grimm’s conclusion that understanding another person is different from third-​personal understanding, which depends only on grasping the right kind of dependence structure (e.g., between a desire and some action), without the further step of understanding not only why it is desired, but why it is desirable. Similarly, functional understanding plausibly involves different “cognitive methods” from mechanistic understanding, though in the case of functional understanding, it suffices to understand what is desirable from the implicit perspective of a functional explanation, without it being desirable in Grimm’s first-​personal sense. The second way in which mechanistic and functional understanding could differ is with respect to the specificity of their commitments regarding the causal structure of the world. As we argued in Section 3, functional understanding is—​in an important sense—​mechanism independent. When we obtain functional understanding, our causal commitments radically underdetermine the actual causal process by which some property or event came to be. We can functionally explain why the alarm clock beeped by appealing to its design—​and be satisfied with our explanation—​even if we remain forever ignorant of whether its inner parts function electronically or pneumatically. For functional explanations, a “how possibly” story goes a long way; it might be enough to know that some process with a consequence etiology (Wright 1976) was at work, without knowing more about what it was or how it manifested. The same can’t be said for mechanistic understanding. We might be satisfied by a vague appeal to the alarm clock’s internal electronic processes, but a mechanistic explanation seems to demand a “how actually”

226  Tania Lombrozo and Daniel Wilkenfeld story. If this is right, then mechanistic and functional understanding are qualitatively distinct in the sense that they are differentially demanding with respect to what the actual causal structure of the world must be like. In sum, we’ve sketched two arguments for the strong differentiation thesis. We’ve suggested that mechanistic and functional understanding involve different epistemic relationships in that the latter has normative and perspectival elements that introduce a world-​to-​mind fit. We’ve also argued that because functional understanding is mechanism-​independent, it makes weaker demands on the causal structure of the world—​possibility is enough for understanding. These claims go beyond the weak differentiation thesis because they posit that mechanistic and functional understanding differ not only in terms of their objects, but in the mind-​world relation that they require.

6.  Conclusion Our aim in this paper has been to review empirical evidence for mechanistic and functional modes of construal, and to relate this evidence to accounts of understanding. First, we argued that these modes of construal support understanding because they play a role in generating the kinds of representations that (at least partially) constitute understanding. Next, we argued that mechanistic and functional understanding are distinct in two ways:  they involve importantly different objects, and (more tentatively) they involve different epistemic relationships. These claims have implications for how to think about understanding in epistemology and philosophy of science. They also invite us to ask a host of empirical questions about the psychological capacities that underwrite these forms of understanding, and about their implications for our interactions with the world.

Acknowledgments The authors gratefully acknowledge the John Templeton Foundation for its support through the Varieties of Understanding project. We are also grateful to Stephen Grimm for relevant discussions and helpful comments on an earlier draft.

Mechanistic versus Functional Understanding  227

References Ahn, Woo-​kyoung. 1998. “Why Are Different Features Central for Natural Kinds and Artifacts? The Role of Causal Status in Determining Feature Centrality.” Cognition 69(2): 135–​178. Allen, Colin. “Teleological Notions in Biology.” In The Stanford Encyclopedia of Philosophy (Winter 2009 edition), Ed. Edward N. Zalta, URL  =  https://​plato.stanford.edu/​ archives/​win2009/​entries/​teleology-​biology/​. Alter, Adam L., Oppenheimer, Daniel M., and Zemla, Jeffrey C. 2010. “Missing the Trees for the Forest: A Construal Level Account of the Illusion of Explanatory Depth.” Journal of Personality and Social Psychology 99(3): 436–​451. Blanchard, Thomas, Vasilyeva, Nadya, and Lombrozo, Tania. (2018). “Stability, Breadth and Guidance.” Philosophical Studies 175(9): 2263–​2283. Cannon, Erin N., and Woodward, Amanda L. 2012. “Infants Generate Goal‐Based Action Predictions.” Developmental Science 15(2): 292–​298. Carey, Susan. 1985. Conceptual Change in Childhood. Cambridge, MA: MIT Press. Dennett, Daniel C. 1971. “Intentional Systems.” Journal of Philosophy 68(4): 87–​106. Dennett, Daniel C. 1989. The Intentional Stance. Cambridge, MA: MIT Press. Dennett, Daniel C. 2009. “Intentional Systems Theory.” In The Oxford Handbook of Philosophy of Mind. Eds. Ansgar Beckermann, Brian P. McLaughlin, and Sven Walter. New York: Oxford University Press. De Regt, Henk W. 2009. “The Epistemic Value of Understanding.” Philosophy of Science 76(5): 585–​597. Gergely, György, Nádasdy, Zoltán, Csibra, Gergely, and Bíró, Szilvia. 1995. “Taking the Intentional Stance at 12 Months of Age.” Cognition 56(2): 165–​193. Gopnik, Alison, Meltzoff, Andrew N., and Bryant, Peter. 1997. Words, Thoughts, and Theories. Cambridge, MA: MIT Press. Gopnik, Alison, and Wellman, Henry M. 2012. “Reconstructing Constructivism: Causal Models, Bayesian Learning Mechanisms, and the Theory Theory.” Psychological Bulletin 138(6): 1085–​1108. Grimm, Stephen R. 2006. “Is Understanding a Species of Knowledge?” British Journal for the Philosophy of Science 57(3): 515–​535. Grimm, Stephen R. 2016. “How Understanding People Differs from Understanding the Natural World.” Philosophical Issues 26(1): 209–​225. Hall, Ned. 2004. “Two Concepts of Causation.” In Causation and Counterfactuals. Eds. John Collins, Ned Hall, and Laurie Paul. Cambridge, MA: MIT Press. Hempel, Carl G. 1965. Aspects of Scientific Explanation and Other Essays in the Philosophy of Science. New York: Free Press. Hills, Alison. 2016. Understanding Why. Noûs 50(4): 661–​688. Hume, David. [1738/​1740] 2003. A Treatise of Human Nature. Courier Corporation. James, W. 1890. The Principles of Psychology. Vol. 1. New York: Holt. Keil, Frank C. 1994. “The Birth and Nurturance of Concepts by Domains: The Origins of Concepts of Living Things.” In Mapping the Mind: Domain Specificity in Cognition and Culture. Eds. Lawrence A. Hirschfeld and Susan A. Gelman. Cambridge: Cambridge University Press. Keil, Frank C. 1995. “The Growth of Causal Understandings of Natural Kinds.” In Causal Cognition:  A Multidisciplinary Debate. Eds. Dan Sperber, David Premack, and Ann James Premack. New York: Oxford University Press.

228  Tania Lombrozo and Daniel Wilkenfeld Keil, Frank C. 2006. “Explanation and Understanding.” Annual Review of Psychology 57: 227–​254. Kelemen, Deborah. 1999. “Function, Goals and Intention:  Children’s Teleological Reasoning about Objects.” Trends in Cognitive Sciences 3(12): 461–​468. Kelemen, Deborah, and DiYanni, C. 2005. “Intuitions about Origins:  Purpose and Intelligent Design in Children’s Reasoning about Nature.” Journal of Cognition and Development 6(1), pp.3–​31. Kelemen, Deborah, and Rosset, Evelyn. 2009. “The Human Function Compunction:  Teleological Explanation in Adults.” Cognition 111(1): 138–​143. Kelemen, Deborah, Rottman, Joshua, and Seston, Rebecca. 2013. “Professional Physical Scientists Display Tenacious Teleological Tendencies: Purpose-​Based Reasoning as a Cognitive Default.” Journal of Experimental Psychology: General 142(4): 1074–​1083. Kim, Jaegwon. 1994. “Explanatory Knowledge and Metaphysical Dependence.” Philosophical Issues 5: 51–​69. Knuuttila, Tarja, and Martina Merz. 2009. “An Objectual Approach to Scientific Understanding:  The Case of Models.” In Scientific Understanding:  Philosophical Perspectives. Eds. Henk De Regt, Sabina Leonelli, and Kai Eigner. Pittsburgh: University of Pittsburgh Press. Liquin, Emily, and Lombrozo, Tania. 2018. “Structure-​Function Fit in the Evaluation of Teleological Explanations.” Cognitive Psychology 107: 22–​43. Lombrozo, Tania. 2009. “Explanation and Categorization: How ‘Why?’ Informs ‘What?.’” Cognition 110(2): 248–​253. Lombrozo, Tania. 2010. “Causal-​Explanatory Pluralism: How Intentions, Functions, and Mechanisms Influence Causal Ascriptions.” Cognitive Psychology 61(4): 303–​332. Lombrozo, Tania, and Carey, Susan. 2006. “Functional Explanation and the Function of Explanation.” Cognition 99(2): 167–​204. Lombrozo, Tania, and Gwynne, Nicholas Z. 2014. “Explanation and Inference: Mechanistic and Functional Explanations Guide Property Generalization.” Frontiers in Human Neuroscience 8: 700. Lombrozo, Tania, Kelemen, Deborah, and Zaitchik, Deborah. 2007. “Inferring Design:  Evidence of a Preference for Teleological Explanations in Patients with Alzheimer’s Disease.” Psychological Science 18(11): 999–​1006. Lombrozo, Tania, and Rehder, Bob. 2012. “Functions in Biological Kind Classification.” Cognitive Psychology 65(4): 457–​485. Malle, Bertram F., and Hodges, Sara D. 2005. Other Minds. New York: Guilford Press. Murphy, Gregory. L., and Medin, Douglas L. 1985. “The Role of Theories in Conceptual Coherence.” Psychological Review 92(3): 289. Murray, D., and Lombrozo, T. 2017. “Effects of Manipulation on Attributions of Causation, Free Will, and Moral Responsibility.” Cognitive Science 41(2): 447–​481. ojalehto, bethany, Waxman, Sandra R., and Medin, Douglas L. 2013. “Teleological Reasoning about Nature:  Intentional Design or Relational Perspectives?” Trends in Cognitive Sciences 17(4): 166–​171. Prasada, Sandeep. 2016. “Mechanisms for Thinking about Kinds, Instances of Kinds, and Kinds of Kinds.” In Core Knowledge and Conceptual Change. Eds. David Barner and Andrew S. Baron. New York: Oxford University Press. Price, Huw. 2011. Naturalism without Mirrors. New York: Oxford University Press. Rozenblit, Leonid, and Keil, Frank. 2002. “The Misunderstood Limits of Folk Science: An Illusion of Explanatory Depth.” Cognitive Science 26(5): 521–​562.

Mechanistic versus Functional Understanding  229 Sánchez Tapia, Ingrid, Gelman, Susan A., Hollander, Michelle A., Manczak, Erika M., Mannheim, Bruce, and Escalante, Carmen. 2016. “Development of Teleological Explanations in Peruvian Quechua‐Speaking and US English‐Speaking Preschoolers and Adults.” Child Development 87(3): 747–​758. Shtulman, Andrew, and Lombrozo, Tania. 2016. “Bundles of Contradiction:  A Coexistence View of Conceptual Change.” In Core Knowledge and Conceptual Change. Eds. David Barner and Andrew S. Baron. New York: Oxford University Press. Steiner, Samantha M., Zemla, Jeffrey C., and Sloman, Steven. 2016. “MP10: Analytical Style Predicts Religious and Teleological Belief.” Journal of Investigative Medicine 64(3): 808–​809. Strevens, Michael. 2013. “No Understanding without Explanation.” Studies in History and Philosophy of Science Part A 44(3): 510–​515. Trout, J. D. 2007. “The Psychology of Scientific Explanation.” Philosophy Compass 2(3): 564–​591. Vasilyeva, Nadya, Blanchard, Thomas, and Lombrozo, Tania. 2018. “Stable Causal Relationships Are Better Causal Relationships.” Cognitive Science 42(4): 1265–​1296. Vasilyeva, Nadya, Wilkenfeld, Daniel, and Lombrozo, Tania. 2017. “Contextual Utility Affects the Perceived Quality of Explanations.” Psychonomic Bulletin and Review 24(5): 1436–​1450. Wilkenfeld, D. A. Forthcoming. Understanding as Compression. Philosophical Studies. Woodward, Amanda L. 1998. “Infants Selectively Encode the Goal Object of an Actor’s Reach.” Cognition 69(1): 1–​34. Woodward, James. 2003. Making Things Happen: A Theory of Causal Explanation. New York:  Oxford University Press. Woodward, James. 2006. “Sensitive and Insensitive Causation.” Philosophical Review 115(1): 1–​50. Wright, Larry. 1976. Teleological Explanations:  An Etiological Analysis of Goals and Functions. Berkeley: University of California Press.

12 Are Humans Intuitive Philosophers? Steven Sloman, Jeffrey C. Zemla, David Lagnado, Christos Bechlivanidis, and Babak Hemmatian

Philosophers have some very compelling ideas about what makes an explanation good. A kind of modal model from philosophy would characterize the best possible explanation as simple and coherent. It should identify the variables that make a difference to the explanandum and show how they do so. In the process, it should offer a unified account of as much data as possible. Are these the criteria that people use to evaluate everyday explanations? We focus on three related explanatory virtues:  simplicity, coherence, and unification. We will first briefly review the treatment of these virtues by philosophers—​leaving detailed discussion to the philosophers who have contributed to this collection—​before considering the psychological reality of each criterion in everyday judgment.

1.  Explanation from the Philosopher’s Perspective Most philosophers have considered simplicity to be an explanatory virtue since William of Occam in the 14th century (see Baker, 2010, for a review). There’s much less agreement on the meaning of simplicity. Aristotle considered an account simpler than another if it made fewer assumptions, a view echoed by Kant among others. But philosophers like Aquinas and even Isaac Newton considered a simpler explanation one that appealed to fewer causes (see Pacer & Lombrozo, 2017). Modern views of simplicity appeal to the length of the code required to generate the explanandum (the minimum description length or Kolmogorov complexity; Kolmogorov, 1960; Chater, 1996). Simplicity has also been defined in terms of the size of the hypothesis space that an explanation draws on (Tenenbaum & Griffiths, 2001). These definitions of simplicity clearly differ, but they all share the claim that less is more. The difference is in the scale of the quantity being measured.

232  Steven Sloman et al. Simpler explanations enjoy two virtues: First, by asserting less, they are more likely to be true. There are fewer ways they can be wrong. Second, all else being equal, they are easier to understand. The less there is to understand, the easier it is to do so. Thus, simplicity sits comfortably with a second explanatory virtue, coherence. A good explanation does not violate what we already believe. It is also consistent with the evidence and with itself (e.g., Harman, 1980; Thagard, 1989). An explanation that coheres with other beliefs and with evidence is also easier to understand because it does not lead to internal contradictions; there is nothing to rectify. To the degree that the purpose of an explanation is to aid understanding—​to help create a mental representation that reveals causes and makes predictions—​both coherence and simplicity are naturally virtuous. Coherence does suffer from a problem analogous to simplicity: There is no agreed-​upon definition of it. Calling an explanation coherent if and only if it is logically valid (e.g., Hempel & Oppenheim, 1948)  does not admit degrees of coherence, and this makes it difficult to rank explanations by their coherence (Millgram, 2000). Moreover, determining logical validity quickly encounters problems of tractability, as the number of possible valid explanations increases exponentially with the number of atomic propositions posited (e.g., Kornblith, 1989). Another potential explanatory virtue concerns the breadth of application of an explanation. One view is that a good explanation is one that unifies the widest possible range of phenomena (Kitcher, 1989). This entails that explanations should be abstract and idealized (e.g., Jorland, 1994; Strevens, 2007). Garfinkel (1981) calls explanations that contain irrelevant details “hyperconcrete.” They are “too good to be true” and also “too true to be good” (Garfinkel, 1981: 58). In contrast, abstract explanations are more robust to slight perturbations and thus can be used to explain a wider range of phenomena. Interventionist and counterfactual views of explanation also favor abstraction (e.g., Hitchcock & Woodward, 2003; Kuorikoski & Ylikoski, 2010; Strevens, 2007; Weslake, 2010). Strevens (2007) focuses on difference makers:  facts or events necessary for the explanandum to occur. Starting from the most detailed description of the event one is trying to explain, the prescription is to keep removing, abstracting, or idealizing information until the event ceases to obtain. So, even though there is a gravitational force operating between a ball and a bat that influences the speed of the ball, that influence is negligible and thus can and should be ignored

Are Humans Intuitive Philosophers?  233 when explaining how a hitter hit a home run. The fact or event one is trying to explain should constrain what is relevant and, thus, what should be included in the explanation by removing information that makes no difference. Although this position that unification is a virtue is relatively common, other philosophers have taken the opposite view, that abstraction forces an undesirable departure from reality. Cartwright (1999) argues that an ideal explanation in science would include all relevant factors, described with as much precision as possible. She argues that abstract explanations cannot be true as they require idealized conditions that are not found in nature. Other philosophers occupy a middle ground: Railton (1981) argues that abstraction represents a compromise position. Nowak (1992) sees the issue in terms of progress in theoretical development: Early theories of a phenomenon are abstract but false. Subsequent theories add more relevant factors at higher degrees of precision, thus approximating reality to a successively greater degree. Not unexpectedly, philosophers don’t agree on what constitutes the set of explanatory virtues. But there are strong arguments that the list should include simplicity, coherence, and unification. There is also much agreement on what constitutes the fundamental element of most scientific explanations: causation (e.g., Salmon, 1984). There are cases of other kinds of explanation (law-​governed, mathematical, categorical, logical). But appeals to causation are extremely common in sciences other than fundamental physics. Some philosophers have gone farther, and made a case that mechanisms provide the fundamental unit of explanation. According to Machamer, Darden, and Craver (2000), “Mechanisms are entities and activities organized such that they are productive of regular changes from start or set-​up to finish or termination conditions” (3). An activity is essentially a causal relation applied to an entity. They argue that an appeal to mechanisms can help to address a variety of philosophical problems, including the nature of explanation. Glennan (1996) argues that mechanisms can be used to make sense of the notion of causation, with the proviso that causation in fundamental physics is different.

2. Everyday Explanation We evaluate each of the philosophers’ criteria in turn.

234  Steven Sloman et al.

2.1.  Simplicity Like philosophers, psychologists have championed the epistemic virtue of simplicity, going so far as to suggest that simplicity may be a fundamental cognitive principle in cognition (Chater & Vitanyi, 2003). However, simplicity can take many forms that are not always congruous with each other. A number of studies have sought to uncover when and how people endorse simple explanations, and when they prefer complexity. Zemla et al. (2017) explored whether people prefer simple explanations of everyday phenomena. The authors collected a small corpus of explanations from an online question answering community (Reddit). The topics chosen to study included explanations in social policy, history, public health, and legal domains. Each participant read one of eight questions (the explanandum) and one of three explanans for that question. For example, one question asked, “Why has the price of higher education skyrocketed in the U.S., and who is profiting from it?” Participants rated the explanans on numerous criteria, including the overall quality of the explanation (“This is a good explanation”) and its complexity (“This is a complex explanation”) on Likert scales. Contrary to the prediction that people prefer simpler explanations, we found that our subjective measure of complexity correlated positively with overall explanation quality (R  =  .49). Simpler explanations were judged as worse than explanations rated as more complex for the same explanandum. This suggests that subjective interpretations of what a simple explanation is do not align with general preferences for what a good explanation is. There are a number of deflationary accounts of this finding. One is that participants preferred complex explanations in spite of their complexity. For instance, it is not unreasonable to think that complex explanations might also have more detail and breadth; they explain more things. If a preference for breadth outweighed a preference for simplicity, people may have endorsed explanations that were not simple despite valuing simplicity. Another possibility is that the intuitive conception of simplicity that drove people’s ratings do not align with philosophers’ notion of a simple explanation. For example, people judge explanations as more complex when they use jargon (Cooper, Bennett, & Sukel, 1996), though no philosophical theory has argued specifically that lack of jargon is an explanatory virtue.

Are Humans Intuitive Philosophers?  235 Zemla et al. (2017) examined several other measures of complexity to see if a simplicity preference holds. Superficial measures of complexity such as overall explanation length (i.e., word count) correlated positively with subjective explanation quality (R  =  .60). However, measures of lexical complexity, such as median word frequency (R = −.01) and Flesch Reading Ease (Flesch, 1948; R = −.02) did not, indicating that judgments of complexity were not primarily determined by linguistic factors. A more pertinent measure of simplicity is causal simplicity. Zemla et al. (2017) evaluated the causal structure of the explanations by translating each explanation into a causal graph (e.g., Sloman, 2005). They found that the number of root causes invoked to explain a phenomenon positively correlated with explanation quality (R  =  .64). These findings suggest that people sometimes prefer explanations that appeal to more causes. One way to reconcile this preference with a simplicity principle is to say that people do not necessarily prefer explanations with the fewest number of causes, but instead prefer simplicity in scope; i.e., they prefer explanations with the fewest causes relative to what’s being explained. If an explanation with two causes explains more than twice as much data as an explanation with one cause, the two-​cause explanation is arguably the simpler explanation. However, Zemla et al. (2017) found that participants did not judge explanations with more causes as “explaining other issues besides what was asked” (R = .08) and did not contain more effect nodes in the causal model (R = −.09). In other words, explanations that invoked more causes were not seen as having more breadth. Moreover, Khemlani, Sussman, and Oppenheimer (2011) call into question whether people have a preference for simplicity in scope at all. In one experiment, participants were presented with imaginary “spells” that varied in scope; for example, “Delimenta causes lumps and spots,” while “Homurula causes lumps, spots, and bumps.” When asked to explain the symptoms of a patient with lumps, spots, and bumps, participants naturally suggested that the symptoms were the result of Homurula, which causes all three symptoms. In another condition, participants were uncertain about the patient’s symptoms: the patient experienced lumps and spots, but it was unknown whether the participant was experiencing bumps. In this condition, participants’ preferences were reversed; they now believed the patient’s symptoms were a result of Delimenta. This suggests that even when the number of causes is held constant, people do not necessarily prefer

236  Steven Sloman et al. explanations with a larger scope, unless what is being explained is known with certainty. One might be tempted to conclude from these results that people do not prefer simple explanations at all. However these results run counter to some previous psychological findings. Read and Marcus-​Newhall (1993) found that people generally prefer explanations that appeal to a single, common cause rather than multiple independent causes to explain the same phenomenon. For example, participants were provided a story about Cheryl and judged several explanations for why Cheryl had an upset stomach, gained weight, and had been feeling tired. A  common cause explanation which explained all of Cheryl’s symptoms (Cheryl was pregnant) was rated significantly better than explanations that appealed to multiple causes that only explained a single symptom each (Cheryl has a stomach virus, has stopped exercising, and has mononucleosis). Lombrozo (2007) came to a similar conclusion. Participants were presented with a scenario about a hypothetical alien planet and were asked to explain why an alien has sore minttels and purple spots. They were told that these symptoms could be caused by several underlying conditions: Trichet’s syndrome always causes sore minttels and purple spots; Morad’s disease always causes sore minttels but never purple spots; and Humel’s infection always causes purple spots but never sore minttels. Participants overwhelmingly preferred the common cause explanation that the alien had Trichet’s syndrome (96%) as opposed to a disjunctive explanation that the alien had both Morad’s disease and Humel’s infection or some other combination of causes. If people do generally prefer explanations with fewer causes, is it due to an inherent desire for simplicity, or because explanations with fewer causes are typically more probable (e.g., Lagnado, 1994)? It may be that people may not care about simplicity per se, but instead are concerned with finding the set of causes that are most likely given the explanandum. To address this, Lombrozo (2007) modified her experiment to provide base rate information about the various syndromes. In one condition, participants were told that out of 750 aliens, 50 had Trichet’s syndrome, 190 had Morad’s disease, and 197 had Humel’s infection. If it is assumed that the presence of each condition is independent of others, then the probability of the complex explanation (the alien has Humel’s infection and Morad’s disease) is roughly equal to the probability of the simple explanation (the alien has Trichet’s syndrome). Nonetheless, more than 80% of participants still preferred the common

Are Humans Intuitive Philosophers?  237 cause explanation, suggesting that people do have an inherent preference for simplicity that goes beyond a preference for “probable” explanations. This finding has been replicated and extended in subsequent work. Bonawitz and Lombrozo (2012) found that preschool-​aged children also display a preference for simplicity using a paradigm analogous to Lombrozo (2007), in which children observed different colored chips fall onto a contraption, activating a light and fan. Children preferred to explain the events by stating that a blue chip activated both the light and fan, as opposed to a red chip activating the light and a blue chip activating the fan. Pacer and Lombrozo (2017) further demonstrated that people prefer explanations with a fewer number of root (unexplained) causes, rather than the absolute number of causes. In one experiment, participants were told that an alien’s two symptoms (purple spots and itchy flippets) could be caused by the combination of Tritchet’s disease and Morad’s disease, but that Hummel’s disease caused both Tritchet’s and Morad’s disease. Given an alien with purple spots and itchy flippets, participants preferred to explain the symptoms with one root cause (Hummel’s disease), as opposed to two (the alien does not have Hummel’s disease, but independently developed Tritchet’s and Morad’s disease, which caused the symptoms). Though the former explanation has a larger number of absolute causes (three diseases compared to two), it has fewer root causes (one compared to two). So far we have examined two lines of research with apparently opposed conclusions—​one set of studies indicates that people prefer simple explanations with few causes that are broad in scope, and the other indicates that people prefer complex explanations with multiple causes that are narrow in scope. These findings are hard to reconcile if one imagines that explanations are judged solely based on the number of causes and effects present in an explanation. However explanations typically do more than expose the causes of the explanandum. Many explanations also provide an account of how the causes generate the explanandum, the mechanism that leads from causes to effect (Glennan, 1996). One virtue of mechanisms is that they reveal the scope of an explanation, the enabling conditions that determine when and where causes should be effective (Machamer, Darden, & Craver, 2000; Sloman, 2005). For instance, it’s one thing to know that a disease causes a symptom, but if you know that the mechanism of the disease involves the multiplication of viruses, then you can infer that symptoms can be reduced by creating conditions that make it hard for viruses to replicate. Affording such inferences and predictions increases

238  Steven Sloman et al. people’s sense of understanding compared to corresponding explanations without mechanisms (Johnson & Ahn, 2017; Vasilyeva & Lombrozo, 2015). The credibility of proposed mechanisms are themselves open to evaluation, providing an additional way to evaluate an explanation, and such evaluations may compete with a preference for simplicity. Zemla and Sloman (2016) tested whether a preference for simplicity depends on whether an explanation contains mechanisms or not. In one study, participants were told that Sally experienced dizziness and redness of the face as a result of eating food from a buffet. Participants were told that this might be due to eating food with niacin (which causes both dizziness and redness of the face) or from eating foods with thiamine (which causes dizziness) and biotin (which causes redness of the face).1 Consistent with the simplicity principle, participants preferred to explain Sally’s symptoms by saying she consumed niacin, which causes both symptoms. However, in a separate condition, participants were provided explanations with the same causal relations that also described the mechanisms of action. For example, niacin causes dizziness and redness of the face through production of antibodies attached to mast cells, stimulating release of chemical mediators into tissue. Two separate mechanisms described how thiamine causes dizziness and how biotin causes redness of the face. When provided explanations with these mechanisms, participants’ preference for the simpler explanation was significantly reduced. When explanations contain mechanisms, causal simplicity may play less of a role in judging explanations. Detailed descriptions of mechanisms focus judges on the plausibility of those mechanisms, making general principles like simplicity less salient. For instance, Zemla et al. (2017) found that well-​articulated explanations were judged as better explanations overall (R = .69). Descriptions of mechanisms may even contain irrelevant details that make for an alluring explanation. Weisberg et al., (2008) found that explanations of psychological phenomena that contain irrelevant neuroscientific details were judged by non-​experts as better than those without irrelevant details. Hopkins, Weisberg, and Taylor (2016) found that this effect generalizes not just to irrelevant neuroscientific details, but reductionist details of any kind. Rhodes, Rodriguez, and Shah (2014) replicated the seductive allure effect and asked participants to make additional judgments in order to determine 1 Pseudo-​doctor’s note: These nutrients are not actually known to cause dizziness and redness of the face, but were fabricated for the purposes of the experiment.

Are Humans Intuitive Philosophers?  239 why these irrelevant details were compelling. Of several dependent measures, the authors found that the addition of irrelevant neuroscientific details led to modest increases in ratings of the quality of the scientist providing the explanation (10%) and substantial increases in self-​reported understanding of the mechanisms involved (26%). This suggests that extraneous details may promote a greater sense of understanding about the causal mechanisms of an explanation despite increasing the complexity of the explanation. Overall, the psychological literature suggests that people do have a preference for simple explanations that have few causes, but that this preference is often moderated by a number of factors when evaluating everyday explanations. In particular, people prefer explanations that elaborate on causal mechanisms and provide a greater sense of understanding, even if this increases their complexity.

2.2.  Coherence Everyone seems to agree that coherence is a crucial virtue of an explanation. Indeed Zemla et al. (2017) found that a subjective measure of external coherence (“this explanation fits with what I already know”) correlated .47 with their measure of explanation quality. The correlation of quality with internal coherence (“the parts of this explanation fit together coherently”) was an astounding .82. But coherence remains an elusive concept to define, conceptually or operationally (Millgram, 2000). Thagard (2002; 2012) presents an account of scientific, legal, and everyday reasoning based on the notion of explanatory coherence. The key idea is that explanations are embedded in a network of interrelated hypotheses and evidence, and are evaluated in terms of how well these hypotheses fit or “cohere” with each other and with the evidence. Thagard stipulates seven principles, including that explanatory coherence is a symmetrical relation, that hypotheses cohere with what they explain, and that alternative explanations for the same evidence compete and are thus incoherent with each other. The acceptability of a hypoth­ esis depends on how well it coheres with the other propositions in the network. Thagard proposes a computational model (ECHO) to capture these principles and simulate the process of reasoning by maximizing coherence. Propositions are represented as units in a connectionist network, with excitatory or inhibitory links corresponding to relations of coherence

240  Steven Sloman et al. or incoherence. Reasoning is captured by constraint satisfaction, with the most coherent explanation emerging as the most highly activated set of propositions or hypotheses. While Thagard (2012) acknowledges that coherence is not a substitute for truth, he argues that it can lead to approximate truth. In particular, explanatory coherence increases as theories broaden their evidence base, and recruit causal hypotheses from deeper levels. Deepening is achieved by introducing lower-​level causal mechanisms to explain the higher-​level hypotheses. Thagard thus accepts the critical role of causal explanation in science, medicine, law and everyday reasoning. But his approach suffers from its inability to represent causal reasoning, because it is based entirely on symmetric relations of explanatory coherence. This undermines its status as either a normative or a descriptive model of explanation. Consider the scientific discovery that h.  pylori bacteria cause peptic ulcers, and thus that the presence of this bacteria in the stomach explains why someone develops ulcers (cf. Thagard, 2000). To capture this in a network, ECHO would represent the relation between bacteria and ulcers with a symmetric excitatory link between units representing them, whereby the activation of one leads to the activation of the other. But this link does not capture causal direction, and thus cannot tell us that we can treat (or prevent) ulcers by removing the bacteria but that the converse is false; nor can it distinguish the counterfactual that if one hadn’t acquired the bacteria one would not have got the ulcers with the converse counterfactual. Indeed a major step in the scientific discovery was Marshall’s heroic self-​experiment in which he ingested bacteria to show that ulcers ensued. This form of causal inference is not supported by a purely associative representation of the relation between bacteria and ulcers. ECHO also fails to account for how laypeople actually represent and explain the world. Extensive psychological research shows that people use causal models, and often make interventional or counterfactual inferences that require causal reasoning specifically (e.g., Sloman & Lagnado, 2005; 2015; Waldmann, 2017). People know that symptoms do not cause diseases, and can draw appropriate inferences, for instance, knowing that to prevent a disease one must act on the causes of disease, not the symptoms. But a coherence-​based network alone does not support these kinds of inference.

Are Humans Intuitive Philosophers?  241 2.2.1.. Coherence in Legal Decision-​Making The notion of coherence also plays a central role in the main psychological theory of legal decision-​making, the story model (Pennington & Hastie, 1986; 1992). Unlike connectionist approaches, the story model puts causal representations center stage, thus avoiding the problems that beset purely associative models. On the story model, people construct explanatory narratives to make sense of the evidence presented in a trial. The acceptability of a story is dictated by several principles: coherence, coverage, and uniqueness. The latter two principles are relatively clear-​cut: Better stories cover more of the evidence, and one should have more confidence in a story that has fewer competitors. The notion of coherence is unpacked into three components:  consistency, completeness, and plausibility. Consistency refers to the lack of internal contradictions, and is construed in terms of logical consistency. Completeness is self-​explanatory (although it is unclear how one compares between stories that are incomplete in different ways). Plausibility is trickier to define; according to Pennington and Hastie (1992), it corresponds to how consistent the story is with our real world knowledge. But it’s unclear how such judgments are made or appraised as this is the key issue under discussion. What are the cognitive processes that make one explanation more or less plausible than another, given what we know or assume about the world? Another problem for the story model is that it neglects a crucial dimension in legal decision-​making: how people assess the credibility and reliability of the evidence presented in court, as revealed through the process of witness testimony and cross-​examination. This process is critical not only to a trial but in many investigative contexts. People need to evaluate the credibility of what is said and integrate this into their assessments of the various stories in play, but the story model offers little guidance here (Lagnado & Gerstenberg, 2017; Connor Desai, Reimers & Lagnado, 2016). More generally, when appraising explanatory stories people must evaluate the evidence as well as the story it supports. Sometimes people are driven by the appeal of an explanatory story to the detriment of the evidence (Anderson, Lepper & Ross, 1980; Kuhn, 1991). But understanding how people appraise and assimilate evidence is critical to gain a fuller psychological picture of how explanations are assessed. Here again the notion of coherence seems relevant. We prefer explanations that cohere with the evidence. But coherence is a placeholder for a complex system of evaluation.

242  Steven Sloman et al.

2.3.  Unification: Abstraction and Breadth Should the explanation of an event of interest provide as much information as possible? Despite the obvious pragmatic limit on the detail one is able to transmit, many philosophers have argued for the need to impose further constraints. As discussed earlier, it has been claimed that the best explanations do not contain irrelevant details or information described at an inappropriate level of detail (“hyperconcrete” explanations; Garfinkel, 1981). The question of what determines relevance is actively debated. A solution proposed by some (Hitchcock & Woodward, 2003; Strevens, 2007) is to make an explanation more abstract by removing only non-​difference makers, the information whose absence would make no difference to whether the explanandum occurs or not. While many philosophers nowadays agree with the general need for abstraction, there has been little indication, until recently, as to whether other people intuitively do. Weisberg et  al.’s (2008) finding that adding neuroscientific details increased the allure of behavioral explanations to naive participants and neuroscience students (but not professionals in the field) can be accounted for as a bias in favor of reductionism: including information about more fundamental processes is desirable because it provides a sense of understanding (Trout, 2008), it has been proven beneficial in the past (Hopkins et al., 2016), or it gives the impression that the explainer is more knowledgeable (Rhodes et al., 2014). Bechlivanidis et  al. (2017) investigated whether details that make no causal difference are desirable in explanations, even when they don’t transcend levels. In their experiments, they presented participants with three competing explanations of the same phenomenon that differed in their degree of concreteness or abstraction. One of the explanations was abstract; it included only difference makers, facts or events necessary for the event to occur. The second explanation became more concrete by replacing some qualitative statements with precise details, though the greater precision made no difference to the explanandum. Finally, a “hyperconcrete” explanation was constructed by adding information that was accurate but causally irrelevant to the event being explained. The events and their explanations did not refer to law-​like regularities but to everyday events found in newspapers or discussed in social settings. For example, participants were presented with three explanations about a landslide. The abstract explanation attributed the landslide to the type of

Are Humans Intuitive Philosophers?  243 soil in the hill, the sparse vegetation, and the steep slope. The concrete explanation mentioned in addition the exact size of the soil particles (2/​64 of an inch), the proportion of the hill covered by vegetation (13%) and the exact degree of the slope (37 degrees). Finally, the hyperconcrete explanation elaborated further with causally irrelevant details by mentioning that the hill was close to a festival, that the soil particles were light brown and that the vegetation on the hill was not edible. Participants were asked first to rate each explanation and then to rate the causal role of each piece of information at its various levels of precision. We found that participants showed no preference between the concrete and the hyperconcrete explanations but they significantly penalized abstraction. This occurred despite the fact that they found the abstract and the concrete descriptions to be equally causally relevant, and judged the hyperconcrete details to play no causal role. For example, even if they thought that the “steep slope” influenced the landslide approximately as much as the “37 degree slope” and that they judged the fact that “the slope was 5 miles north of the premises of the annual Lilac festival” to have no causal role, they strongly preferred the explanation that mentioned the precise degree of the slope and did not penalize the inclusion of information about the local festival. These results are reminiscent of Fernbach et  al. (2013), who found an overall preference for explanations for novel consumer products that described how the products worked with a moderate degree of detail. In that study, preferences were shown to depend on individual’s cognitive style. Those who were more reflective (as indicated by the Cognitive Reflection Test; Frederick, 2005) preferred more detail, those who were less reflective preferred less. Bechlivanidis et al. (2017) also found that abstraction was preferred only when concrete information failed to communicate the nature of the mechanism. For example, when explaining a warehouse fire, an abstract explanation referred to the mishandling of a flammable material while a concrete explanation attributed the fire to the mishandling of ethyl chloride. Since most people are not aware of the flammability of ethyl chloride, the concrete explanation failed to convey the critical causal properties that brought the fire about and was penalized by the participants. Together with previous findings (Hopkins et al., 2016; Rhodes et al., 2014), these results show that, in contrast to recent philosophical prescriptions (Garfinkel, 1981; Hitchcock & Woodward, 2003; Strevens, 2007; Weslake, 2010), people do not value abstraction for its own sake. In general, there

244  Steven Sloman et al. appears to be a preference for detail, even if that detail does not transcend levels (Hopkins et al., 2016) and, often, even if it is not obviously related to the fact or event being explained. Thus, although causality—​or difference-​ making—​appears to be a required feature of explanations, it is not the sole determinant of the information that is valued. Given that detail reduces the robustness and generalizability of explanations, and may even potentially misinform, what explains people’s consistent preference for it? If, for example, one includes the particular size of soil particles when explaining a landslide, the explanation will fail to generalize to landslides in hills with a slightly different soil size. More worryingly, the listener might interpret the precise information as a difference maker and be led to conclude that slightly smaller or bigger particles would not result in a landslide. The answer might be related to the assumptions people make about the purpose of an explanation. While a policymaker will prefer robust, generalizable information about difference makers, a private investigator will demand as much detail as possible in order to construct a rich, flexible, and vivid model of the particular event. Abstraction is useful for its generalizability, for applying current explanations to future occurrences but detail might enhance our understanding of particular events. Causally irrelevant details might help us better visualize mechanisms and thus promote a feeling of understanding. At the same time, details that are not causally related to the particular event might still inform us about other aspects of the situation. For example, even if the edibility of the vegetation had no causal role in the landslide, it provides information about different aspects we might be interested in the future, such as whether the local agriculture will be affected.

2.4. Unification: Pure Abstraction Perhaps the most abstract type of explanation is a categorical one, whereby the name of a category is provided as an explanation for a feature that has some connection to that category.2 For example, if someone asks why Chewbacca is hairy, a natural answer would be “because he is a Wookie.” With a single phrase, or even a single word, a categorical explanation can 2 Prasada and Dillingham (2006) call this formal explanation in reference to the idea of a formal cause in Aristotelian philosophy.

Are Humans Intuitive Philosophers?  245 summarize what we know about the instances of a category. It excludes or includes certain features, determines the allowed interconnections among features and categories, and points to evidence for a variety of category-​ related assertions. Prasada and Dillingham (2006) show that there are reliable differences in the subjective goodness of categorical explanations, which they attribute to the presence or absence of “principled connections” between category members and the feature to be explained. Giffin, Wilkenfeld, and Lombrozo (2017) have demonstrated that a label can serve as a better explanation than the property that the label refers to by implying reliable causes that link the category and the explanandum. Other causal considerations, such as the number of possible pathways to the feature to be explained, and the persistence of the causal link under changes in background conditions (Woodward, 2006), can also impact the subjective goodness of categorical explanations (Hemmatian & Sloman, 2018). In fact, Wilkenfeld, Asselin, and Lombrozo (2016) suggest that a label can lead people to believe there’s a cause even when they are explicitly told that there is not. Hemmatian and Sloman (2018) examined whether judgments of the quality of categorical explanations were sensitive to social considerations. Our individual experiences and knowledge are limited, even when dealing with familiar categories (Rozenblit & Keil, 2002). Instead, we rely on the more extensive knowledge that other members of our community possess (Sloman & Fernbach, 2017). We are so dependent on this communal aspect of knowl­ edge that philosophers like Wittgenstein (1953) do not consider utterances meaningful unless they use references consented to by a community of speakers. Furthermore, Putnam (1979) notes that a division of linguistic labor is in place among speakers: As each individual focuses on a certain subset of categories, learning what there is to know about their boundaries, features, and interconnections, the community as a whole becomes capable of using that information to further its goals. There is accumulating evidence that people utilize this vast communal information on a regular basis, not just in language, but also in thought (Sloman & Fernbach, 2017). Hemmatian and Sloman (2018) address the role of community in categorical explanation directly. Participants were provided with scenarios in which a character observed several instances of a social or natural category, for example, stars with variable brightness. In some conditions, the character proceeded to name the category herself (e.g., Evans star), and later used it in a categorical explanation (“Because it is an Evans star”). In other conditions,

246  Steven Sloman et al. the character learned the name of the category from community members. Participants were then told about a property of the category and asked to rate how natural the label sounded as an explanation for the property, as well as the degree to which the label served as an explanation. Labels entrenched in the community resulted in significantly higher ratings in response to both questions. Attending to these community cues makes a lot of sense normally, as the community often has access to information unknown to the individual. But community cues can be used even when they do not carry informational value. In a different experiment, Hemmatian and Sloman (2018) explicitly told participants that nothing other than the property identified by the label is known about the category in each scenario. In this case, regardless of labels’ community entrenchment, the categorical explanation reduces to a reiteration of the question: Evans star simply means a star with variable brightness. Despite its lack of information content, the effect of community entrenchment persisted. That community considerations can overshadow informational criteria in judgments of explanations is a testament to the fundamentally social nature of human cognition. Even though humans are potentially sensitive to various information sources discussed in philosophical theories, it appears that the status of a category within the community takes precedence, with individuals trusting that the vast communal intelligence around them will provide satisfactory answers to the difficult questions that the world presents.

3.  Conclusion Humans are not merely intuitive philosophers. How a person evaluates an explanation depends on what that person is trying to achieve. Philosophers sometimes presuppose a specific goal: that the explainer is trying to offer an explanation that is most likely to be correct while being understood and providing an account of the facts, usually a causal account. Explanations are likely to the degree they are simple and unifying. They are understood to the degree they are coherent. If you’re a philosopher—​indeed, any agent—​with these goals, the standard virtues of simplicity, coherence, and unification will help you achieve them. But people don’t always prefer the most likely explanation. This may be because they often have other goals. Sometimes they don’t care about the

Are Humans Intuitive Philosophers?  247 probability that the explanation is correct; they just want to make sure the explanandum seems likely. If you’re looking for an explanation to persuade others, you might not be concerned about the truth of the explanation, but that it makes some outcome seem inevitable. Sometimes people don’t know what the difference makers are, so they want to include detail that provides as much coverage as possible of the variables that might be relevant. Many of us are curious about how a man like Donald Trump could have won the 2016 presidential election in the United States. There is no dearth of hypotheses—​the economic situation in the Rust Belt, the failure of the Democrats to offer a positive message, the appeal of a reality-​TV star, the modern-​day appeal of contentless consumer branding, Trump’s willingness to speak his mind, etc.—​but nobody knows which hypotheses are correct or how much weight each had in the final outcome. And presumably each accounts for some of the dynamics in a horribly complex system that doesn’t lend itself to simple explanations. Perhaps the best we can do is list all the potentially relevant contributors. Sometimes people want to convince listeners by spelling out mechanisms to increase coherence even though there’s not strong evidence for those mechanisms. Sometimes they want to tell a good story, and this requires adding extraneous detail that helps the listener construct a narrative, not merely an explanation. This is why historical novels can give a better sense of understanding than bland but accurate history texts. And people may have other goals as well. One we have not even considered is that people sometimes just want to impress listeners with their knowledge and erudition. Explanation is a multipurpose tool that serves functions beyond those imagined by both philosophers and psychologists. People seem to appreciate those different functions and, in that sense, are not merely intuitive philosophers.

Acknowledgments This publication was made possible through a grant from the Varieties of Understanding Project at Fordham University and the John Templeton Foundation. The opinions expressed in this publication are those of the authors and do not necessarily reflect the views of the Varieties of Understanding Project, Fordham University, or the John Templeton Foundation.

248  Steven Sloman et al.

References Anderson, C. A., Lepper, M. R., & Ross, L. (1980). Perseverance of social theories: The role of explanation in the persistence of discredited information. Journal of Personality and Social Psychology, 39(6), 1037. Bechlivanidis, C., Lagnado, D. A., Zemla, J. C., & Sloman, S. (2017). Concreteness and abstraction in everyday explanation. Psychonomic Bulletin and Review, 24, 1451–​1464. Bonawitz, E. B., & Lombrozo, T. (2012). Occam’s rattle: Children’s use of simplicity and probability to constrain inference. Developmental Psychology, 48(4), 1156–​1164. Cartwright, N. (1999). The Dappled World:  A Study of the Boundaries of Science. Cambridge University Press. Chater, N. (1996). Reconciling simplicity and likelihood principles in perceptual organization. Psychological Review, 103(3), 566–​581. Chater, N., & Vitányi, P. (2003). Simplicity: A unifying principle in cognitive science? Trends in Cognitive Sciences, 7(1),  19–​22. Connor Desai, S., Reimers, S., & Lagnado, D. (2016). Consistency and credibility in legal reasoning: A Bayesian network approach. In Proceedings of the 38th Annual Conference of the Cognitive Science Society. Austin, TX: Cognitive Science Society, 626–​631. Cooper, J., Bennett, E. A., & Sukel, H. L. (1996). Complex scientific testimony: How do jurors make decisions? Law and Human Behavior, 20(4), 379–​394. Fernbach, P. M., Sloman, S. A., Louis, R. S., & Shube, J. N. (2013). Explanation fiends and foes: How mechanistic detail determines understanding and preference. Journal of Consumer Research, 39(5), 1115–​1131. Flesch, R. (1948). A new readability yardstick. Journal of Applied Psychology, 32(3), 221–​233. Frederick, S. (2005). Cognitive reflection and decision making. Journal of Economic Perspectives, 19(4), 25–​42. Garfinkel, A. (1981). Forms of Explanation:  Rethinking the Questions in Social Theory. New Haven, CT: Yale University Press. Giffin, C., Wilkenfeld, D. A., & Lombrozo, T. (2017). The explanatory effect of a label: Explanations with named categories are more satisfying. Cognition, 168, 357–​369. Glennan, S. S. (1996). Mechanisms and the nature of causation. Erkenntnis, 44(1),  49–​71. Harman, G. (1980). Reasoning and explanatory coherence. American Philosophical Quarterly, 17(2), 151–​157. Hemmatian, B., & Sloman, S. A. (2018). Community appeal: Explanation without information. Journal of Experimental Psychology: General, 147(11), 1677–​1712. Hempel, C. G., & Oppenheim, P. (1948). Studies in the logic of explanation. Philosophy of Science, 15(2), 135–​175. Hitchcock, C., & Woodward, J. (2003). Explanatory generalizations, part II: Plumbing explanatory depth. Noûs, 37(2), 181–​199. doi:10.1111/​1468-​0068.00435. Hopkins, E. J., Weisberg, D. S., & Taylor, J. C. (2016). The seductive allure is a reductive allure: People prefer scientific explanations that contain logically irrelevant reductive information. Cognition, 155,  7–​76. Johnson, S. G. B. & Ahn, W.-​K. (2017). Causal mechanisms. In Michael R. Waldmann (Ed.), The Oxford Handbook of Causal Reasoning (pp. 127–​146). New  York:  Oxford University Press. Jorland, G. (1994). Idealization and transformation. In B. Hamming & N. B. De Marchi (Eds.), Idealization VI: Idealization in Economics (pp. 265–​275). Atlanta, GA: Rodopi.

Are Humans Intuitive Philosophers?  249 Khemlani, S. S., Sussman, A. B., & Oppenheimer, D. M. (2011). Harry Potter and the sorcerer’s scope: Latent scope biases in explanatory reasoning. Memory and Cognition, 39(3), 527–​535. Kitcher, P. (1989). Explanatory unification and the causal structure of the world. Scientific Explanation, 13, 410–​505. Kolmogorov, A. N. (1963). On tables of random numbers. Sankhyā: The Indian Journal of Statistics, Series A, 369–​376. Kornblith, H. (1989). The unattainability of coherence. In The Current State of the Coherence Theory (pp. 207–​214). Dordrecht: Springer Netherlands. Kuhn, D. (1991). The Skills of Argument. New York: Cambridge University Press. Kuorikoski, J., & Ylikoski, P. (2010). Explanatory relevance across disciplinary boundaries:  The case of neuroeconomics. Journal of Economic Methodology, 17(2), 219–​228. Lagnado, D. A. 1994. The Psychology of Explanation: A Bayesian Approach. Master’s thesis, Schools of Psychology and Computer Science, University of Birmingham. Lagnado, D. A., & Gerstenberg, T. (2017). Causation in legal and moral reasoning. In M. Waldmann (Ed.), Oxford Handbook of Causal Reasoning (pp. 565–​ 602). New York: Oxford University Press. Lombrozo, T. (2007). Simplicity and probability in causal explanation. Cognitive Psychology, 55(3), 232–​257. Machamer, P., Darden, L., & Craver, C. F. (2000). Thinking about mechanisms. Philosophy of Science, 67(1),  1–​25. Millgram, E. (2000). Coherence:  The price of the ticket. Journal of Philosophy, 97(2),  82–​93. Nowak, L. (1992). The idealization approach to science: A survey. In J. Brezinski & L. Nowak (Eds.), Idealization III:  Approximation and Truth (pp. 9–​63). Amsterdam; Atlanda GA: Rodopi. Pacer, M., & Lombrozo, T. (2017). Ockham’s razor cuts to the root: Simplicity in causal explanation. Journal of Experimental Psychology: General, 1761–​1780. Pennington, N., & Hastie, R. (1986). Evidence evaluation in complex decision making. Journal of Personality and Social Psychology, 51(2): 242. Pennington, N. & Hastie, R. (1992). Explaining the evidence: Tests of the story model for juror decision making. Journal of Personality and Social Psychology, 62(2):189–​206. Prasada, S., & Dillingham, E. M. (2006). Principled and statistical connections in common sense conception. Cognition, 99(1), 73–​112. Putnam, H. (1979). Philosophical Papers, vol. 2, Mind, Language and Reality. New York: Cambridge University Press. Railton, P. (1981). Probability, explanation, and information. Synthese, 48(2), 233–​256. Read, S. J., & Marcus-​Newhall, A. (1993). Explanatory coherence in social explanations: A parallel distributed processing account. Journal of Personality and Social Psychology, 65(3), 429–​447. Rhodes, R. E., Rodriguez, F., & Shah, P. (2014). Explaining the alluring influence of neuroscience information on scientific reasoning. Journal of Experimental Psychology: Learning, Memory, and Cognition, 40(5), 1432–​1440. Rozenblit, L., & Keil, F. (2002). The misunderstood limits of folk science: An illusion of explanatory depth. Cognitive Science, 26(5), 521–​562. Salmon, W. (1984). Scientific Explanation and the Causal Structure of the World. Princeton, NJ: Princeton University Press.

250  Steven Sloman et al. Sloman, S. A. (2005). Causal Models:  How People Think about the World and Its Alternatives. New York: Oxford University Press. Sloman, S. A., & Fernbach, P. (2017). The Knowledge Illusion: Why We Never Think Alone. New York: Riverhead Books. Strevens, M. (2007). Why explanations lie: Idealization in explanation. Unpublished manuscript. Retrieved from http://​www.strevens.org/​research/​expln/​Idealization.pdf. Tenenbaum, J. B., & Griffiths, T. L. (2001). Generalization, similarity, and Bayesian Inference. Behavioral and Brain Sciences, 24(4), 629–​640. Thagard, P. (1989). Explanatory coherence. Behavioral and Brain Sciences, 12(3), 435–​467. Thagard, P. (2000). How Scientists Explain Disease. Princeton, NJ:  Princeton University Press. Thagard, P. (2002). Coherence in Thought and Action. Cambridge, MA: MIT Press. Thagard, P. (2012). The Cognitive Science of Science:  Explanation, Discovery, and Conceptual Change. Cambridge, MA: MIT Press. Trout, J. D. (2008). Seduction without cause: Uncovering explanatory neurophilia. Trends in Cognitive Sciences, 12(8), 281–​282. Vasilyeva, N., & Lombrozo, T. (2015). Explanations and causal judgments are differentially sensitive to covariation and mechanism information. In Proceedings of the 37th Annual Conference of the Cognitive Science Society (pp. 2475–​2480). Austin, TX: Cognitive Science Society. Waldmann, M. R. (2017) (Ed.). The Oxford Handbook of Causal Reasoning. New York:  Oxford University Press. Weisberg, D. S., Keil, F. C., Goodstein, J., Rawson, E., & Gray, J. R. (2008). The seductive allure of neuroscience explanations. Journal of Cognitive Neuroscience, 20(3), 470–​477. Weslake, B. (2010). Explanatory depth. Philosophy of Science, 77(2), 273–​294. doi:10.1086/​ 651316. Wilkenfeld, D. A., Asselin, J., & Lombrozo, T. (2016). Are symptom clusters explanatory? A study in mental disorders and non-​causal explanation. In A. Papafragou, D. Grodner, D. Mirman, & J. C. Trueswell (Eds.), Proceedings of the 38th Annual Conference of the Cognitive Science Society (pp. 1979–​1984). Austin, TX: Cognitive Science Society. Wittgenstein, L. (1953:2010). Philosophical Investigations. Hoboken, NJ:  John Wiley and Sons. Woodward, J. (2006). Some varieties of robustness. Journal of Economic Methodology, 13(2), 219–​240. Zemla, J. C., & Sloman, S. (2016). The role of mechanistic information in explanatory preference. Talk presented at the 16th International Conference on Thinking, Providence, RI. Zemla, J. C., Sloman, S., Bechlivanidis, C., & Lagnado, D. A. (2017). Evaluating everyday explanations. Psychonomic Bulletin and Review, 24, 1488–​1500.

III

T HE OLO G Y OF U N DE R STA NDI NG

13 Religious Understanding and Cultured Practices Terrence W. Tilley

In a letter to the editor of the New  York Times, soi-​disant “free-​thinker” Richard C. Johnson wrote that both believers and atheists are afflicted with doubt about the existence of God. Both are “hamstrung by the seeming fact that the existence of God may neither be proved or disproved.” He then continues: There is a compelling argument, however, that God is an artifact of human consciousness. Consciousness creates within us a serious identity problem. We strive to know how we came to be. God is just the latest version of our relentless search for identity. It is human nature for us to search for God in the natural world about us. But it’s the wrong place. God is just a product of our consciousness.1

Like Ludwig Feuerbach and Sigmund Freud, Johnson is one of many who assume or imply that because we “only” imagine God, there is no reason to think that God is there to be imagined. We project into the world what is only in our minds. “God” represents no reality other than providing a terminus in our quest for identity. When we think are engaging in a form of religious understanding, we are actually engaging in a confused practice of seeking our identities. In this paper, I simply accept the claim that religious practitioners imaginatively produce concepts of God shaped by their desires.2 But I reject the 1 Richard C. Johnson, Letter to the Editor, “Sunday Review,” New York Times, April 3, 2016, 8. Of course, it seems empirically false that many members of either group are afflicted with doubt. 2 For a sophisticated analysis and critique of historic projection theories, see Van A. Harvey, Feuerbach and the Interpretation of Religion (Cambridge: Cambridge University Press, 1995). I do not share all his conclusions, and the present strategy—​directed at more vulgar accounts than Harvey’s foils—​is much simpler.

254  Terrence W. Tilley claim that an imaginative origin for religious belief implies that “God is just a product of our consciousness.” I explore a way that people come to understand the unseen in practice, and provide an indirect way for appraising those claims about the transcendent dimension.

1.  The Problems with Projection Theories Projection theories have two fundamental problems, one about the flawed origin of religious beliefs and one about the unwarrantability of religious beliefs. First, projection theories presume that religious concepts and beliefs are formed irrationally.3 Freud, in The Future of an Illusion (1927), finds that belief in God is an illusion.4 Illusions are beliefs with two defining characteristics: we hold them because we want them to be true, and we do not examine them to see if they are either warranted or likely to be true. What Freud said about belief in God can be understood mutatis mutandis about other beliefs that emerge from religious understanding such as belief in nirvana, karma-​ samsara, or Amida Buddha. By engaging in specific religious practices such as prayer, meditation, repentance, pilgrimage, charity, etc., religious practitioners learn who and where they are in the grand order. It is not merely that a practitioner wishes to be saved from death or to believe that the unfair evil endured, and the good deeds done, have a place in the great cosmic scheme. Rather, religious practitioners learn how to have such desires in the practice of believing in karma and samsara or in a divine judgment. Freud and Johnson are partially right: we build our beliefs about the transcendent out of our needs and desires. But such an account is too simple. The very participation in religious (and quasi-​religious) practices teaches us what to want—​nirvana or heaven 3 As Fiona Ellis notes, the standards for “rational” understanding are disputed and arbitrary. See “Religious Understanding, Naturalism and Desire,” in Making Sense of the World: New Essays in the Philosophy of Understanding, ed. Stephen R. Grimm (New  York:  Oxford University Press, 2017), especially her analysis of the approach of John Cottingham. Using historical understanding as the test case, I have argued that the problem is the unwarranted and often simply assumed extension of the necessary presumptions for engaging in critical historical research (especially methodological atheism) to metaphysical assertions about the nature of reality (substantive atheism). See my History, Theology and Faith: Dissolving the Modern Problematic (Maryknoll, NY: Orbis Books, 2004), 37–​48. I would argue that this move is common to reductionist naturalisms, but that is beyond the scope of this paper, and Ellis has made similar arguments in her paper. 4 See Sigmund Freud, The Future of an Illusion, trans. James Strachey (New York: Norton, 1961).

Religious Understanding and Cultured Practices  255 or money and power. And in so doing we come to learn how to walk paths to such goals. In walking such paths, we create our identities.5 Of course, some of us sometimes believe what we wish to be true only because we wish it, but that is hardly limited to religious concepts—​a point altogether too obvious in the political debacle that was the U.S. presidential race in 2016. However, on Freud’s account—​and in the view of many reductive naturalists—​religious belief is as irrationally founded as my belief that the world is fair because I wish it to be so. If religious beliefs be Freudian illusions, then the rational thing to do would be to abandon them. Reasonable persons believe what experience, evidence, and argument shows to be true, not what they wish to be true. More interesting is Freud’s rebuttal of a counter-​ argument that he envisioned religious believers would make against his views. Freud wrote that believers would accuse him of substituting his own emotionally unsatisfying and cold illusion—​the illusion of science—​for the satisfying and warm illusion of religion. The belief in science, they would argue, is just as much based in the scientists’ wish that science would solve the problems of humanity as is some religious believer’s faith that God will solve all human problems is based in the believer’s wish. Freud responded that science, if it were an illusion, would be a different sort of illusion.6 Unlike religious belief and practice, he said, science is self-​ correcting. If a scientist held false beliefs, the practice of science can and does correct them. In contrast, Freud believed, faith is not merely an illusion but an irrational belief, a delusion because it cannot admit of correction. Hence, even if science were an illusion (and he did not think it was, but accepted the point for the purpose of argument), it would be a healthy, self-​correcting illusion, not a diseased delusion. Freud was surely correct about some religious beliefs—​those that are fossilized, frozen into unchangeable beliefs. Yet he was wrong about other beliefs. Most religious traditions recognize that their formulations of their beliefs, even some of their most central convictions, can require modification in the 5 Some philosophers of religion have recognized that religious believing is a practice and that we can understand the significance of religious beliefs through participating in a panoply of religious practices (please note the absence of “only” between “can” and “understand”; this limitation, characteristic of some forms of fideism, is, in my view, unwarranted). See William P. Alston, Perceiving God: The Epistemology of Religious Experience (Ithaca, NY: Cornell University Press, 1991); Terrence W. Tilley, The Wisdom of Religious Commitment (Washington, DC: Georgetown University Press, 1995); and Ellis, “Religious Understanding.” 6 See Freud, Future of an Illusion,  54–​56.

256  Terrence W. Tilley light of further experience and understanding. While religious traditions are not scientific and do not make scientific progress, those faith traditions that recognize the need for the development of doctrine live on in a changing world and are not frozen into delusion.7 Moreover, religious people grow in wisdom.8 Many scholars write of faith as a journey.9 We begin with a faith we inherit, a naive faith. We generally accept what we are told. We engage in religious practices. We develop religious beliefs. We learn how to be good religious people. But then our faiths are challenged. We realize, often in adolescence, that the comforting stories told us in our youth are simply not very credible. The evils in the world challenge the image of God as all-​good and all-​powerful. The failures of religious or political leaders to be worthy of our trust undermine not only their credibility, but the credibility of what they taught us. We learn other practices and develop virtues of the mind characteristic of critical investigative practices in the humanities, sciences, and social sciences.10 Engaging in such practices may lead us to lose a religious faith, i.e., to lose the ability to see the point of religious practice and beliefs. The causes of such “loss” (or “gain,” if one is more skeptical) are varied. Or we may discover that other people practice different faiths and realize that what we were told was “certain” is opposed by people from other traditions who are as wise and good as members of our own tradition. We have a “crisis of faith.” Some of us resolve that crisis by affirming the faith we were taught in a nuanced and critical manner. We integrate the critical practices we have learned with our religious practices. We accept the tradition we were given as a whole, but may not buy into some of its parts. Others of us reject the tradition we were taught. Perhaps we practice and believe non-​religious humanism or scientific naturalism. We come to a “new,” “mature” faith, appropriate for a

7 The Roman Catholic Church, for example, condemned freedom of conscience in the nineteenth century, but advocated for it as a political right in the final third of the twentieth century. For a Lakatosian account of development in theology, see Nancey Murphy, Theology in the Age of Scientific Reasoning (Ithaca, NY: Cornell University Press, 1990). 8 The next four paragraphs are indebted to Paul Ricoeur’s understanding of “second naivete,” modified somewhat for the present argument. See Paul Ricoeur, The Symbolism of Evil, trans. E. Buchanan (New York: Harper & Row, 1967), 349. 9 This is a major theme in Ellis, “Religious Understanding.” 10 The allusion here is to Linda Trinkaus Zagzebski, The Virtues of the Mind: An Inquiry into the Nature of Virtue and the Ethical Foundations of Knowledge (Cambridge: Cambridge University Press, 1996); also see her “Religious Knowledge and the Virtues of the Mind,” in Rational Faith: Catholic Responses to Reformed Epistemology, ed. Linda Zagzebski (Notre Dame: University of Notre Dame Press, 1993), 199–​225.

Religious Understanding and Cultured Practices  257 “rational” person. In effect, we fully immerse ourselves in other practices and abandon our religious practice and the beliefs it generates.11 Those faiths that incorporate religious practices that enable them to develop over time—​whether personal or communal faiths—​show that even if religious understanding can produce illusions, religious understanding is not in itself irrational. Freud also implicitly concedes that practicing science is at least quasi-​religious worship of Ananke and logos.12 Are we “hamstrung,” as Johnson put it, because there is no conclusive proof about the ultimate efficacy of either scientific or religious practices? I address this in the third section of this paper. Second, many arguments against particular religious understandings exemplify the genetic fallacy.13 The genetic fallacy claims that if a belief has a particular origin or genesis, that makes it false. While the genesis of a belief may affect its warrant or justification, a particular origin does not make it true—​or false. Johnson claims that because we desire to find God in (or through or under or behind, I’d add) the natural world, and because the concept of God arose in our heads out of our need for identity, the natural world is the wrong place to look for God. What Johnson—​following Freud—​does is deny any plausibility to a religious concept to its genesis in the illusion producing practice that is religious understanding. But then all our concepts are mental—​they all originate in our minds. We develop them because we desire something—​perhaps fame, perhaps fortune, perhaps power, perhaps knowledge. If we were to say that because nuclear fusion is a concept that originates in our minds because of our desire to harness atomic power, or because its originators wanted fame or fortune, we should not look for nuclear fusion in theworld, but in our minds, we would be ludicrous. Johnson’s argument can be applied mutatis mutandis to every concept, belief, or claim, religious or not. The scientific community did not commit the genetic fallacy and reject claims to cold fusion because they came from

11 These two journeys have different shapes. But they are both journeys of developing religious understanding, one to a reinvigorated appreciation of “the old faith” and the other to a commitment to a “new creed.” Many find new ways to live out their old creeds. But some move beyond their tradi­ tion. Indeed, Freud himself began his life in a religious environment and wound up with faith in science. His practice of psychoanalysis generated beliefs he found incompatible with religious believing. 12 Freud, Future of an Illusion,  53–​54. 13 Some genetic fallacies are of the “ad hominem” variety:  “She believes in Christianity simply because she was raised in a Christian home and thus you cannot take her seriously” or “She is an atheist because she was raised in an atheistic household and her thinking is therefore corrupted.” Unfortunately, many such fallacies crop up in popular religious debates.

258  Terrence W. Tilley Utah in 1989. They rejected the claims because the experiments were flawed and the results could not be replicated. In sum, arguing against a claim because it can be reduced to a particular origin is a fallacy. The legitimate argument against a concept is to use the best available methods and criteria to evaluate its use. To assess all “teenage love” as the result of hormonal mayhem combines fallacies of composition and genesis. Similarly, to assess all religious understanding as a result of wishing to be saved or to have meaning and identity commits fallacies of composition and genesis. At least some teenagers may truly love; and some religious believers may truly seek, and perhaps find, something that is true about the transcendent.

2.  Understanding and Religious Understanding So, if all religious understanding and concepts cannot be reduced to illusions or delusions and shown to be false by their genesis, how can we understand them? First, fundamental to my view is that literal language is parasitic on figurative language.14 If you doubt this, listen to the present paper and the others 14 I cannot argue this point here. I have explored it for decades (for example, see my Story Theology [Wilmington, DE:  Michael Glazier, 1985] 1 et passim). Recent work in cognitive linguistics has offered an elegant account of how humans expand their understandings via figurative language. Here I adapt the work of Robert Masson, Without Metaphor, No Saving God: Theology after Cognitive Linguistics, Studies in Philosophical Theology 54 (Leuven: Peeters, 2014). Metaphorical language is an expression of a brain process, viz., cognitive mapping across disparate domains (the metaphoric process) that is “foundational and ubiquitous for [the] embodied mind” (Masson, 93). “Seeing is touching,” for example, joins two different sense domains. “The dollar is rising against the euro” invokes domains of both money and conflict in the context of a very conventional cognitive move that maps “rising” as “up” even though no pieces of currency need move a millimeter up for the sentence to be literally true in the domain of motion. Yet this latter sentence not only engages in cross-​domain transfers of meanings via the metaphoric process, but we also take the sentence as literally true or false, depending not on motion, but on activity in the currency markets, forgetting that “rising” was metaphorical and becomes literally true or false only when it is used in its new domain. If everyday language is littered with metaphor and everyday thinking includes imaginative cross-​ domain transfers of meaning, then a claim that religious and theological thinking is inferior because it does so as well is as bogus. As Robert Masson put it, “Religious and theological assertions, doctrines, analogies, and symbols can be metaphorical or figurative and at the same time can be semantically proper, logically warranted, and factually the case, in other words, can qualify as ‘literal’ truths” (Masson, 8). In other words, religious belief and language, qua religious, are at least analogous to other practices that generate particular uses of language and belief. Second, cross-​domain transfers can be “one-​way” or “two-​way.” “The dollar is rising” or “The temperature is rising” or “Her temper is rising” does not fundamentally change the sense of “rising.” If one understands “The balloon is rising” or “The sun is rising,” the other uses are easily understandable. They extend the meaning of “rising” to a new domain. They are “one-​way” shifts. Perhaps the “rising dollar” is only possible when currencies float against each other, rather than being tied to something like the gold standard. During the classical period of the gold standard,

Religious Understanding and Cultured Practices  259 in this conference and notice how many sentences speakers utter are both “literally” descriptive and require recognizing metonyms and metaphors—​ like “parasitic” in the previous sentence—​if we are to understand them. We easily—​almost naturally—​learn how to understand figurative language by utilizing a concept in a new domain. “Juliet is the sun,” as uttered by Romeo, is not about astronomy, but an application of an image in a new domain—​ what cognitive linguists label the “metaphorical process.” However, skeptics might go so far as to concede all this. The cultured despiser could point out that good historians and scientists operate with the presumption of naturalism, that is, that the transcendent is not a subject or agent in their academic practices. The gaps are closed and there is no need for

roughly 1870–​1914, the claim that the “dollar is rising” might be nonsense or false. The metaphor works when currencies float. This is a one-​way cross-​domain transfer. The concepts of “rising” and “falling” now blend into the economics of currency. The metaphoric process is one that enables us to extend the analogy or similitudes to a further conceptual domain. It seems to me that the notion that Jesus ransomed us from sin and death is a one-​way transfer. The concept of “ransom” is extended if Jesus’s crucifixion is the ransom the Father paid to the devil to free humanity. But the concept of “ransom” is not changed. Here is one of many examples of one-​way cross-​domain transfer as an expression of religious understanding. However, some such shifts in meaning create whole new ways (“tectonic shifts”) of conceptualizing the world and change the meanings of the concepts used. What Thomas S. Kuhn characterized as “scientific revolutions” can be seen as tectonic shifts in science (whether or not one accepts Kuhn’s full account or another like Pepper’s or Lakatos’s). The tectonic shift changes both the origin domain and the target domain. We see something similar in religious understanding. James Cone’s title The Cross and the Lynching Tree makes perfect sense as a paradigm of cross-​domain transfer. The identity claim that “Jesus is Messiah,” as understood by his early disciples, articulates a tectonic shift. In Second Temple Judaism, the Messiah was expected to be a political and religious victor. But Jesus was a Jewish prophet executed for treason. Identifying this man as “Messiah” shifts the significance of who Jesus is and what Messiah is. The meanings of “Jesus” and of “Messiah” shift from what they were for Second Temple Jews to what they meant for Second Temple Jews for Jesus. Perhaps the best illustration of this shift is in the “Caesarea Philippi” scene as assembled by the author of the Gospel of Mark, where Peter both confesses Jesus as the Anointed One—​Messiah or Christ—​(attributed by Jesus to divine inspiration), and is identified as “a Satan” later by Jesus for failing to understand what the Anointed One does. Peter got the title right, but its significance wrong since he had not realized that “Messiah” did not have the same meaning for Jesus’s followers and his opponents. It had changed its meaning in its new domain. Eventually monotheistic Jews saw Christians as blasphemers, and even as bitheists, for proclaiming Jesus as divine. Some Muslims later construed Christians as tritheists. What appears to Christians as Trinitarian monotheism is construed in other domains as something quite different. As with floating currencies, a host of practical and political factors are involved in creating new domains, but understanding that tectonic shifts in meaning occurred as constituents or results of such domain change helps us understand the confusions between those traditions of religious understanding. Key concepts have importantly different significance in their differing domains. In supporting religious believing as a phronetic practice, Zagzebski concludes that “persons with phronesis do not act by following a specifiable procedure, and I suggest that typically they do not form beliefs by following a specifiable argument” (“Religious Knowledge and the Virtues,” 223). If Masson is correct, then at least some religious believers do follow a roughly specifiable procedure in forming religious beliefs. But these procedures are not arguments, but a different doxastic practice.

260  Terrence W. Tilley the God hypothesis, even to explain religion. Why think we can responsibly develop beliefs that transcend the world of nature and history?15 By definition, whatever might be “transcendent” is to be found beyond our “ordinary” dimensions. How can we use the ordinary to understand the extraordinary? We use history, physical sciences, social sciences, and some humanistic disciplines to understand, explore, and even attempt to transform the four-​dimensional world of space-​time in which we live. How can we understand a dimension that transcends what we see as natural? A scene in Edwin Abbott’s brilliant 1884 satire, Flatland, provides a model. Flatland is a world with two and only two dimensions, length and breadth. Earlier in the book, the “author,” A Square, has a vision or dream of a conversation with a denizen of a one-​dimensional world, called “Lineland.” A Square’s conversation with the King of Lineland produces only incomprehension and anger in the King. The King is ready to kill the heretic from Flatland who insists that there are not one, but two, dimensions. But this dream sets up what A Square calls “the facts.” The key scene in this part is A Square’s discussion with his grandson. Insisting that geometry can only have two dimensions, A Square says this: I began to show the boy how a Point by moving through a length of three inches makes a Line of three inches, which may be represented by 3; and how a Line of three inches, moving parallel to itself through a length of three inches, makes a Square of three inches every way, which may be represented by 32. Upon this, my Grandson . . . took me up rather suddenly, and exclaimed, “Well then, if a Point by moving three inches, makes a line of three inches, represented by 3; and if a straight line of three inches, moving parallel to itself, makes a square of three inches every way, represented by 32; it must be that a Square of three inches every way, moving somehow parallel to itself (but I don’t see how) must make Something else (but I don’t see what) of three inches every way—​and this must be represented by 33.” “Go to bed,” said I, a little ruffled by this interruption: “if you would talk less nonsense, you would remember more sense.”16

Talking of a third dimension in two dimensional Flatland is nonsense, isn’t it? 15 Harvey, Feuerbach, recognizes that religious practices such as hymns and prayers reveal the content and significance of the practitioners’ faith (309). I would add other practices such as instructing, storytelling, and meditation to the list and note that the “illusions” thus produced are not delusions. That requires, as noted later, a different kind of appraisal. 16 Edwin A. Abbott, Flatland: A Romance of Many Dimensions, with illustrations by the Author, A  SQUARE, 6th ed., revised with an introduction by Banesh Hoffmann (New  York:  Dover Publications, 1952), 66.

Religious Understanding and Cultured Practices  261 A Square encounters what appears to him to be a circle. But the circle is actually a sphere whom A Square calls “the stranger.” A Sphere tries an argument much like the rejected grandson’s to no avail. He claims that Flatland is not really flat, but that its height is simply infinitesimal, but real. A Square is unconvinced. So A Sphere introduces A Square from Flatland to three-​dimensional Spaceland. A Square can see Flatland from above, spread out before him. Suffice it to say that A Square is both awed and baffled by what he sees. He returns to Flatland. When he tries to spread the gospel of three dimensions in Flatland, he is imprisoned. Having failed to convince the Council which judged him by his analogies and having failed to get them to accept his explanation of shifting circular phenomena as two-​dimensional representations of A Sphere moving and intersecting the plane of Flatland at different points, he is frustrated and distraught. He is imprisoned. He concludes his prison memoir seemingly hamstrung by his uncertainty and confusion about the reality of Spaceland.17 Is this the end of the story? I think not. Flatland illustrates what cognitive linguists call the metaphorical process, but to a transcendent domain. The grandson transfers arithmetical concepts of squares and cubes into the domain of geometry. Whereas A  Square at first could not conceive of spheres and cubes in geometry, eventually his “experience” with A  Sphere and the vision that he had forced him to think otherwise. What Flatland illustrates is that there is no good reason to think that the practice of extrapolating or analogizing from the dimensions we know to those we do not is inherently invalid. It is a process of understanding shared by religious believers, astrophysicists, and even Darwinians—​or do you think “selection” is not a transfer from the realm of human choice to evolutionary biology or “strings” from twine to physics? We frequently move from a known dimension to an unknown dimension by metaphorical processes, extrapolation, and analogy—​or, in the case of string theory, to eleven dimensions (something I simply cannot imagine). But, someone might say, yes advanced scientific theories are difficult to understand. But we have ways to appraise competing scientific theories even 17 “It is part of the martyrdom which I endure for the cause of Truth that there are seasons of mental weakness, when Cubes and Spheres flit away into the background of scarce-​possible existences; when the Land of Three Dimensions seems almost as visionary as the Land of One or None; nay, that when even this hard wall which bars me from my freedom, these very tablets on which I am writing, and all the substantial realities of Flatland itself, appear no better than the offspring of a diseased imagination, or the baseless fabric of a dream” (Abbott, Flatland, 103).

262  Terrence W. Tilley if we cannot distinguish them by purely empirical means. This hardly applies to religious beliefs. Religious practices are all over the place—​prayer, pilgrimage, confession, instruction, meditation, acts of charity, storytelling—​all of these are vehicles, sometimes contested, for developing religious understanding. We can weed out bad science. Can we discriminate truthful and obscurantist religious practice?

3.  Appraisals Many practices include a practice of appraising.18 The contents of appraisals are truth claims. Realtors appraise real estate, editors appraise writing, reviewers appraise arguments, consumers appraise merchandise, juries appraise guilt, athletic judges appraise performances. Each of these appraisals results in practical truth claims: “This house is worth $247,000,” “This manuscript will never be a book we can sell,” “This essayist has demonstrated her point convincingly,” “This shirt is not worth $22.95,” “He is guilty of involuntary manslaughter, not of murder,” “Her floor exercise is worth 9.7 points.” These claims may be formed by people well trained in appraisals (and thus their appraisals should be reliably formed unless other factors interfere), but that alone does not make those appraisals acceptable. All of our explicit truth claims, from simple to complex, can be construed as the linguistic expression of practical appraisals, from “This house is worthless,” to “This argument is valid,” to “God’s in heaven and all’s well in God’s world.” We appraise the results, often verbal, of practices. The project of appraising or assessing generally requires multiple standards. Real estate appraisers, for example, use multiple measures.

18 I have been developing an appraisal account of truth and truthfulness since my “Moments of Truth: The Standards of Truth and Narrative Theology,” Annual Meeting of the American Academy of Religion, Roundtable Session, December 1982, revised and published in Story Theology, 182–​ 218; also see Wisdom of Religious Commitment, 121–​154; Inventing Catholic Tradition (Maryknoll, NY: Orbis Books, 2000), 156–​170. This section is developed from the most recent appraisal account, Faith: What It Is and What It Isn’t (Maryknoll, NY: Orbis Books, 2010), 102–​128, where I develop a theory of faith as a relationship between a person or community and its ultimate center of meaning and value. In doing so, the criteria I developed there can also be applied to practices that are “anti-​ religious” or “unreligious” insofar as these are the root of our ultimate commitments, or, to use cognitive linguistics, provide the metaphors we use to represent all there is. I also argue there that there are forms of faith that are “polytheistic,” involving no “ultimate commitment,” but a set of “proximate, incommensurable commitments.” That the present text is limited to religious understanding and religious practice does not preclude the use of such criteria for appraising other metaphysical or anti-​ metaphysical claims or concepts of the transcendent (or the absence thereof).

Religious Understanding and Cultured Practices  263 They look at the recent sale prices of comparable properties. They evaluate other properties on the market. They look to pricing trends. They extrapolate market value from fair rental pricing. Appraising is not purely objective nor purely subjective. It is a judgment call given by a skilled practitioner whose judgment has been shaped by a thorough understanding of a practice. Something analogous can be said about understanding religious practices; we understand religious practices and can appraise the claims they generate. The practice of appraisal is not “neutral.” We do not have a “God’s-​eye view.” Rather, the practice of appraisal begins by using criteria that reflect the ways we use the term “true” in our everyday language. Nonetheless, we assess many claims by common sense. We assess others by the skilled use of criteria developed by experts in a practice. Each practice—​from investigating crimes, to appraising real estate, to solving linear equations, to psychological and physical diagnosis of clients and patients—​has developed distinct criteria for evaluating claims. The question is what are the conditions and criteria for appraising religious practice and belief when religious practices and beliefs differ so widely. People have faiths that were formed in the context of different faith traditions. A given Baptist may be a truly good Baptist. A given Buddhist may be a truly good Buddhist. Yet each has engaged in different practices and developed different beliefs. The situation of diversity means that being a good practitioner in a particular tradition is not a sufficient criterion for appraising religious practice and belief.19 Of course, a reductionist naturalist might say that science should judge. But that begs the question. It assumes the practice of scientific materialism as sufficient for understanding whatever there is needs no justification. That assumption is not sustainable.20 After all, even if all our concepts are human constructs, including our concepts of the transcendent, is it that religious practice gets religious people to “see” something that is 19 In this, I differentiate my view from those of reformed epistemologists, like Alvin Plantinga, who seem to think that religious beliefs properly formed in a particular religious practice, such as listening to God speak through the Bible, is sufficient for appraising them as warranted. Either a Zen Buddhist who practices meditation and achieves the insight into the universe characterized as satori is an epistemic peer of the Christian who reads the Bible—​in which case the situation of diversity as described here obtains and warrant in Plantinga’s sense is not sufficient for a full appraisal of these diverse beliefs and ways of understanding—​or the Buddhist is not the Christian’s epistemic peer and cannot have warranted beliefs as a non-​Christian—​a position for which I see no non-​question-​begging argument. See my “Reformed Epistemology in a Jamesian Perspective,” Horizons 19/​1 (Spring 1992), 84–​98; “Reformed Epistemology and Religious Fundamentalism: How Basic Are Our Basic Beliefs?” Modern Theology 6/​3 (April 1990), 237–​257; and my review of Alvin Plantinga, Warranted Christian Belief (New York: Oxford University Press, 2000), Theological Studies 62/​2 (June 2001), 388–​90. 20 See Ellis, “Religious Understanding.”

264  Terrence W. Tilley not there or that reductionist materialism inclines materialists to “missing” something that is there? That’s the question! We recognize that we appraise doxastic practices. We typically appraise the claims that these practices produce as an indirect way of appraising the practices. The question is what standards or criteria can we use to appraise religious practices? I find five criteria useful for appraising the claims and acts religious practices generate, all rooted in our ordinary uses of “true.”

3.1.  Revealing the Hidden; or, “Oh my God: It’s true!!!” A religious practice can be appraised as truthful if it enables the practitioners to understand the world in which we live, or a part of it, in a revealing way. Heidegger captured this notion of truth as revealing the hidden. He reflected on the ancient Greek word for truth, aletheia. The letter alpha that begins the word is an alpha privative. To tell the truth or to show the truth is to uncover what has been hidden. Parables—​stories that upset our world or part of it—​are designed to be revelatory. If they do not expose the cracks in the foundations of our world or oversights in our concepts, then they simply fail to do what they should. Nagasena’s chariot parable worked with King Menander to undermine his belief in a soul.21 Indeed, even if I do believe I have a soul, this parable uncovers the fact that we have to think long and hard about just what a soul is. Any metaphor may show us previously unnoticed connections among the events and people of the world in which we live. Such expressions can be evaluated as true by this standard of “uncovering the hidden.” Of course, this criterion cannot stand alone. That would unhappily equate the shocking, the surprising, or the revealing with the true! There’s more to truth than just unveiling. Moreover, sometimes shocking stories may be told in order to deceive the hearer, or shocking gestures—​like those of a magician—​may be performed to distract a viewer from what is there.22 21 “The Chariot Parable” from Questions of King Milinda (aka Menander), probably second century b.c.e., trans. John Bowker, Problems of Suffering in Religions of the World (Cambridge: Cambridge University Press, 1970), 242–​44. 22 Assessing a religious practice as being able to truly reveal what has been hidden is never final. Some expressions become so hackneyed that they no longer “uncover” something we overlooked. Some symbols also become hackneyed. Some images can no longer be used or used rightly—​if a Catholic did indeed pray to a statue, that statue, in becoming “opaque” to God for her or him, would fail to be revelatory. Practices can become deformed and thus distorting.

Religious Understanding and Cultured Practices  265

3.2.  Fittingness If the first criterion focuses on the effectiveness of a religious utterance to reveal, the second focuses on the content. An expression that fits well with other facts we recognize can be appraised as true. If a religious practice generates claims that do not fit what we know independently of it, then this counts against its being truthful. A religious practice can be called truthful, then, if (1)  the beliefs it develops are not inconsistent with the other facts we recognize; (2)  the beliefs it develops refer adequately; and (3) the beliefs attribute accurately. This criterion picks up central claims correspondence and coherence theories of truth.

3.2.1.  Cohering First, we can check a religious understanding to see if it can cohere with the other facts we recognize. The most obviously false story is the self-​deceptive one. Blind faith can be a form of self-​deception or the result of being deceived by someone you trust.23 Consider Ivan Karamazov. Ivan lives in a “Euclidean universe.” He cannot accept God’s “Euclidean” creation. Yet his Euclidean geometry cannot cohere with some facts he might recognize. In reference to Lobachevskian geometry, which he has refused to accept, Ivan says, “Let the parallel lines meet [at infinity] even before my own eyes: I shall look and say, yes, they meet, and still I will not accept it. That is my essence, Alyosha, that is my thesis.”24 Ivan’s acceptance of a Euclidean universe may be a common-​ sense view, but his refusal to accept evidence that would undermine his worldview marks Ivan’s understanding of science as a blind faith. The Brothers Karamazov portrays Ivan’s indefeasible scientific faith in a Euclidean world as a Freudian delusion. The results of scientific and historical investigations are often seen as challenges to faith. Taking them seriously can require folk to change their 23 To take a classic religious example, consider reading the Bible. A book neither interprets nor appraises itself. The Bible cannot say what the Bible means or appraise it as true. Appraisals and interpretations are acts; the Bible is a text or set of texts. We have to do the work of interpreting what it means, and whether God—​who can neither deceive nor be deceived in the Christian view—​speaks through it. To think it is self-​interpreting is at best confused, at worst self-​deceptive. 24 Fyodor Dostoevsky, The Brothers Karamazov, trans. Richard Peavey and Larissa Volokhonsky (New York: Farrar, Straus and Giroux, 1990), 206.

266  Terrence W. Tilley expressions of faith or to reformulate their reasons for holding their faith claims. Some forms of faith become implausible when those who hold them refuse to take seriously challenges from critical disciplines. Neither science nor history can be necessary and sufficient criteria for assessing the truth of a faith claim.25 But to ignore or refuse in some way to accommodate history and science undermines the truthfulness of religious practice and belief. A religious practice that generates claims that do not fit with other claims we recognize—​including scientific and historical claims—​cannot be appraised as “truthful.”

3.2.2.  Accuracy of Reference A second factor is the “accuracy of reference” of a faith claim. A more difficult issue usually involves a transcendent agent or the absence of any such agent. Did Moses lead the Exodus from Egypt? Or did God? Was Moses an agent empowered by God, who is really responsible? Or was Moses a powerful leader who led his people to wander around in the desert for forty years? Is the Buddha-​essence what produces good among people? Or is it something else? Is the God that Jews, Christians, and Muslims worship the creator and sustainer of the universe? Or is the material universe all that there is? In many instances, resolving disputed questions about transcendent actors is practically impossible. There is no direct evidence to determine whether we fail to see what is really going on or if we are imagining something that just is not there. If “accuracy of reference” were the only criterion for justifying faith, that would be a problem. But we do have other standards to use when we cannot apply this one. But notice that Ivan Karamazov remarks that the devil who appeared to him late in the novel was a projection of his imagination, yet says of the encounter, “That was no dream! No, I swear it was no dream, it all just happened!”26 Ivan’s evident inability to identify the agent that brought about his nightmare counts against his “scientific” practice. Limited as he is to science, anecdote, and history, he cannot refer clearly to the agent that tormented him. The incident seems to drive Ivan insane.

25 See Ellis, “Religious Understanding.”

26 Dostoevsky, The Brothers Karamazov, 650.

Religious Understanding and Cultured Practices  267

3.2.2.  Accuracy of Attribution Third, the more accurate an expression is in attributing a property or quality to something, the more reasonable it is to call it true. As accuracy of reference centered on naming the right agent, so accuracy of attribution centers on describing an agent, place, event, or thing accurately. To say that it is typically hotter in summer in Phoenix, Arizona, than it is in Washington, DC, is accurate. But to say that it feels hotter in Washington than in Phoenix may also be accurate (due to the humidity). To call the creator and sustainer of the universe “good” seems to fly in the face of the mixed good and evil character of the universe. Clearly, Ivan Karamazov in the great “Rebellion” chapter lays this out, and his brother Alyosha understands it. However, the vision of the whole novel shows how the world could be a paradise if everyone were “guilty” before all and were forgiven by all.27 Father Zosima and Alyosha Karamazov are willing to risk suffering (as Ivan is not). Such risk is reasonable if they are to have any possibility of alleviating suffering, solving problems, or reconciling enemies. Such reconciliation requires not merely being “guilty” before all, but responsible to and for all. They have to understand how they stand before God in order to understand the world as a sin-​damaged icon which humans can work to restore if they are “realists” like Alyosha rather than “materialists” like Ivan (these are Dostoevsky’s descriptors). The practical realist like Alyosha can see the world as God’s creation because he takes responsibility for it and acts to restore it, as Ivan cannot. His religious practice generates a coherent vision of world as a damaged icon. Obviously, Dostoevsky (despite claims to the contrary in the literature) values Alyosha’s realism over Ivan’s materialism. Alyosha does not refuse Ivan’s views, but sees a world that Ivan does not. In sum, coherence, accuracy of reference, and accuracy of attribution must all be taken into account before we can claim the beliefs we use to express our religious understanding. Religious understanding is not merely a matter of acquiring beliefs. It is also a way of relating to what is ultimate. Hence, we have to ask how “true” applies to relationship practices.28 27 Dostoevsky, The Brothers Karamazov, 290; also see Rowan Williams, Dostoevsky:  Language, Faith and Fiction (Waco, TX: Baylor University Press, 2008), 163. 28 Perhaps the gravest misattribution is to believe that some of the gods we (mistakenly) worship can be truly dependable sources of meaning and value—​of happiness—​in our lives. See David Foster Wallace (“Transcription of the 2005 Kenyon College—​May 21, 2005,” https://​web.ics.purdue.edu/​ ~drkelly/​DFWKenyonAddress2005.pdf, accessed 19 June 2016). Wallace offers no way to choose among the gods worthy of faith, a point to which we will return. He does, however, argue that the

268  Terrence W. Tilley

3.4.  Promoting Authenticity and Fidelity “Be true to yourself.” “Be true to the one you love.” A practice that fails to enable us to understand how to be “who we are” fails to enable us to be authentic and faithful, people of integrity and honesty. Religious understandings should be constituted in practices that enable communities or individuals to be “true to themselves” and “true to others.” The complement to being true to yourself is being “true to others.” We may be true to our lovers or we may be unfaithful. We may be true to our ideals or blow with the wind. We may be true to our gods or have little faith. If authenticity is not purely subjective, as argued previously, it involves us with others. Because we are not monadic individuals, authenticity is connected with fidelity. Being true to oneself is connected to being true to others.29 We sometimes forget that fidelity and authenticity are closely linked.30 It is easy to be true to one’s friends or one’s country. Some of us are captured by finite gods that will be consumed by the passage of time, the twists of fate, or the reality of death are not worth worshiping. Beauty fades with time, power and money are vulnerable to fate, and every nation fades, loses its way or is defeated in time. These gods have no power over life unless we choose to give it to them, and no power at all to save us from death. Perhaps those who understand divinities as finite (e.g., henotheists) should have considered Blaise Pascal’s wager argument. For his loss of love and real meaning shows one of the ways that worship of the finite gods cannot fulfill our real desires, even if they promise to fulfill our immediate desires. These are the gods of the rat race, not of real freedom. Or so Wallace proclaims. To understand what these finite gods really offer requires us to see if the myth that renders them divine is coherent. In short, it seems not: finite gods threaten to eat us alive. It seems inaccurate, and even foolhardy, to attribute power to them to make our lives whole and true; they are not worthy of worship. 29 Again, Dostoevsky provides an illustrative vignette. Gruschenka is narrating a folk tale to Alyosha: Once upon a time there was a woman, and she was wicked as wicked could be, and she died. And not one good deed was left behind her. The devils took her and threw her into the lake of fire. And her guardian angel stood thinking: what good deed of hers can I remember to tell God? Then he remembered and said to God: once she pulled up an onion and gave it to a beggar woman. And God answered: now take that same onion, hold it out to her in the lake, let her take hold of it, and pull, and if you pull her out of the lake she can go to paradise, but if the onion breaks, she can stay where she is. The angel ran to the woman and held out the onion to her: here, woman, he said, take hold of it and I’ll pull. And he began pulling carefully, and had almost pulled her all the way out, when other sinners in the lake saw her being pulled out and all began holding on to her so as to be pulled out with her. But the woman was wicked as wicked could be, and she began to kick them with her feet: “It’s me who’s getting pulled out, not you; it’s my onion, not yours.” No sooner did she say it than the onion broke. And the women fell back into the lake and is burning there to this day. And the angel wept and went away. (Dostoevsky, The Brothers Karamazov, 352) Like Ivan later in novel, who performed only one good deed (633) of picking up a peasant he had knocked down, the woman had not understood properly that she could grasp the onion in solidarity because she had not developed the habitus of fidelity that would enable her to understand the concept of being pulled out of the lake of punishment by a guardian angel and to receive her desired salvation. 30 For example, President John F. Kennedy famously said, “Ask not what your country can do for you. Ask what you can do for your country.” Although his remarks sound as if the first injunction was

Religious Understanding and Cultured Practices  269 the Peanuts character Linus van Pelt, who said, “I love humanity; it’s people I can’t stand.” The real test of fidelity is not merely how one deals with one’s friends, but with those one despises or opposes.31 The issue is how broadly a “golden rule” applies. How wide or how narrow is the range of one’s fidelity? In short, the greater the range of one’s fidelity, the better one meets this criterion of fidelity—​and, I would claim, but not argue here, that it measures authenticity as well. For a practice to promote authenticity, it has to promote the development of an authentic self, a self that is living a life worth living.32 The worship of the god of money enmeshes us in the art of the deal; commitment to allure places

opposed to the second, I think that reading is inaccurate. Rather, I’d say he was advocating the way to begin to find meaning in one’s life as a patriot. When asked what one can do for one’s country, one will find the country a source of meaning and value (perhaps irreducible, perhaps not). 31 Here the practice of nonviolence seems relevant. The practice of nonviolence has four characteristics. First, the nonviolent person or community is committed to not harming, maiming, or killing another person. Second, the nonviolent person wants to confront the opponent. To let the opponent continue practices that are harmful without resistance is not to deal with the other, but to ignore the other. Third, nonviolence is not the same as passivity. Nonviolent people can use demonstrations or strikes or other communal acts of resistance in order to confront the opponent. The purpose is to confront the opponent as persons in order to help them to see the evil effects of what they do. Fourth, the ultimate goal of nonviolence cannot be victory over an enemy, but the establishment of a community that if peaceful, fair, and just to all its members. Nonviolence requires sacrificing seeking control or domination of others because control and domination are committed to use violence if necessary. William James recognized the power of nonviolence. He used the term “non-​resistance,” but the point is the same. In The Varieties of Religious Experience, he discussed the power of saintly love: “Force destroys enemies; and the best that can be said for prudence is that it keeps whatever we already have in safety. But non-​resistance, when successful, turns enemies into friends; and charity regenerates its objects.” The point is not to say that a faith that is nonviolent at its heart is the only way to be truly faithful to others. But the practice of nonviolence exemplifies the real test of fidelity: What does the faith, the story, the symbol teach us about our fidelity not only to our friends, but also to our opponents and those we despise? Perhaps the most humiliating example of the failure of fidelity is the Christian tradition of anti-​ Semitism, captured in the accusation that “the Jews,” in executing Jesus, were “Christ-​killers.” The gospels say that some members who were leaders of the Jewish community in Roman-​occupied Palestine may have helped to bring about the execution of Jesus. However, many other fellow Jews revered Jesus—​not the least of whom were his disciples. The gospels and the Apostles’ Creed say that he was executed under Pontius Pilate. The obvious fact that the Romans executed Jesus should have made calling Jews “Christ-​killers” clearly unacceptable. But if fidelity is measured by how one treats one’s opponents and enemies, then the historic anti-​Semitism of the Christian tradition provides a good counter-​example. Repudiating that conviction and acting to rectify, insofar as possible, the horrible results of that view enables Christian communities to be more truly themselves. And it should be noted that most Christian churches have now come to know better—​they repudiate anti-​Semitism and grieve over their historic failures to be true to all God’s children as God is, as the faith they hold teaches them. 32 As in note 28, Wallace argued that devotion to some of the gods of our culture will destroy us. If that is so, is the self developed in the worship of money or pleasure or allure a self worth becoming? If you think so, then you will find that such a devotion truly promotes living an authentic human life. If you do not, then you must find that another form of faith will more truly promote living authentically (unless, of course, authenticity means nothing to you).

270  Terrence W. Tilley our center in the world of beauty that time will inevitably erode; henotheistic patriotism (“My country, right or wrong”) makes us prefer our nation (or its government or political party) and despise, ignore, or sacrifice whatever seems not of benefit to us. I would claim that secular humanism and universalist religious traditions link us to all the best in all humanity and thus grade out well on this criterion.

3.5.  Constancy in Seeking Truth A practice that shapes lives of creative truthfulness can be appraised as truthful. A person who is “true to her ideals” lives out a faith that can be appraised as more true than the person who has no ideals or who is an opportunist in practice. A faith that promotes the pursuit of truth, the uncovering of truth, and the telling of truth is a practice that encourages constancy. A truthful faith has seeking truth as an ideal. Gandhi’s concept of satyagraha as the power of truth expresses this criterion for assessing the truth of faith.33 Gandhi described satyagraha as involving renunciation, hard work, suffering, and perhaps even death in the service of seeking what is true. He described it as selfless, as concern with self is not primary. He found that a selfless search for truth meant that one could not wander off the path to truth very long. If one honestly keeps one’s eyes on the goal and ignores the difficulties encountered in seeking truth, the search is self-​correcting. If one stumbles off the path, the stumbling and difficulties will alert one to return to the right path. A faith that encourages truth seeking and truth telling can be assessed as truthful. This standard alone may not pick out some faiths that encourage people to seek “my truth” or “our truth” rather than “the truth.” But such faiths will not be ranked very high by other standards. The faith that was German Nazism, for example, pursued “our truth” of blood and soil. But even if someone were to find it encouraged truth seeking, it might not do so well on the standards of coherence, authenticity, and fidelity. The practice of universalist humanists who have faith in science also appraises well by this standard. The scientific virtues of honesty, carefulness, 33 Gandhi wrote: “Satyagraha is literally holding on to Truth, and it means, therefore, Truth-​force. Truth is soul or spirit. It is, therefore, known as soul-​force. It excludes the use of violence because man is not capable of knowing the absolute truth and, therefore, not competent to punish.” Mohandas K. Gandhi, Non-​violent Resistance (New York: Schocken Books, 1967), 3.

Religious Understanding and Cultured Practices  271 and tenacity are an exemplary form of constancy. Scientists seek the truth wherever it may be found. The results of their work are scrutinized by other scientists before they are accepted as true. Moreover, scientists know that they never have the final truth, so they must always continue to seek the truth. Those who seek what is true are also typically humble. To pursue truth means that one recognizes that one does not have the whole truth. One may be rightly proud of personal accomplishments. One may be satisfied that one has understood and communicated something previously unknown. But this criterion for assessing one’s pursuit of truth requires something like humility: “The more I learn, the more I realize how little I have learned.” Previously I claimed that some religious practices and the traditions that carry them were self-​correcting. A faith that has truth seeking as one of its goals can be recognized by the practice of self-​correction. Such a religious or quasi-​religious practice can be appraised as meeting the criterion of constancy. Like science, these faiths know that they must always seek what is true because they do not have it, at least not yet in its fullness. Even if we believe that God has revealed eternal and universal truths to humanity, those truths are necessarily expressed in language and other symbols that are temporal and particular. Our concepts cannot be eternal and universal because of historical changes and because the languages in which we express them are not. Hence, a faith tradition is truly constant only if it seeks the truth and is willing to rethink its inheritance in order to more fully express what is true, and is not constant if it clings to verbal formulae as if they were final truths.

4.  Conclusion By exploring the criteria for appraising expressions of faith, I have sought to show that, however contingent and disputable they may be, appraisals cannot be “purely” subjective. Appraisals are not arbitrary, capricious, or subject to no standards. Appraisals are judgments subject to the standards of practices, and good appraisers are people well-​trained in the practice of appraising. Religious practices include implicit criteria for appraisal and I have tried here to make some of them explicit as part of the process of religious understanding. Appraising religious understanding is necessarily contextual. That is, we appraise and assess claims, lives, and stories in a context and for a purpose. Appraisals can change. What has been accepted as a tolerable practice in the

272  Terrence W. Tilley past, such as just war-​making and capital punishment, may no longer be appraised as tolerable in the present. A claim that is accepted as true today might not have been accepted or appraised as true in the past, such as the claim that religious freedom is based in the dignity of the human person created by God—​a view rejected by the teaching authority of the Roman Catholic Church in the nineteenth century but then endorsed by the teaching authority of the Roman Catholic Church in the twentieth century. As cultural contexts change and as faith traditions change in response to the contexts in which they are inculturated, appraisals can change. Contextualism does not imply that “all truth is relative.” Such a claim is either unobjectionable or empty. If it means that claims are formed in particular times and places and appraisals of them are equally contextual, it is unobjectionable.34 If it means that the approach is inferior to a universalist approach, the claim is empty, for we cannot get to a non-​contextual context, an Archimedean platform, at least this side of the grave. I have claimed, then, that religious understanding does not necessarily produce illusions. Rather, it is one form of the metaphorical or imaginative practice that begins with imaginative constructions from the immanent to the transcendent, and that different patterns of religious practice and understanding can be appraised by standards that are implied by our ordinary uses of “true” which help us to appraise which practices are truthful.

References Abbott, Edwin A. Flatland:  A Romance of Many Dimensions, with Illustrations by the Author, A  SQUARE. 6th ed. Revised with introduction by Banesh Hoffmann. New York: Dover Publications, 1952. Alston, William P. Perceiving God:  The Epistemology of Religious Experience. Ithaca, NY: Cornell University Press, 1991.

34 My shaving mirror, a polished piece of brass, the Hubble telescope, and the tiny mirrors in video projectors, all, if they are well constructed, truly reflect reality in different ways. But all are used for different purposes in different contexts. What we will count as a true reflection will vary in each case. Hence, if one protests against contextualism as developed here as an inappropriate “relativism,” I would ask, “What can the alternative be?” A universal eye, God’s point of view, and a mirror for all occasions are inaccessible to us, just as concepts of reality independent of the faiths expressed in myths are, even if they can exist. Some will also find such a view fragile, unguaranteed by God or the world. But living faith traditions are fragile. Faith is risky. To have faith requires having the courage to risk commitment, knowing that it is possible that one’s pattern of faith in worship and sacrifice may be wrong. Coping with fragile traditions and risky faith does not require better epistemological foundations, but open-​ eyed courage tempered with thoughtful reflection on faith and faiths.

Religious Understanding and Cultured Practices  273 Bowker, John. Problems of Suffering in Religions of the World. Cambridge:  Cambridge University Press, 1970. Dostoevsky, Fyodor. The Brothers Karamazov. Translated by Richard Peavey and Larissa Volokhonsky. New York: Farrar, Straus and Giroux, 1990. Ellis, Fiona. “Religious Understanding, Naturalism and Desire.” In Making Sense of the World: New Essays in the Philosophy of Understanding, edited by Stephen R. Grimm. New York: Oxford University Press, 2017. Freud, Sigmund. The Future of an Illusion. Translated by James Strachey. New  York:  Norton, 1961. Gandhi, Mohandas K. Non-​violent Resistance. New York: Schocken Books, 1967. Harvey, Van A. Feuerbach and the Interpretation of Religion. Cambridge:  Cambridge University Press, 1995. Johnson, Richard C. Letter to the Editor. “Sunday Review.” New York Times, April 3, 2016. Masson, Robert. “Without Metaphor, No Saving God:  Theology after Cognitive Linguistics.” In Studies in Philosophical Theology, 54. Leuven: Peeters, 2014. Murphy, Nancey. Theology in the Age of Scientific Reasoning. Ithaca, NY: Cornell University Press, 1990. Plantinga, Alvin. Warranted Christian Belief. New  York, NY:  Oxford University Press, 2000. Ricoeur, Paul. The Symbolism of Evil. Translated by E. Buchanan. New York: Harper & Row, 1967. Tilley, Terrence W. “Moments of Truth: The Standards of Truth and Narrative Theology.” Roundtable session at the Annual Meeting of the American Academy of Religion, December 1982. Revised and published in Story Theology. Tilley, Terrence W. Story Theology. Wilmington, DE: Michael Glazier, 1985. Tilley, Terrence W. “Reformed Epistemology and Religious Fundamentalism: How Basic Are Our Basic Beliefs?” Modern Theology, 6 no. 3 (April 1990): 237–​257. Tilley, Terrence W. “Reformed Epistemology in a Jamesian Perspective.” Horizons 19, no. 1 (Spring 1992): 84–​98. Tilley, Terrence W. The Wisdom of Religious Commitment. Washington, DC: Georgetown University Press, 1995. Tilley, Terrence W. Inventing Catholic Tradition. Maryknoll, NY: Orbis Books, 2000. Tilley, Terrence W. Review of Alvin Plantinga, Warranted Christian Belief. In Theological Studies 62 no. 2 (June 2001): 388–​90. Tilley, Terrence W. History, Theology and Faith:  Dissolving the Modern Problematic. Maryknoll, NY: Orbis Books, 2004. Tilley, Terrence W. Faith: What It Is and What It Isn’t. Maryknoll, NY: Orbis Books, 2010. Wallace, David Foster. “Transcription of the 2005 Kenyon College Address.” May 21, 2005. Accessed June 19, 2016. https://​web.ics.purdue.edu/​~drkelly/​ DFWKenyonAddress2005.pdf. Williams, Rowan. Dostoevsky: Language, Faith and Fiction. Waco, TX: Baylor University Press, 2008. Zagzebski, Linda Trinkaus. “Religious Knowledge and the Virtues of the Mind.” In Rational Faith:  Catholic Responses to Reformed Epistemology, edited by Linda Zagzebski. Notre Dame, IN: University of Notre Dame Press, 1993. Zagzebski, Linda Trinkaus. The Virtues of the Mind:  An Inquiry into the Nature of Virtue and the Ethical Foundations of Knowledge. Cambridge: Cambridge University Press, 1996.

Index For the benefit of digital users, indexed terms that span two pages (e.g., 52–​53) may, on occasion, appear on only one of those pages. Note: References to figures are denoted by an italic f following the page number. References to notes are denoted by an n, followed by a period and the note number, following the page number.   Abbott, Edwin, 260–​61 Appelfeld, Aharon, 69–​70 abstraction, as explanatory virtue, appraising religious practices 232–​33,  242–​46 accuracy of attribution, 267 accuracy of attribution, in appraisal of religious accuracy of reference, 266 practices, 267 coherence,  265–​66 accuracy of reference, in appraisal of religious constancy in seeking truth, 270–​71 practices, 266, 267 fittingness, 265 aesthetic approach to literary understanding, general discussion, 271–​72 67–​68, 70–​71,  77–​80 overview,  262–​64 aesthetics promoting authenticity and fidelity, 268–​70 deferential judgment, appropriateness of, 117 revealing the hidden, 264 rational understanding of, 121 appreciation of reasons, in understanding of understanding why in, 118 people, 1 worth of secondhand judgment in, aptness, varieties of in virtue 113–​14,  115 epistemology, 111 affirmations, in virtue epistemology, 110–​11 apt success, Aristotelian, 109–​10 “After Great Pain, a Formal Feeling Comes” Aquinas, Thomas, 100, 135, 231 (Dickinson),  72–​73 “Archaic Torso of Apollo” (Rilke), 82 agency Aristotle, 74, 100, 109–​10, 125–​26, 231 children’s attributions of to robots, 152 Arnold, Matthew, 95–​96, 102, 103 perceived, effect on children’s learning, 153–​ articulable knowledge, epistemic basis 57, 154f, 155f of,  118–​21 and trust of children in teachers, 151 Asselin, J., 245 Aibo robot dog, 144 associative thinking. See characterizations; alethic affirmations, in virtue epistemology, frames; perspectives 110, 111 Atkins, Peter, 1–​2n.4 Alexander, Stephon, 132–​33, 134 attribution, accuracy of, in appraisal of religious Alter, Adam L., 219 practices, 267 analogical discovery, repeating structures Attridge, Dick, 77 and,  131–​34 Augustine, 100 analogical equations, 31–​33 Austen, Jane, 102 androids, 149 (see also robots, children’s authenticity, in appraisal of religious understanding of) practices,  268–​70 antecedent fact, worth of secondhand judgment authority, intellectual, in sciences versus in questions of, 113–​14 humanities,  61–​63 anti-​Semitism, Christian tradition of,   268–​69n.31 base-​level judgments, perspectives as antisocial behaviors, impact of robots influencing,  30–​31 on,  161–​62 Beardsley, Monroe, 87

276 Index Bechlivanidis, C., 242, 243 beliefs. See also religious understanding defining, 24–​25n.2 true, and understanding as grasp of structure,  126–​31 Berman, Russell, 77, 78n.45 Bible, interpretation of, 265n.23 biology, subjectivity in, 104–​5, See also science(s) Blanchot, Maurice, 77–​78 blind faith, 265 Bloch, Marc, 57–​58 Bonawitz, E. B., 237 Booth, Wayne, 74 Boyd, Richard, 32 breadth of explanations, 232–​33, 242–​44 Brink, Kimberly A., 8, 9 Brothers Karamazov, The (Dostoevsky), 265, 266, 267, 268n.29 Buridan, Jean, 101   Callahan, Laura, 10–​11n.11 Camp, Elisabeth, 11 Carey, Susan, 215 Carnap, Rudolf, 36 Cartwright, N., 233 categorical explanations, 244–​46 causal ascriptions, mechanism independence of,  218–​19 causal centrality, 20–​21, 202, 204 causal complexity, mechanistic understanding and sense of, 200 causal dependence relations, 210–​11 causal explanations children’s preference for, 197–​99 mechanistic and functional explanations as,  214–​16 causality and coherence of explanations, 240 as fundamental element of explanations, 233 kinds of, as cognitive trace from exposure to mechanism, 202 and preference for simple explanations,  235–​39 causal potency, 202 causal processes, normative evaluations of, 224 causal reasoning, in social learning, 172–​73 causal relevance, 202, 204 causal schematic expectations, 202 causal structures desire to grasp, 130–​31 functional understanding’s demands on,  225–​26

Cave, Terence, 86 centrality causal, 20–​21, 202, 204 as dimension of mattering, 20–​21 and ultimate characterization, 38 chance, success by, in Aristotelian ethics,  109–​10 characterizations. See also frames; perspectives versus concepts, 19–​23, 28–​29 contextual malleability of, 21–​22 defined,  18–​19 epistemic status of, 28–​29 as holistic affair, 23–​24 multidimensional structure of, 20–​21 perspectives and, 24–​26 reflective endorsement of, 22–​23 ultimate,  37–​41 voluntary control of, 23 children. See also robots, children’s understanding of; social learning as communicators of information, 182–​84 as evaluators of others’ informativeness, 177–​82,  179f as interpreters of information provided by others, 172–​77, 174f intuitions about mechanistic complexity, 200 preference for mechanistic information, 7,  197–​99 preference for simple explanations, 237 Christianity. See also religious understanding cross-​domain transfers of meanings in, 258–​59n.14 tradition of anti-​Semitism in, 268–​69n.31 Cicero, 100 Citizen (Rankine), 82–​84 cognitive approach to literary understanding, 67–​68, 70–​71,  81–​87 cognitive development. See also robots, children’s understanding of; social learning cognitive ecology, literary work as part of, 86 cognitively privileged, functional explanations as, 220 cognitive resources strong differentiation thesis, 6–​7 and varieties of knowledge, 3–​4 and varieties of understanding, 4–​5 cognitive traces from exposure to mechanism,  200–​4 coherence in appraisal of religious practices, 265–​66,  267 as explanatory virtue, 232, 239–​41

Index  277 Coll, Richard K., 17–​18 collective knowledge, in sciences versus humanities, 48, 56–​58 Collingwood, R. G., 1 common cause explanations, preference for,  236–​37 communication, frames as tool for, 17–​18, 29 communicative contexts, studying social learning in, 169–​72, 171f, See also social learning communicators of information, children as,  182–​84 community epistemic obligations in, 10–​11 role in categorical explanation, 245–​46 competence as ability to assess testimony of experts,  115–​16 in Aristotelian apt success, 109–​10 expertise as variety of, 111 complexity assumptions about in sciences versus humanities,  63–​65 of explanations, preferences regarding, 7,  234–​35 intuitions about mechanistic, 200 limitations imposed on science by, 103–​4 component parts, understanding in terms of. See mechanistic understanding computational models of pedagogical reasoning, 170–​71, 175, 180, 181 conceptions, 19, See also characterizations concepts, versus characterizations, 19–​23,  28–​29 conceptual taxonomies, 30, 36–​37 concrete explanations, 242–​43 Cone, James, 258–​59n.14 consciousness, as impregnable to science, 98–​99,  104–​7 consensus, in sciences versus humanities, 48,  56–​58 consequence etiology, 214–​15 constancy in seeking truth, 270–​71 contextualism, and religious understanding,  271–​72 contextualization in humanities research, 50, 52 contextual malleability of characterizations,  21–​22 convention, worth of secondhand judgment in questions of, 113–​14 cooperative principles, Grice’s, 178–​79

cost-​benefit analysis of information transfer, 182–​83,  184–​86 counterfactual dependence, causal ascriptions based on, 218–​19 Craver, C. F., 233 creepiness of humanlike robots, 9, 144–​46, 145f, 146f, 147f,  147–​49 crises of faith, 256–​57 Cross and the Lynching Tree, The (Cone), 258–​59n.14 cross-​domain transfers of meanings, 258–​59n.14,  259–​61 Currie, Gregory, 68–​69   Darden, L., 233 decay-​neglect phenomenon,  194–​95 decline of knowledge, 194–​95 deep differences in dependence relations, 6–​7 and varieties of understanding, 2–​5 deference to others, 9–​11 deferential judgment, appropriateness of,  116–​17 in humanities, 57–​58 and utility of cognitive traces from exposure to mechanism, 204 worth of secondhand judgment, 113–​16 delusions, religious beliefs as, 255 Dennett, Daniel, 212, 216–​17 dependence relations, understanding as representing, 5–​7, 210–​11, 213, 221 Derrida, Jacques, 69–​70n.13 Descartes, René, 10, 100 descriptive claims, in sciences versus humanities, 48, 54–​56 design stance, 212–​14, 216–​17 detailed explanations, 242–​44 developmental trajectory of children’s understanding of robots, 142–​44,  158 of perceived agency on learning, 156–​57 of uncanny valley, 147f,  147–​49 diagnosticity, in prominence, 20, 37–​38 Dickens, Charles, 74, 75–​76 Dickinson, Emily, 72–​73 Dictionary (Johnson), 101 difference makers, in explanations, 242–​43 differences, inquiry focused on, 63–​64 Dillingham, E. M., 244–​45n.2, 245 Dilthey, Wilhelm, 1 dispensability of frames and perspectives,  34–​37 divinities, finite, 267n.28

278 Index DiYanni, C., 215 domains cross-​domain transfers of meanings, 258–​59n.14,  259–​61 grasp of structures repeating in different,  131–​34 understanding of different structures in same,  129–​31 domain theories, modes of construal and, 212 Dostoevsky, Fyodor, 265, 266, 267, 268n.29   early childhood. See robots, children’s understanding of; social learning ECHO computational model, 239–​40 ecology, cognitive, 86 education. See also social learning related to grasping structure, 134 social robots, role in, 150–​57 Elgin, Catherine, 39n.4, 125 Ellis, Fiona, 254n.3 emotions about robots, children’s, 144–​50, 147f role in literary understanding, 75 empathy, literary, 74, 75 empirical literary cognitivism, 81 empirical psychology, 105 end of inquiry, perspectives at end of, 33–​37 epistemic basis of knowledge, inarticulacy and,  118–​21 epistemic deference. See deference to others epistemic obligations in communities, 10–​11 epistemologies of humanities and sciences assumptions about simplicity and complexity,  63–​65 indexical versus non-​indexical claims, 47,  49–​51 individual versus collective focus, 48, 56–​58 intellectual authority, role of, 61–​63 intellectual progress, notions of, 59–​60 natural endpoint to inquiry, 58–​59 overview,  47–​49 perspectival versus non-​perspectival insights, 47–​48,  51–​54 prescriptive versus descriptive claims, 48,  54–​56 epistemology firsthand understanding, 9–​11 frames, status of, 28–​33 literary understanding, 11–​12 perspectives, at end of inquiry, 33–​37 psychology of understanding, 4–​9 reformed, 263–​64n.19 religious understanding, 11, 12

secondhand understanding, 8–​9, 10 ultimate characterization, 37–​41 understanding as representing (explanatory) dependence,  210–​11 varieties of understanding, 2–​5 virtue,  110–​13 essentially secondhand knowledge, 115–​16 ethics, Aristotelian, 109–​10 evaluation of explanations abstraction, 232–​33,  242–​46 coherence, 232, 239–​41 explanatory virtues in philosophy, 231–​33 general discussion, 246–​47 overview, 231 simplicity, 231–​32,  234–​39 unification,  242–​46 evaluators of others’ informativeness, children as, 177–​82, 179f evolutionary psychology, 97 experiential dimension of literary understanding, 67–​68,  75–​77 expert authority, in sciences versus humanities,  61–​63 expertise, in virtue epistemology, 111 explaining, understanding versus, 196–​97 explanation choice paradigm, 198–​99 explanations. See also evaluation of explanations breadth of, 232–​33, 242–​44 categorical,  244–​46 causal, 197–​99,  214–​16 complex versus simple, 7 concrete,  242–​43 detailed,  242–​44 formal, 244–​45n.2 functional, 212, 214–​21 hyperconcrete, 232, 242–​43 mechanistic, 212, 214–​21, 233, 237–​39 modes of construal and, 212 as object of knowledge constituting understanding, 211 probable,  236–​37 simple, 7 explanatory dependence relations, understanding as representing, 210–​11, 213, 221 explanatory depth, illusion of, 194–​95, 219 explanatory interests related to conceptual taxonomies,  36–​37 explanatory knowledge, decline of, 194–​95 expressive dimension of literary understanding, 67–​68, 86–​87, See also cognitive approach to literary understanding

Index  279 faith. See religious understanding Faith (Tilley), 262n.18 feature introduction by system completion,  31–​33 feelings role in literary understanding, 75 toward robots, children’s, 144–​50, 147f Fernbach, P. M., 243 fiction. See literary understanding fidelity, in appraisal of religious practices,  268–​70 figurative language, literal language as parasitic on, 258–​59, 258–​59n.14 finite gods, 267n.28 firsthand knowledge and understanding, 9–​11 Aristotelian apt success, 109–​10 deferential judgment, 116–​17 overview, 109 rational understanding, 121 versus secondhand knowledge, 111–​13 understanding and inarticulacy, 118–​21 understanding why, 117–​18 virtue epistemology, 110–​13 worth of secondhand judgment, 113–​16 first-​personal understanding, 225 fittingness, in appraisal of religious practices, 265 Flatland (Abbott), 260–​61 formal explanation, 244–​45n.2 forms, grasp of, 125–​26, See also structure, understanding as grasp of frames characterizations, defining, 18–​24 definitions related to, 18–​19 dispensability of, 34–​37 as double-​edged swords, 18 as expressing perspectives, 27–​28 general discussion, 41–​42 generative,  32–​33 as guiding search for information, 31–​33 as instruments for inquiry, 28–​33 in intersubjective understanding, 29 overview,  17–​18 perspectives, defining, 24–​26 perspectives at end of inquiry, 33–​37 programmatic research-​orienting role of,  32–​33 in self-​understanding, 29 as tools for thought, 29–​31 ultimate characterization, 37–​41 Freud, Sigmund, 254–​56, 256–​57n.11, 257 full understanding, 191–​94 functional explanations, 212, 214–​21

functional mode of construal empirical evidence for, 214–​21 as path to understanding, 212–​14 functional understanding, 5–​7 empirical evidence for, 214–​21 general discussion, 226 overview,  209–​10 stances or modes of construal as paths to understanding,  212–​14 strong differentiation thesis, 209, 223–​26 understanding as representing (explanatory) dependence, 210–​11, 213, 221 weak differentiation thesis, 209, 221–​23 Future of an Illusion, The (Freud), 254   Gadamer, Hans-​Georg, 69–​70n.13 Galilei, Galileo, 95 Gandhi, Mohandas K., 270 Garfinkel, A., 232 generalization based on functional explanations, 217–​18 in sciences versus humanities, 47, 49–​51 generative dimension of literary understanding, 67–​68,  78–​80 genetic fallacy, 257–​58 Geography of Insight, The (Foley), 47, 65 Gergely, György, 219 gestalt perception, 23–​24, 39 Gibson, John, 77 Giffin, C., 245 Glennan, S. S., 233 goal-​directed  action children’s inferences related to, 173–​75, 174f infants’ perception of, 219 goals, role in evaluation of explanations, 246–​ 47, See also functional understanding God, belief in, 253–​54, See also religious understanding gods, finite, 267n.28 golden ratio, 123–​24, 132 Gosetti-​Ferencei, Jennifer,  11–​12 grasping, in simple view of understanding, 211 Grice, H. P., 178–​79 Grimm, Stephen, 1–​2n.2, 225 Gweon, Hyowon, 8 Gwynne, Nicholas Z., 217–​18   Harper, Kyle, 130–​31 Harvey, Van A., 253–​54n.2, 259–​60n.15 Hastie, R., 241 Hawking, Stephen, 93–​94 Heidegger, Martin, 69–​70n.13, 84–​85, 264 Hemmatian, B., 245–​46

280 Index hidden, revealing, in appraisal of religious practices, 264 higher-​order judgments and characterizations, 39–​41 perspectives as influencing, 30–​31 Hills, Alison, 41, 210n.1 Hirsch, E.D., 75 historians challenges to faith caused by, 265–​66 indexical claims by, 50 interest in causal structures, 130–​31 Hobbes, Thomas, 101 Hopkins, E. J., 238 humanism dispute between naturalism and, 1–​2 varieties of understanding, 2–​5 humanities. See also firsthand knowledge and understanding assumptions about simplicity and complexity,  63–​65 debate over scientism, 94–​99 indexical versus non-​indexical claims, 47,  49–​51 individual versus collective focus, 48, 54–​56 intellectual authority, role in, 59–​60 intellectual progress, notions of, 59–​60 limitations of sciences versus, 103–​4 natural endpoint to inquiry in, 58–​59 perspectival versus non-​perspectival insights, 47–​48,  51–​54 prescriptive versus descriptive claims, 48,  54–​56 versus sciences, overview of, 47–​49 humanlike robots. See also robots, children’s understanding of likeliness of children to learn from, 152–​53 in special populations, need for research on, 159 terminology related to, 149–​50 uncanniness of, 144–​46, 145f, 146f,  147–​49 human social learning. See social learning Hume, David, 224 humility, and constancy in seeking truth, 271 Huxley, T. H., 95–​96 hyperconcrete explanations, 232, 242–​43   ideal theory, 35–​37 illusion of explanatory depth (IOED), 194–​95,  219 illusions, idea of religious beliefs as, 254–​57 imagined listening (perceptual mimesis), 12, 72 imitation, in social learning, 172–​73

inarticulacy, understanding and, 118–​21 incomplete understandings. See partial understandings indexical claims, in sciences versus humanities, 47,  49–​51 individual insights, in sciences versus humanities, 48, 56–​58 infants, perception of goal-​directed action by, 219, See also children; robots, children’s understanding of; social learning inferences relation to modes of construal, 220 requiring causal reasoning, 240 in social learning, 168–​69, 172–​84, 174f (see also social learning) supporting different kinds of, 222 information children as communicators of, 182–​84 mechanistic, children’s preference for, 7,  197–​99 informativeness, children as evaluators of others’, 177–​82, 179f Ingarden, Roman, 78 inquiry frames as instruments for, 28–​33 natural endpoint to, in sciences versus humanities,  58–​59 perspectives at end of, 33–​37 insight, interchangeable terms for, 48–​49, See also understanding instrumental value of frames and perspectives,  34–​37 intellectual authority, in sciences versus humanities,  61–​63 intellectual progress, in sciences versus humanities,  59–​60 intensity, in prominence, 20, 37–​38 intentional fallacy, 87 intentionality, in cognitive approach to literary understanding,  86–​87 intentional stance. See functional understanding interactive robot-​teaching styles, 156–​57 internal complexity, inferences about, 200 interpretation, literary. See literary understanding interpreters of information provided by others, children as, 172–​77, 174f interpretive frames. See frames intersubjective understanding, frames as contributing to, 29 intuitive theories, 213–​14

Index  281 intuitive thinking. See characterizations; frames; perspectives IOED (illusion of explanatory depth), 194–​95,  219 iPal robot, 139–​40, 140f irrational religious beliefs, presumption of,  254–​57 irrelevant details, in explanations, 238–​39,  242–​43 Iser, Wolfgang, 78–​80   Jackson, Frank, 105 James, Henry, 74, 75–​76 James, William, 216, 268–​69n.31 Jesus anti-​Semitism based on death of, 268–​69n.31 cross-​domain transfers of meanings related to, 258–​59n.14 Jibo robot, 139–​40, 140f Johnson, Richard C., 253, 254–​55, 257 Johnson, Samuel, 101 journey, faith as, 256–​57 Judaism, cross-​domain transfers of meanings in, 258–​59n.14 judgments, in virtue epistemology, 111 just-​so stories, 17, 27, 31   Kafka, Franz, 85–​86 Kaspar robot, 144–​46, 146f, 147f Keil, Frank, 7–​8, 212 Kelemen, Deborah, 215, 220 Keller, Evelyn Fox, 32 Kennedy, John F., 268–​69n.30 Kepler, Johannes, 132–​33 Khemlani, S. S., 235–​36 knowledge. See also firsthand knowledge and understanding; secondhand knowledge and understanding; understanding accounts of understanding based on, 210–​11 decline of, 194–​95 interchangeable terms for, 48–​49 relationship between literature and, 68–​70 seeking understanding, 10–​11 as special case of understanding, 128–​31 varieties of, 3–​4 Köppe, Tilman, 68–​69n.10 Krioukov, Dmitri, 132 Kuhn, Thomas S., 258–​59n.14 Kvanvig, Jonathan, 39, 40–​41   Lakoff, George, 17–​18 Lamarque, Peter, 81–​82

language in cognitive approach to literary understanding,  84–​85 and religious understanding, 12, 258–​59n.14 large scale schematic patterns, 202 Laws, The (Plato), 95 learning, social. See social learning learning from robots developmental trajectory of perceived agency on,  156–​57 impact of children’s understanding of robots on, 153–​56, 154f, 155f Leavis, F. R., 96 legal decision-​making, coherence in, 241 Leibniz, G. W., 95 Lindemann, Frederick, 93 linguistic articulation, understanding and,  118–​21 Liquin, Emily, 219–​20 listening, imagined, 12, 72 literalization of metaphors, 34–​35 literal language, as parasitic on figurative language, 258–​59, 258–​59n.14 literal meaning, in literary understanding, 71 literary cognitivism, 67–​68, 70–​71, 81–​87 literary empathy, 74, 75 literary understanding, 11–​12 aesthetic approach to, 67–​68, 70–​71, 77–​80 cognitive approach to, 67–​68, 70–​71, 81–​87 elements of, 71–​73 experiential dimension, 67–​68, 75–​77 expressive dimension, 67–​68, 86–​87 general discussion, 68–​73 generative dimension, 67–​68, 78–​80 moral approach to, 67–​68, 74–​77, 84 overview, 67–​68, 88 literature defining, 68 relationship between knowledge and, 68–​70 relationship between understanding and,  70–​71 Locke, John, 10, 18, 101, 111–​12n.1 Lombrozo, Tania, 5–​7, 215, 217–​20, 222, 236–​37,  245   Machamer, P., 233 machine-​like robots, children’s perceptions of,  147–​48 Marcus-​Newhall, A., 236 Masson, Robert, 258–​59n.14 mattering, in characterization, 20–​21 meaning, in literary understanding, 71–​73, 75

282 Index mechanism independence of functional explanations, 215–​20,  225–​26 mechanistic explanations, 212, 214–​21, 233,  237–​39 mechanistic/​physical mode of construal empirical evidence for, 214–​21 as path to understanding, 212–​14 mechanistic understanding, 5–​8 children’s preference for mechanistic information,  197–​99 cognitive traces from exposure to mechanism,  200–​4 empirical evidence for, 214–​21 general discussion, 226 intuitions about mechanistic complexity, 200 overview, 192–​94,  209–​10 stances or modes of construal, 212–​14 strong differentiation thesis, 209, 223–​26 understanding as representing dependence, 210–​11, 213, 221 weak differentiation thesis, 209, 221–​23 mental abilities, children’s perceptions of in robots, 148, 157–​58 mental models, 193–​94 Merleau-​Ponty, Maurice, 80 Messiah, Jesus as, 258–​59n.14 Metamorphosis, The (Kafka), 85–​86 metaphorical process, 258–​61, 258–​59n.14 metaphors frames, 17, 31, 32, 34–​35 in literary understanding, 72–​73, 80, 85–​86 mimetic function of literature, 79–​80 mind(s) children’s perceptions of in robots, 148,  157–​58 understanding other, 167–​68, 176–​77, 184–​ 85 (see also social learning) mirror neurons, 4–​5n.9 modal implications of mechanistic and functional understanding, 225–​26 modes of construal, as paths to understanding,  212–​14 moral approach to literary understanding, 67–​ 68, 70–​71, 74–​77, 84 moral development, impact of robots on,  161–​62 morality deferential judgment, appropriateness of, 117 firsthand understanding in, 9–​11 understanding why in, 118 worth of secondhand judgment in, 113–​15 mores knowledge, 120–​21 Mrs. Dalloway (Woolf), 11–​12, 71–​73, 76–​77

multidimensional structure of characterizations,  20–​21 music, analogical reasoning related to, 132–​33 mutability,  20–​21   Nagel, Thomas, 53, 98–​99, 105, 106–​7 Nao robots, 144, 147f, 153–​54, 159 National Robotics Initiative, 162 natural endpoint to inquiry, 58–​59 naturalism dispute between humanism and, 1–​2 presumption of, and understanding of transcendent,  259–​60 varieties of understanding, 2–​5 natural pedagogy (trust in testimony), 140–​41, 151–​56, 154f, 155f natural sciences. See science(s) nature, organizational structure of, 132–​35 Newton, Isaac, 231 Nicomachean Ethics (Aristotle), 109–​10 Noe, Alva, 84–​85 non-​indexical claims, in sciences versus humanities, 47, 49–​51 non-​perspectival insights, in sciences versus humanities, 47–​48,  51–​54 non-​propositional nature of perspectives,  24–​26 non-​propositional structure,  127–​31 nonviolence, 268–​69n.31 normative considerations of mechanistic and functional understanding, 224–​25 normative-​evaluative aspect of intentional stance,  6–​7 Notebooks of Malte Laurids Brigge, The (Rilke), 82 Nowak, L., 233 Nussbaum, Martha, 74, 75   object of cognition, deep differences in, 4–​5 objects of understanding dependence relations as, 5–​7 of mechanistic versus functional understanding, 6, 209, 221–​23 object structure, understanding of. See structure, understanding as grasp of Ockham, William of, 101 omission, sins of, in pedagogical context, 177–​81,  179f one-​way cross-​domain transfers, 258–​59n.14 open-​ended nature of perspectives, 24–​26 Oppenheimer, D. M., 219, 235–​36 others, understanding, 167–​68, 176–​77, 184–​ 85, See also social learning

Index  283 Pacer, M., 237 parables, 264 partial understandings children’s preference for mechanistic information,  197–​99 cognitive traces from exposure to mechanism,  200–​4 decline of knowledge, 194–​95 general discussion, 206 intuitions about mechanistic complexity, 200 overview,  191–​94 understanding versus explaining, 196–​97 utility of cognitive traces, 204 Pascal, Blaise, 267n.28 pedagogical contexts, studying social learning in, 169–​72, See also social learning pedagogical discourse, frames in, 17–​18 pedagogical reasoning, 170–​71 children as communicators of information,  182–​84 children as evaluators of others’ informativeness, 177–​82, 179f children as interpreters of information provided by others, 175–​77 Pember, Margaret, 93 Pennington, N., 241 Pepperberg, Irene, 103–​4 perceptual mimesis (imagined listening), 12, 72 perspectival insights, in sciences versus humanities, 47–​48,  51–​54 perspectives. See also characterizations; frames in cognitive approach to literary understanding,  86–​87 defined,  18–​19 dispensability of, 34–​37 at end of inquiry, 33–​37 epistemic status of, 28–​29 frames as expressing, 27–​28 in literary understanding, 11–​12, 71–​73 and mechanistic versus functional understanding,  224–​25 non-​propositional nature of, 24–​26 open-​ended nature of, 24–​26 in scientific accounts, 105 as tools for thought, 29–​31 Phaedo (Plato), 94 phenomenology of mechanistic and functional understanding, 222 philosophy. See also specific areas of philosophical research on understanding deferential judgment, appropriateness of, 117 explanatory virtues in, 231–​33 non-​indexical focus of, 49–​50

prejudice of science against, 93–​94 understanding as representing (explanatory) dependence,  210–​11 understanding why in, 118 union with reality in, 135 worth of secondhand judgment in, 115 phronetic practice, religious believing as, 258–​59n.14 physical mode of construal empirical evidence for, 214–​21 as path to understanding, 212–​14 physical sciences. See science(s) Pinker, Steven, 96–​97 Pippin, Robert, 74 planetary motion, Kepler’s laws of, 132–​33 Plantinga, Alvin, 263–​64n.19 Plato, 68–​69, 94–​95, 100 plausibility, in story model, 241 poetry. See literary understanding politics deferential judgment, appropriateness of, 117 frames in, 17–​18 understanding why in, 118 Pólya, George, 34 Posterior Analytics (Aristotle), 100 potency, causal, 202 powers of mind strong differentiation thesis, 6–​7 and varieties of knowledge, 3–​4 and varieties of understanding, 4–​5 practices, religious, 12, 254–​57, 259–​60n.15, See also appraising religious practices; religious understanding pragmatic implicature, 178–​80 Prasada, S., 244–​45n.2, 245 prescriptive claims, in sciences versus humanities, 48, 54–​56 Pride and Prejudice (Austen), 102 probabilistic inferences, in social learning, 173–​75,  174f probable explanations, preference for, 236–​37 Problems of Philosophy, The (Russell), 135 programmatic research-​orienting role of frames,  32–​33 progress, notions of in sciences versus humanities,  59–​60 projection theories of religion overview,  253–​54 problems with, 254–​58 prominence as dimension of mattering, 20, 21 and ultimate characterization, 37–​38

284 Index propositional attitudes, perspectives versus,  24–​26 propositional structure, 126–​27, 128–​31 prose. See literary understanding proximal causes, understanding in terms of. See mechanistic understanding psychological agency children’s attributions of to robots, 152 perceived, effect of on children’s learning, 153–​57, 154f, 155f and trust of children in teachers, 151 psychologically privileged, functional explanations as, 220 psychology. See also specific areas of psychological research on understanding empirical, 105 evolutionary, 97 of understanding, 4–​9 public affirmation, in virtue epistemology, 110 pure abstraction, 244–​46 purposes, understanding in terms of. See functional understanding Putnam, H., 245 Pythagoras,  123–​24   Railton, P., 233 Ramsification, 32 Rankine, Claudia, 82–​85 ransom concept, 258–​59n.14 rational inference in social context, 172–​84,  174f rational understanding, questions calling for, 121 Read, S. J., 236 realist fiction, and literary understanding, 74,  75–​77 reality in aesthetic approach to literary understanding,  77–​80 in cognitive approach to literary understanding, 84 in moral approach to literary understanding, 84 unity of all, 134–​35 reasoning, pedagogical. See pedagogical reasoning; social learning reasons, in understanding of human actions, 1 reference, accuracy of, in appraisal of religious practices, 266, 267 reflective endorsement of concepts versus characterizations,  22–​23 reformed epistemology, 263–​64n.19 relativism, and religious understanding, 272

relearning, 200, 204 relevance, causal, 202, 204 religious understanding, 11, 12 accuracy of attribution, 267 accuracy of reference, 266 appraising,  262–​71 coherence,  265–​66 constancy in seeking truth, 270–​71 correction of beliefs over time, 255–​57 fittingness, 265 general discussion, 271–​72 genetic fallacy, 257–​58 overview,  253–​54 presumptions of irrationality related to,  254–​57 problems with projection theories, 254–​58 promoting authenticity and fidelity, 268–​70 revealing the hidden, 264 understanding and, 258–​62 repeating structures and analogical discovery,  131–​34 repertoire, in literature, 79–​80 representational content of understanding, 210–​11,  213 revealing the hidden, in appraisal of religious practices, 264 Rexroth, Kenneth, 75–​76 Rhodes, R. E., 238–​39 Ricoeur, Paul, 79, 256n.8 Rilke, Rainer Maria, 82, 84–​85 Robinson, Jenefer, 75 Robosem robot, 150–​51 robots, children’s understanding of and children’s feelings toward robots, 144–​50,  147f developmental trajectory of perceived agency on learning, 156–​57 development of, 142–​44 impact on learning from robots, 153–​56, 154f, 155f natural pedagogy, 140–​41, 151–​56 overview, 139–​41, 140f and robot designs and features, 159–​61 robots as teachers, 9, 150–​57 suggestions for further research, 148–​50,  157–​62 uncanny valley, 144–​46, 145f, 146f, 147f,  147–​49 Robovie robot, 142–​43, 156–​57 Rodriguez, F., 238–​39 Roman Empire, decline and fall of, 130–​31 root causes, preference for explanations with fewer, 237

Index  285 RUBI robot, 150–​51 Ruskin, John, 102–​3 Russell, Bertrand, 68–​69, 135   salience, 20 sampling processes and children as evaluators of others’ informativeness, 177–​81, 179f children’s inferences related to, 173–​75, 174f Sartre, Jean-​Paul, 74, 75, 83–​84 satyagraha, 270 schematic expectations, causal, 202 science(s) in appraisal of religious practices, 263–​64 assumptions about simplicity and complexity,  63–​65 challenges to faith caused by, 265–​66 charges of scientism, 94–​99 complaints about limited scope of, 93–​94,  103–​7 consciousness and subjectivity impregnable to, 98–​99,  104–​7 and constancy in seeking truth, 270–​71 evolving definitions of science, 99–​103 frames in, 17–​18 versus humanities, overview of, 47–​49 as illusion, 255 indexical versus non-​indexical claims, 47,  49–​51 individual versus collective focus, 48, 54–​56 intellectual authority, role in, 61–​63 intellectual progress, notions of, 59–​60 natural endpoint to inquiry in, 58–​59 perspectival versus non-​perspectival insights, 47–​48,  51–​54 prescriptive versus descriptive claims, 48,  54–​56 rethinking notion of intrinsic limits to, 103–​7 understanding in, 39n.4 use of frames in, 34–​35 Sciences of the Artificial (Simons), 202 scientia,  99–​102 scientism, debate over, 94–​99, See also naturalism scientist, coinage of term, 99 scope, simplicity in, 235–​36 Scruton, Roger, 69–​70n.13, 104–​5 secondhand knowledge and understanding versus firsthand knowledge, 111–​13 overview, 8–​9, 10 understanding why, 118 worth of secondhand judgment, 113–​16 second naivete, 256n.8

Second Temple Judaism, 258–​59n.14 seeking truth, constancy in, 270–​71 self, authentic, 269–​70 self-​correcting religious practices, 255–​57, 271 self-​deception,  265 selfless search for truth, 270 self-​understanding, frames as contributing to, 29 Shah, P., 238–​39 Shklovsky, Victor, 80 similarities, inquiry focused on, 63–​64 Simons, Herb, 202 simple explanations, 7 simple view of understanding, 211 simplicity assumptions about in sciences versus humanities,  63–​65 as explanatory virtue, 231–​32, 234–​39 simplification, 125 sins of omission, in pedagogical context, 177–​81,  179f Sloman, S., 7, 238, 245–​46 Small, Helen, 52 smart technology, 139, See also robots, children’s understanding of Snow, C. P., 96 social development, impact of robots on,  161–​62 social learning children as communicators of information,  182–​84 children as evaluators of others’ informativeness, 177–​82, 179f children as interpreters of information provided by others, 172–​77, 174f in communicative, pedagogical contexts, 169–​72,  171f general discussion, 184–​86 overview,  167–​69 rational inference in social context, 172–​84 social nature of human cognition, 245–​46 social robots. See robots, children’s understanding of Socratic influence, 57–​58, 62 Sosa, Ernest, 9–​11 sources of information, evaluation of, 8–​9 special populations, child-​robot interactions in, 159 stability, of concepts versus characterizations,  21–​22 stances, as paths to understanding, 212–​14 statistical information, children’s probabilistic inferences based on, 173–​75, 174f

286 Index Stecker, Robert, 68–​69n.10 stereotypes, 19 “stops-​at-​the-​neck” fallacy,  97 story model, 241 Strange Tools (Noe), 84–​85 Strevens, M., 211, 232–​33 string cosmology, 133, 134 strong differentiation thesis, 6–​7, 209, 223–​26 structural instantiation, through characterization,  38–​41 structural judgments, perspectives as influencing, 30 structure, understanding as grasp of different structures in same domain, 129–​31 general discussion, 134–​35 knowledge as special case of understanding,  128–​31 non-​propositional structure,  127–​31 overview,  123–​26 propositional structure, 126–​27, 128–​31 repeating structures and analogical discovery,  131–​34 simplification, 125 true belief and understanding, 126–​31 structures of concepts versus characterizations, 20–​21 and dependence relations, 5n.10 subjective goodness of categorical explanations, 245 subjectivity, inability of sciences to capture, 98–​99,  104–​7 success, apt, in Aristotelian ethics, 109–​10 Sussman, A. B., 235–​36 Swirski, Peter, 86n.75 sympathetic reconstruction, in understanding of people, 1 system completion, feature introduction by,  31–​33   taxonomies, conceptual, 30, 36–​37 Taylor, J. C., 238 teachers. See also social learning children as, 182–​84 evaluation of, 8–​9, 177–​82, 179f pedagogical reasoning, 175–​77 robots as, 9, 150–​57 technology, literature as cognitive, 84–​85, See also robots, children’s understanding of tectonic shifts, 258–​59n.14 Telenoid robot, 144–​46, 146f teleological/​functional mode of construal empirical evidence for, 214–​21 as path to understanding, 212–​14

telling details, 17, 27, 31 temporal patterns, as cognitive trace from exposure to mechanism, 202 testimony. See secondhand knowledge and understanding testing, limits on scientific, 103 Thagard, P., 239–​40 theory, ideal, and dispensability of perspectives,  35–​37 theory of mind, 87, 184–​85 third-​personal understanding, 225 Tilley, Terrence W., 12, 254n.3, 262n.18 training related to grasping structure, 134 transcendent, use of ordinary to understand,  259–​61 transcendent actors, and accuracy of reference, 266 Treatise of Human Nature (Hume), 224 Trout, J. D., 211n.2 true beliefs, 126–​31 trust in testimony (natural pedagogy), 140–​41, 151–​56, 154f, 155f truth, appraisal account of accuracy of attribution, 267 accuracy of reference, 266 coherence,  265–​66 constancy in seeking truth, 270–​71 fittingness, 265 overview, 262, 263, 264 promoting authenticity and fidelity, 268–​70 revealing the hidden, 264 Tversky, Amos, 20 two-​way cross-​domain transfers, 258–​59n.14   ultimate characterization, 37–​41 ultimate theory, 35–​37 uncanny valley, 9, 144–​46, 145f, 146f, 147f,  147–​49 uncovering the hidden, in appraisal of religious practices, 264 underinformative teachers, children’s evaluation of, 177–​82, 179f understanding. See also partial understandings; robots, children’s understanding of; specific types of understanding characterizations in, 38–​41 versus explaining, 196–​97 frames as contributing to, 29–​33 as grasped relatedness of items, 39 as grasp of structure, 123–​35 humanism versus naturalism, 1–​2 inarticulacy and, 118–​21 interchangeable terms for, 48–​49

Index  287 knowledge as special case of, 128–​31 knowledge seeking, 10–​11 others, 167–​68, 176–​77,  184–​85 perspectives as contributing to, 29–​33, 34–​37 psychology of, 4–​9 relationship between literature and, 70–​71 as representing (explanatory) dependence, 210–​11, 213, 221 simple view of, 211 stances or modes of construal as paths to,  212–​14 understanding why, 117–​18 varieties of, 2–​5, 209–​10 unification, as explanatory virtue, 232–​33,  242–​46 unity of all reality, 134–​35 as derived from structure, 124, 131 universe, organizational structure of, 132–​35 Ursus robot, 150–​51 utility-​based social reasoning, 182–​83, 184–​86   varieties of aptness, 111 Varieties of Religious Experience, The (James), 268–​69n.31

varieties of understanding, 2–​5, 209–​10 Vasilyeva, Nadya, 222 Velleman, David, 68–​69 veritism, 39n.4 verstehen tradition, 1–​2n.2, See also humanism virtue epistemology, 110–​13   Wallace, David Foster, 267n.28, 269–​70n.32 weak differentiation thesis, 6, 209, 221–​23 Weisberg, D. S., 238, 242 Wellman, Henry M., 8, 9 why, understanding, 117–​18 Wieseltier, Leon, 96–​97, 98, 99, 104 Wilkenfeld, Daniel, 5–​7, 222, 245 Wittgenstein, L., 49–​50, 71, 245 Woodward, Amanda L., 219 Woodward, James, 210–​11 Woolf, Virginia, 11–​12, 71–​73, 76–​77, 82 Wright, Larry, 214–​15 Wright, Richard, 74, 83–​84   Zagzebski, Linda, 5n.10, 258–​59n.14 Zemla, J. C., 219, 234–​35, 238, 239–​40 Zenbo robot, 139–​40, 140f