247 28 28MB
English Pages 288 Year 2019
WHAT’S THE POINT OF
KNOWLEDGE? A Function-First Epistemology
MICHAEL HANNON
What’s the Point of Knowledge?
What’s the Point of Knowledge? A Function-First Epistemology
MICHAEL HANNON
OXFORD UNIVERSITY PRESS
UNIVERSITY PRESS
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.
Library of Congress Cataloging-in-Publication Data Names: Hannon, Michael (Michael J.), author. Title: What’s the point of knowledge? : a function-first epistemology / Michael Hannon. Description: New York, NY : Oxford University Press, [2019] I Includes bibliographical references and index. Identifiers: LCCN 2018032483 (print) ILCCN 2018051505 (ebook) I ISBN 9780190914752 (online content) I ISBN 9780190914738 (updf) I ISBN 9780190914745 (epub) I ISBN 9780190914721 (cloth: alk. paper) Subjects: LCSH: Knowledge, Theory of. Classification: LCC BD161 (ebook) I LCC BD161 .H36 2019 (print) I DDC 121—dc23 LC record available at https://lccn.loc.gov/2018032483
CONTENTS
Preface vii
Introduction
1
CHAPTER 1 Methodologies in Epistemology
11
CHAPTER 2 The Point of Knowledge 35 CHAPTER 3 The Value of Fallible Knowledge CHAPTER 4 Impure Knowledge
81
CHAPTER 5 Pluralism about Knowledge CHAPTER 6 Epistemic Diversity
56
103
139
CHAPTER? Epistemic Pragmatism
157
CHAPTER 8 Skepticism and the Point of Knowledge CHAPTER 9 What’s the Point of Understanding? Bibliography 257 Index 271
222
189
PREFACE
hy do humans use words like “knowing,” “understanding,” and “ra tional”? Put more broadly, what is the point of epistemic evalua tion? This book attempts to answer this question. I began working on this topic in 2009 at the University of Cambridge, where my doctoral thesis explored the ways in which speaking of “knowing” facilitates human cooperation, survival, and flourishing. Drawing on the work of Edward Craig, I hypothesized that humans think and talk about knowledge primarily to identify reliable sources of information to members of their community. I claimed (and still claim) that identifying reliable informants is essential for pooling, sharing, and retrieving useful information. This idea might appear simple, but I believe it has wide-reaching implications. In subsequent years, I have continued investigating the nature, purpose, and value of knowledge, building on my previous work while pushing deeper to extend it in new directions. The result is this book. In a book about our epistemic dependence on others, I must thank the many people to whom I am indebted for their comments, criticisms, and encouragement while working on this project. Two of the most influen tial people are Hallvard Lillehammer (my PhD supervisor) and Stephen Grimm (my postdoc supervisor). Together their incisive comments have shaped many ideas in this book. I cannot imagine a better pair of mentors and I owe each of them an immeasurable debt. I am also grateful to Krista Lawlor for agreeing—on very short notice—to supervise part of this proj ect while I was a visiting postdoc at Stanford University in the spring of 2017.
W
Others who deserve thanks are Byron Alvares, Manuel Alves, Nathan Ballantyne, James Beebe, Jessica Brown, Chris Cowie, Edward Craig, Tim Crane, Keith DeRose, Sinan Dogramaci, Kate Elgin, Sam Elgin, Georgi Gardiner, Mikkel Gerken, John Greco, David Henderson, Nick Hughes, Mark Kaplan, Klemens Kappel, Chris Kelp, Martin Kusch, Jess Kwong, Tania Lombrozo, Helen Marsh, Robin McKenna, Joshua Mozersky, Gareth Nellis, Corey Nishio, Huw Price, Duncan Pritchard, Kevin Roberts, Patrick Rysiew, Anand Shan, Barry Smith, Ernest Sosa, and Nick Treanor. I am especially thankful to Kate Elgin and Robin McKenna for reading and providing comments on a full draft of this manuscript. My family, too, is owed more than deep thanks. Their influence on my work is less obvious but no less real. I also want to apologize to anyone I forgot to mention by name and say thank you for your help. This book is the result of many years of thinking about the social role of knowledge, and I am sure I have learned things from more people than I can now remember. Audiences at the following events have also provided helpful feed back at various stages of writing: the Bled Epistemology Conference (2017), Queen’s University Colloquium Series (2017), University of Vienna (2017), Varieties of Understanding Seminar (2016), Concepts and Cognition Lab at UC Berkeley (2016), APA Pacific Division Meeting (2015), Buffalo Experimental Philosophy Conference (2015), Helsinki Epistemology Workshop (2015), Central States Philosophical Association (2015), APA Eastern Division Meeting (2014), NYC Philosophy of Language Workshop (2013), Midsummer Philosophy Workshop (2013), European Epistemology Network Meeting (2012), and the Cambridge Faculty of Philosophy (2010-2013). I have drawn, with kind permission, on previously published work. Parts of chapter 2 are taken from “The Practical Origins of Epistemic Contextualism,” Erkenntnis (2014) 78 (4): 899-919. Chapter 3 draws on my paper “Fallibilism and the Value of Knowledge,” Synthese (2014) 191 (6): 1119-1146. Chapter 4 is based on “A Solution to Knowledge’s Threshold Problem,” Philosophical Studies (2017) 174 (3): 607-629. Parts of chapter 5 are taken from “The Importance of Knowledge Ascriptions,” Philosophy Compass (2015) 10 (12): 856-866. Chapter 6 is a substantially revised and expanded version of “The Universal Core of Knowledge,” Synthese (2015) 192 (3): 769-786.1 want to thank the relevant editors and publishers for letting me use this material. I am also grateful to the Social Sciences and Humanities Research Council of Canada for providing me with a doctoral fellowship that
viii
|
Preface
enabled me to study at the University of Cambridge, where the ideas for this book first took shape. I am also thankful to King’s College and the Faculty of Philosophy in Cambridge for supporting my research. My former colleagues at Queen’s University, especially Josh Mozersky, also deserve thanks for providing me with a fellowship that allowed me to bring this book much closer to completion. I would also like to thank Barry Smith and the Institute of Philosophy for giving me the support needed to finish this project, as well as Lucy Randall at Oxford University Press and two anonymous reviewers for all their helpful feedback. While I am thankful to many people, I could not have written this book without the love, support, and encouragement of Elizabeth Edenberg. Thank you for your relentless optimism, your generosity of spirit, and for editing and improving everything I write.
Preface
|
ix
I Introduction
knowledge and its value. My aim is to reveal the na ture, purpose, and significance of knowledge by investigating why humans think and speak of knowing. The word “know” is remarkable for a number of reasons. It is one of the 10 most commonly used verbs in English, outpaced only by utterly basic verbs like “be,” “have,” “do,” “go,” and “say.” It is also the first cog nitive verb that children leam and the most frequently used term in epistemic appraisal.1 For example, it is far more common for us to speak of “knowledge” than it is for us to speak of “reliability,” “justification,” “understanding,” “wisdom,” and an array of other epistemological notions. Perhaps most strikingly, the word “know” seems to find a comfort able meaning-equivalent in every human language. This is especially noteworthy because empirical evidence from cross-linguistic semantics indicates that almost every word in the English lexicon does not have an equivalent in many, perhaps most, other languages. Even words that refer to common emotional states like “sad” and “angry,” as well as words for seemingly universal mental states and processes like “believe” and “re member,” are language and culture specific. In contrast, linguists have iso lated “know” as one of a very small number of words that are allegedly culturally universal.12 This all suggests that knowledge is deeply important to human life. this book is about
1 See Davies and Gardner (2010), Shatz, Wellman, and Silber (1983), and Gerken (2015). 2 See Goddard (2010). More precisely, the sense of “know” that embeds a propositional or whcomplement is arguably universal. Not every language follows English by making “know” perform multiple tasks; for example, the French have “savoir’7“connaitre” and Germans have “wissen”/“kennen.”
This book aims to shed light on the nature and importance of knowl edge by investigating what our epistemic words, concepts, norms, and practices are for. I call this methodological approach function-first epistemology. Although the name is new, the methodology is not. Edward Craig took this approach in his insightful and (until recently) underappreciated book Knowledge and the State of Nature, hi that work, Craig argues that our prac tice of attributing knowledge plays a vital role in human cooperation. More specifically, he says we speak of “knowing” in order to recommend good sources of information to members of our community. This practice is neces sary, or at least deeply important, because humans are information-dependent creatures that often rely on the testimony of others. I find this idea highly plausible and I will attempt to show that we can de rive substantial epistemological payoffs by adopting it. By reflecting on the purpose of epistemic evaluation, we can make headway on topics such as the nature and value of knowledge, the intractability of skepticism, the relation ship between knowledge and practical reasoning, the Gettier problem, as well as the nature and value of human understanding. The extent of my theoretical debt to Craig is sometimes embarrassing, but my focus in this book is not on exegesis. The philosophical approach ar ticulated and defended here is distinct from Craig’s proposal in a number of ways. For example, Craig’s inquiry is directed exclusively at explicating the concept of knowledge, whereas function-first epistemology is a method of in quiry directed at epistemic evaluation more broadly. Further, Craig provides a genealogical story that traces the development of the concept of knowledge from the more primitive concept of a good informant, which in turn arose in a “state of nature.” In contrast, function-first epistemology is not a genealogy and makes no reference to a fictional state of nature. I’ll say more about this in chapter 2. These are just two of many ways in which I modify Craig’s initial proposal. I will also respond to unanswered criticisms, fill gaps he left open, and extend this approach to areas of the epistemological landscape left untouched by his work. Chapter 9, for example, explores the nature, purpose, and value of human understanding. One of my central claims will be that understanding and knowledge play different social roles. By examining these two cognitive achievements from the point of view of their function, we can throw much light on epistemic value—the kind of value that attaches to cognitive successes. I will argue that, contrary to popular belief, knowledge plays a more important role than under standing in human life.
2
|
What’s the Point of Knowledge?
Within the past decade or so, a number of scholars have started thinking and writing about the function of epistemic evaluation.3 This has made it an exciting and rewarding time to work on this otherwise neglected topic. But despite the surge of interest in the function of epistemic evaluation, this approach has faced two major obstacles. First, epistemologists have tended to adopt Craig’s general framework without defending this meth odology. They merely assume this approach is a viable starting point for further theorizing, but this very starting point is controversial. One of my goals is to make function-first epistemology more palatable by answering a wide range of objections that have accumulated in recent years. Second, people often find it unclear what this method really amounts to. I have met many scholars who warmly approve of this research program while also admitting they don’t quite know what it is. They ask questions like: What does it mean to say knowledge has a “function”? Doesn’t it have a variety of functions? How are they related? One of the goals of this book is to answer these sorts of questions and, in doing so, to put function-first epistemology on a sounder methodolog ical basis. I also hope to show that we can answer many interesting and difficult questions in epistemology by reflecting on the purpose of epi stemic evaluation. At the core of this book are two broad proposals. First, I suggest that we can make progress in epistemology by taking a function-first approach. Second, I hypothesize that the function of the concept of knowledge is to identify reliable informants—a practice that is vital for human survival, cooperation, and flourishing. These two proposals are not inseparable. You might endorse the method of function-first epistemology while rejecting my hypothesis about the function of the concept of knowledge; or you might reject my methodological approach and yet think there is an im portant conceptual connection between knowing and being a reliable in formant. Nevertheless, I will defend both proposals. I want to emphasize, however, that the usefulness of function-first epistemology would not be undermined if it fumed out that my particular hypothesis about the con cept of knowledge were mistaken. I hope to demonstrate the power of this methodology without necessarily relying on my hypothesis about our need to identify reliable informants. Indeed, some chapters in this book are
3 See Williams (2002), Neta (2006), Weinberg (2006), Fricker (2008), Greco (2008), Henderson (2009), Gelfert (2011), Kelp (2011), Pritchard (2012), Dogramaci (2012), McKenna (2013), Lawlor (2013), Hannon (2013, 2017a), Greco and Hedden (2016), and the contributions in Greco and Henderson (2015).
INTRODUCTION
|
3
mainly concerned with the general method but not the specific hypothesis (such as the chapter on epistemic diversity). This book provides a deeply social picture of knowledge, one that places our reHance on others at center stage. The details of this idea will be filled out as we proceed. For now, I’ll simply emphasize that a very small part of our knowledge of the world comes to us from our immediate experience. In vast areas of life we depend on what others tell us, and this dependence is not restricted to what we are told in face-to-face interactions. As Richard Moran observes, “we also take ourselves to know all sorts of things that only reached us through a long chain of utterances and documents, whose evidential status we have never investigated for ourselves and will never be in a position to investigate” (2005: 1). Testimony, or secondhand in formation, is a basic way in which knowledge gets around. If the only in formation available to us came directly, our stock of knowledge would be significantly impoverished. This goes against the tradition of “epistemic individualism” in epistemology according to which the legitimate springs of knowledge are located in firsthand experience. It is clear, however, that we cannot get by without relying on other people. Recent work in evolutionary theory supports this idea. In The Evolved Apprentice, Kim Sterelny emphasizes our refiance on others to learn valu able skills for survival. One of the most distinctive features of human life, he claims, “is our dependence on intricate networks of cooperation and the division of labor” (Sterelny 2012: 101). He goes on to say that “No living humans gather the resources needed for a successful life by their own efforts.” In other words, we are not isolated inquirers: our cognitive competence is a collective achievement that depends on many ordinary individuals who gather and share the informational resources on which human life depends. Further, this type of cooperation and information ex change is what ushered along our sharp divergence from our great-ape relatives—it is what led to our distinctive intelligence and cognitive power. I will not attempt to integrate my account of knowledge with empirical theories of how human cognitive behavior evolved, but this book can be seen as providing a framework for bringing evolutionary considerations to bear on our understanding of the social role of knowledge. This book has nine chapters. I myself don’t enjoy reading (or writing) chapter summaries, but so few people agree with me about this that I have decided to provide the reader with a road map. Chapter 1, “Methodologies in Epistemology,” outlines the method of function-first epistemology and highlights some of its benefits. This methodology involves three broad steps: we start with a prima facie
4
|
What’s the Point of Knowledge?
plausible hypothesis about the role of some epistemic concept (norm, practice) in human life; then we try to determine what a concept (norm, practice) having this role must be like; finally, we examine the ex tent to which the concept (norm, practice) we have described matches our everyday judgments. In order to highlight what is distinctive and fruitful about function-first epistemology, I also compare it with four alternative approaches: reductive conceptual analysis, knowledge-first epistemology, reverse engineering of epistemic evaluations, and epis temological naturalism. I argue that each of these methods faces lim itations that my approach does not. For example, I argue that certain methodologies rely too heavily on linguistic usage or intuitive judgments about cases; as a result, they only end up describing our actual practices and not characterizing those that would be best for us. In contrast, my view leaves space for the normative project of evaluating how well or poorly our epistemic concepts, norms, and practices actually satisfy our needs and goals. I also argue that knowledge is a socially constructed kind—it is not something we discover in the natural world but rather something we impose upon it. I then explore the ramifications of this proposal for how we ought to investigate knowledge and knowledge claims. This metaepistemological groundwork provides the basis for the rest of the book, which uses the function-first method to answer some of the most challenging questions in epistemology. Chapter 2, “The Point of Knowledge,” provides a preliminary statement and defense of this book’s central hypothesis, namely: the purpose of the concept of knowledge is to identify reliable informants. I use this hypoth esis to yield a systematic account of knowledge and knowledge claims. There are two key players in my story. The first is “the inquirer” who is trying to find accurate information; the second is “the reliable informant” who provides the information the inquirer seeks. We start with the idea of an individual inquirer who needs an informant to satisfy a need for infor mation; then we imagine a more complex situation in which a community of individuals collaborate to pool and share information. As our interest in information becomes more socially directed, it becomes vital that we iden tify individuals who are sufficiently reliable for members of our commu nity. This results in an “objectivized” standard for knowledge. According to this idea, a “knower” is someone reliable enough for the community at large. After outlining this thought, I try to defend it from objections. For example, what should we say about people who seem not to qualify as re liable informants and yet clearly have knowledge? Also, does the notion of “objectivized” knowledge lead to skepticism? And doesn’t all this rely
INTRODUCTION
|
5
on dubious quasi-historical postulations? This chapter provides answers to these (and other) questions. Chapter 3, “The Value of Fallible Knowledge,” attempts to solve a problem that vexes all fallibilist theories of knowledge, which I call the “threshold problem.” This is the problem of how to provide a plausible ac count of what fixes the threshold (level, degree) of justification (evidence, probability, warrant, supporting ground) for knowledge in a nonarbitrary way—and in a way that also makes sense of the perceived value of knowl edge. Although it is widely accepted that knowledge does not demand in fallibility, epistemologists have been largely silent about how strong the justificatory component of knowledge must be. As Laurence BonJour writes, “it is fair to say that nothing like a precise specification of this [level of justification] has ever been seriously suggested, let alone more widely endorsed” (2010: 61). This chapter attempts to answer this challenge. By appealing to the hypothesis that the concept of knowledge is used to flag reliable informants, I argue that we can reasonably determine the level of justification required for fallible knowledge; further, we may explain why this level of justification has the significance that makes knowledge valu able. I also explore the alleged payoffs of rejecting fallibilism and I argue these benefits are illusory. Chapter 4, “Impure Knowledge,” defends a view that I call “communal impurism.” Purism is the view that whether or not a true belief amounts to knowledge depends exclusively on epistemic factors—for example, on whether the true belief was formed in a reliable way or was supported by good evidence. Impurists deny this. My goal in this chapter is to defend a new version of impurism that is based on our communal need to identify reliable informants. To do this, I must reconcile two seemingly conflicting ideas. On the one hand, I claim that knowers must be sufficiently reli able to serve as actionable sources of information for members of their community (i.e., they must meet a communal threshold for knowledge). On the other hand, impurists claim the level of justification required for knowledge is partly fixed by an individual’s practical reasoning situation (i.e., it becomes more difficult to know as the stakes go up for the puta tive knower). We thus face a puzzle: it seems that a knower must meet a communal standard for knowledge and yet the threshold for knowledge is partly determined by an individual’s practical reasoning situation. But how can there be both a general (communal) standard for knowledge and a fluctuating standard that is partly determined by individual practical interests? I reconcile these ideas by arguing for the following view: there is a communal standard for knowledge that is needed to promote a deep kind
6
|
What’s the Point of Knowledge?
of coordination in our epistemic practices, and yet an individual’s stakes may trump the communal standard when it is not sufficiently demanding for the relevant practical reasoning situation. In this way, the concept of knowledge continues to serve the practical function of identifying suffi ciently reliable informants. Chapter 5, “Pluralism about Knowledge,” tackles one of the most common objections to my view, namely: why think the concept of knowl edge has just one purpose? There seem to be many uses of this concept and there is no a priori reason to favor one use over any others. Indeed, there have been a variety of proposals about what role our knowledge con cept plays in epistemic evaluation; for example, it has been suggested that we think and speak of knowing in order to signal the appropriate end of inquiry; to track the epistemic norm governing warranted assertion and practical reasoning; to distinguish between blameworthy and blameless behavior; to provide assurance to others; to encourage good testimony; and so forth. This plurality of views should make us wonder whether the concept of knowledge has just one function. And even if we suppose it does have a single function, how do we decide between these various proposals? In this chapter, I try to ascertain what the concept of knowledge must be like to accommodate all these plausible functions. I defend a ver sion of “epistemic pluralism” that countenances the idea that we speak of knowing for a variety of purposes, but I also claim the primary purpose of the concept of knowledge is to identify reliable informants. Chapter 6, “Epistemic Diversity,” provides a new argument against epi stemic relativism. Traditionally, epistemological investigation has aspired to general conclusions, not culturally local ones. As Rebecca Kukla observes, epistemologists customarily take questions about the nature of knowledge, justification, fallibility, and the like to be questions about “uni versal principles and concepts unmarked by any particular demographic or geographic limitations” (2015: 202). For example, skepticism is supposed to be worrying and interesting because it purports to show that there is no human knowledge, or almost none. According to the epistemic relativist, however, the cognitive norms that determine what counts as knowledge (or whether a belief is rational, justified, etc.) vary with, and are dependent on, local conceptual or cultural frameworks. In recent years, a group of exper imental philosophers have claimed to provide empirical evidence for this sort of epistemic diversity. In this chapter, I argue the data actually supports the existence of a universal folk epistemology. I further support this idea by providing a theoretical argument for why language users living in social communities would develop a cross-culturally (and cross-linguistically)
INTRODUCTION
|
7
shared concept of knowledge. More specifically, I argue that all humans have certain basic practical needs that the concept of knowledge is used to satisfy and we could not meet these needs without a word that is (near-) synonymous with “know.” Thus, we should expect every culture to have a word to express the concept of knowledge. Function-first epistemology is thereby exonerated from the charge of relativism. Chapter 7, “Epistemic Pragmatism,” centers on a hotly debated ques tion in epistemology: what is the correct semantic account of “knows”? Scholars have recently appealed to putative facts about the purpose of knowledge ascriptions to adjudicate this dispute. I am generally sympa thetic with the idea that new epistemological insights can be gained by thinking about the purpose of having a concept of knowledge, but I raise doubts about the viability of appealing to facts about the purpose of the concept of knowledge to determine the correct semantic theory of “knows.” In fact, I argue this entire debate about the semantics of “knows” rests on an error: it presupposes that we should account for the meaning of epi stemic claims by determining their truth conditions. A more natural way to approach the meaning of epistemic claims, I argue, is to ask what prac tical functions they serve us in communicating with each other. Drawing on the work of J. L. Austin, I articulate a view called “epistemic pragma tism.” This view questions the pervasive presupposition that the best way to account for the meaning of epistemic claims is to look for referents or truth conditions. Instead, we should look for the social or practical roles of words and concepts in order to understand their meaning. If this proposal is correct, it will undercut a variety of popular theories and could poten tially reorient contemporary debates about the semantics of knowledge ascriptions. Chapter 8, “Skepticism and the Point of Knowledge,” explores the rela tionship between philosophical skepticism and the concerns of daily fife. By reflecting on the clash between our everyday outlook and the attitude to which we are led by philosophical reflection, I hope to show that functionfirst epistemology can augment an argument against skepticism. I ex plain the force of the skeptic’s argument as well as our desire to reject the skeptical conclusion in the following way: our need to share information pushes us to accept stricter epistemic standards that might logically end at skepticism, but practical factors encourage us to formulate standards that stop short of skepticism. This tension creates an area of indeterminacy in which controversies about skepticism take place. However, I do not at tempt to resolve the clash between skepticism and our everyday outlook by appealing to one perspective against the other. My theory of knowledge
8
|
What’s the Point of Knowledge?
explains how (and to what extent) knowledge is possible, but without as suming there is something drastically mistaken with the skeptical chal lenge. This diagnosis leaves room for the idea that there may be contexts in which it is appropriate to make skeptical challenges. Some fallibilists may begrudge the skeptic even this limited and temporary victory, but I am not one of them. I believe we can explain the persuasive power of skepti cism while also explaining why skeptical worries do not (and should not) threaten our everyday knowledge. Chapter 9, “What’s the Point of Understanding?, ” extends the functionfirst approach to human understanding. I argue that the concept of un derstanding serves the practical function of identifying good explainers, which is an important dimension of epistemic evaluation. This hypothesis throws light on a variety of issues, including the role of explanation in understanding, the relationship between understanding and knowledge, epistemic luck, and the value of understanding. I also argue that under standing and knowledge play different social roles: roughly, knowledge is closely tied to answering our need for true beliefs, whereas understanding answers our need for good explanations. Everyday inquiry is typically aimed at true beliefs, which is why knowledge matters, but sometimes we need more than just true beliefs to get by in the world. This book is written in the belief that new epistemological insights can be gained by thinking about the point of having epistemic concepts like knowledge and understanding. I aim to show that some major issues in epistemology can be resolved by taking a function-first approach, thereby illustrating the significant role that this method can play in contemporary philosophical debates. What’s more, the “function-first” approach can be applied to areas be yond epistemology. There are many objects of philosophical inquiry that may, like knowledge, serve some important function or role in human life, and the kind of theoretical investigation practiced here might with equal profit be applied to other areas of philosophy. Indeed, there is an active re search program in moral theory that treats moral concepts exactly as I treat epistemic concepts. Function-first epistemology is therefore just one ap plication of a much broader philosophical methodology, which we might call “function-first philosophy.” I hope others will take up this research program and that this book will be useful to those interested in extending this method to other philosophical topics. Finally, a few words about the current epistemological landscape. The idea that knowledge is deeply important has been roundly criticized in recent epistemology; for instance, it has been argued that we can fully
INTRODUCTION
|
9
explain rational behavior without appealing to knowledge; that knowledge isn’t more valuable than justified true belief; that knowledge is an inco herent notion or a myth; and that we should instead focus on other epi stemic states that are allegedly more important, such as understanding or wisdom.4 Many philosophers now believe the myopic focus on knowledge in the history of epistemology cannot be defended. Contrary to this trend, I believe that knowledge is fundamental to our epistemic life and belongs at the heart of epistemological theory. It will emerge from this book that life without knowledge, whether nasty and brutish or not, would be more solitary and almost certainly shorter.
4 Schiffer (1996), Kaplan (2003), Kvanvig (2003), Elgin (2006), BonJour (2010), and Grimm (2015a) defend versions of these claims.
10
|
What’s the Point of Knowledge?
chapter 1 | Methodologies in Epistemology
It is by its methods rather than by its subject matter that philosophy is to be distinguished from the other arts and sciences. —A. J. AYER (1956: 7)
A. J. ayer opens The Problem of Knowledge by remarking that philos ophy is to be distinguished from other disciplines not by its subject matter but rather by its methods. Yet philosophy is increasingly preoccupied with determining the validity of its own methods. It is widely assumed that a deeper understanding of philosophical methodology is necessary for clarity and progress, but there is no consensus about which methodology is the right one. Perhaps the most traditional method of philosophy is reductive con ceptual analysis, the goal of which is to enumerate the necessary and suf ficient conditions of some philosophically interesting concept. This style of philosophy goes all the way back to Socrates. In recent years, however, this method has come under attack from various perspectives. A result of these attacks has been increased disagreement about which methods to use and what types of data support philosophical claims. We now find on offer a variety of methods from which to choose; for instance, there is conceptual analysis, philosophical naturalism, experimental philosophy, practical explication, reverse engineering, and ordinary language philos ophy, to name a few.1 Further, the evidence used to support philosoph ical theories includes intuitions elicited by thought experiments, linguistic data from ordinary usage, reflective judgments, and empirical science1 1 Proponents of these views in epistemology include Jackson (1998), Komblith (2002), Weinberg et al. (2001), Craig (1990), Dogramaci (2012), and Austin (1979).
(broadly construed). But there is little agreement about how to best eval uate these methods and how much support the various sources of evidence provide philosophical claims. For example, how easily should we reject our intuitive judgments? How much stock do we put in ordinary language usage? Do we require an Archimedean perspective from which to evaluate competing methodologies and theories? There is no better way to test a proposed philosophical method than by examining how well it does in solving thorny philosophical problems. As Bertrand Russell once said, “nothing of any value can be said on method except through examples” (1914: 240). Following Russell’s advice, the main goal of this book is to make use of a fairly new methodology in epis temology that I call function-first epistemology. But what is function-first epistemology?
1.1 Function-First Epistemology A function-first epistemologist seeks to explain the nature and value of an epistemic concept, norm, or practice by reflecting on its function or purposes. The guiding idea is that we will better understand our epistemic concepts, norms, and practices by investigating what they are for. Thus, a function-first epistemologist will ask questions like: Why do humans speak and think in terms of “knowing,” “understanding,” and “ration ality”? What epistemological distinctions and norms would best facilitate human survival and flourishing? What might life be like without our cur rent practices of epistemic evaluation? Do our concepts of epistemic eval uation carry much weight in science, philosophy, or daily life? This way of doing epistemology is inspired by the work of Edward Craig, who developed a similar approach in his book Knowledge and the State of Nature.2 My view is deeply indebted to Craig, but the method of function-first epistemology goes beyond his proposal in many ways. Before I distinguish my view from Craig’s, however, I want to foreground our deepest and perhaps most important point of agreement. The central
2 This approach has interesting similarities and dissimilarities with Carnap’s (1950) method of “practical explication.” The task of explication is to transform an inexact concept (the explicandum) into an exact one (the explicatum), effectively replacing the former with the latter. The adequacy of the explicatum is assessed in light of the theoretical purpose it is intended to serve in the target theory. Thus, like the function-first approach, the method of explication is related to a purpose; but whereas the former refers to the purpose that the concept in question serves in our life, the latter focuses on the purpose that the concept is intended to serve within a target theory.
12
|
What’s the Point of Knowledge?
hypothesis of my book is that the primary function of the concept of knowledge is to identify informants who are reliable enough to appropri ately serve as sources of actionable information for members of our com munity. Craig pithily describes this basic idea as follows: the concept of knowledge is used to “flag approved sources of information” (1990: 11). This puts the matter very roughly, of course; the chapters that follow will develop this hypothesis in great detail. For now, I will simply assert that this core idea has the potential to allow progress on a wide range of epis temological issues. Whether I eventually make good on this claim, the reader is left to decide. I have said that my general strategy is to explore the nature and impor tance of epistemic evaluation by investigating the point or purposes of our epistemic concepts, norms, or practices. But how do we determine what role (or roles) our concepts, norms, or practices play and what needs they satisfy? Craig describes his methodology as follows: We take some prima facie plausible hypothesis about what the concept of knowledge does for us, what its role in our life might be, and then ask what a concept having that role would be like, what conditions would govern its application... then see to what extent it matches our everyday practice with the concept of knowledge as actually found. (1990: 2-3)
As this passage makes clear, Craig uses this methodology to investi gate the concept of knowledge. But we may generalize this approach by applying it to epistemic evaluation more broadly. After all, humans have a range of predicates that serve to evaluate the epistemic position of others, such as “knows,” “understands,” and “justifies.” We can there fore ask what the point is of each term of epistemological appraisal. I will focus primarily on knowledge in this book, but I will also extend function-first epistemology to investigate the nature, purpose, and value of human understanding. Taking our cue from the passage quoted above, we may characterize the “function-first” strategy as involving three broad steps. First, we start with some hypothesis about the role of some epistemic concept (norm, practice) in our life. (For simplicity, I will henceforth speak primarily of concepts and set aside norms and practices.) In order for this hypothesis to be plausible, it must be compatible with certain facts about human life, such as facts about our physical environment, our social organization, our cognitive capacities, and the basic aims and interests humans typically have. These facts about humans and their circumstances will then give rise
METHODOLOGIES IN EPISTEMOLOGY
|
13
to a certain conceptual need that is supposed to be satisfied by the purpose described by our hypothesis. Second, we try to determine what a concept having this role would be like (i.e., what conditions would govern its application). At this second stage we can perhaps gain new insights about familiar conceptual issues such as the factivity of knowledge, the relationship between knowledge and epistemic luck, and further questions about the necessary conditions for knowing, understanding, and other epistemic states. Once we better appreciate the role of the relevant concept in human fife, we can ask what features a concept that satisfies this role would have; for instance, would the presence of epistemic luck threaten the function of the concept of knowledge? If so, then we have a reason to regard knowledge as incom patible with epistemic luck. I will not pretend to resolve every issue about the nature of knowledge that is raised in the course of this book, but at certain points I will try to indicate where my account could be developed in different ways. Third, we must examine the extent to which the concept we have constructed matches our everyday notion. In this way, our investigation is anchored by the ordinary concept that we are looking to explicate. Suppose, for example, that we are investigating the concept of knowledge. If our investigation were to reach a result quite different from the intuitive extension of the word “knows,” then, as Craig says, “barring some special and especially plausible explanation of the mismatch, our original hypoth esis about the role that the concept plays in our life would of course be the first casualty” (1990: 2). Our aim is to construct a concept that not only functions in the manner suggested by our hypothesis, but also one that fits our intuitions (or explains why our intuitions diverge). Thus, while func tion comes “first” on this approach, it is not the last word. This methodology has many benefits. Indeed, I aim to show that some of the most pressing issues in epistemology can be resolved by taking a function-first approach. It will take considerable time and argument to display these benefits in full view, but five of these benefits can be stated up front. First, this method allows us to illuminate the importance of knowl edge, understanding, and other epistemic states. After all, we evaluate individuals as having or lacking knowledge (understanding, etc.) and these evaluations seem to matter, so our epistemological theory should explain why they matter. Second, this approach can shed light on the na ture of knowledge (understanding, etc.), since we can ask what knowledge (understanding, etc.) must be like in order to play its distinctive role in
14
|
What’s the Point of Knowledge?
epistemic evaluation. Third, function-first epistemology can help us re solve conflicts of intuitions about cases, since we can appeal to claims about the role of our epistemic concepts to adjudicate between conflicting intuitions. In other words, it offers a way of doing “postintuitive episte mology.” Fourth, we can avoid verbal disputes about how to use words like “knows,” “understands,” and so on. As David Chalmers (2011) argues, one way to avoid descent into purely terminological debates is to clearly indi cate what theoretical roles the concept in question is supposed to play, and then to theorize about what would enable the concept to best play those roles.3 Finally, this approach will allow me to explain what is right about other theories in the literature in epistemology—or so I will argue. This is all fairly abstract. A way to further improve our understanding of function-first epistemology is by comparing it with other approaches and highlighting the differences between them. The next four sections will look at the following philosophical frameworks: traditional concep tual analysis, knowledge-first epistemology, reverse engineering epistemic evaluations, and epistemological naturalism. I make no attempt to pro vide a comprehensive survey of the available alternatives. Rather, I discuss these approaches because the juxtaposition illuminates what is distinctive about function-first epistemology. By comparing and contrasting these approaches with my own, we will get a better grip on what function-first epistemology is and why it is useful. This lays the metaepistemological foundation for the rest of the book, which uses this methodology to make progress on a wide range of issues in epistemology.
1.2 Conceptual Analysis in Epistemology Epistemologists have largely been interested in analyzing propositional knowledge. This type of knowledge is commonly expressed using the schema “S knows that p” (or “S knows whether p”), where “S” refers to the knowing subject and “p” to the proposition that is known. Analyses of propositional knowledge have largely overshadowed attempts to elu cidate other types of knowledge, such as knowledge-how (e.g., knowing how to ride a bike) and knowledge by acquaintance (e.g., knowing John). One might question whether it is right to make propositional knowledge the central focus of epistemology, but this book will concentrate mainly on propositional knowledge. I am interested in propositional knowledge
3 Smithies (2015) mentions similar benefits.
METHODOLOGIES IN EPISTEMOLOGY
|
15
because, first, I want to contribute to existing debates that share this focus and, second, propositions are the most common vehicle of transmittable information. While my aim is to develop a unified theory of propositional knowledge, I am open to the possibility of a nonreducible variety of uses of “know,” not all of which can be assimilated into my account. A philosophical “analysis” has traditionally been understood as enumerating a concept’s individually necessary and jointly sufficient conditions. The basic idea is that questions like “What is knowledge?” or “What is truth?” can be answered on the basis of one’s grasp of the relevant concepts, and the ideal result is an analysis formulated as a biconditional that states necessary and sufficient conditions. This method can be traced back to ancient Greece. Socrates believed that in order to get clear about some concept we had to find the element(s) necessarily common in all its application conditions. The most famous analysis of this form is the “justified true belief’ theory of knowledge, which Edmund Gettier (1963) attributes to Plato, A. J. Ayer, and Roderick Chisholm. According to this view, for any subject S and any proposition p, S knows that p if and only if S has a justified true belief that p. Knowledge is therefore understood in terms of three individually necessary and jointly sufficient conditions: jus tification, truth, and belief. I will call this the “traditional analysis of knowledge.” Today it is widely accepted that Gettier refuted the traditional analysis of knowledge in 1963.4 He described two cases in which a person satisfies the three putative conditions for knowledge and yet this person does not seem to qualify as a knower, thereby demonstrating that these conditions are unsatisfactory. In subsequent decades much philosophical work was devoted to finding a Gettier-proof analysis of knowledge in terms of true belief plus some other factor (or factors). While these solutions differ in their details, they all presumed the traditional account of knowledge was roughly right but in need of modification to deal with Gettier-type counterexamples. Even critics of the proposed solutions to the so-called Gettier problem tended to leave open the possibility that some other set of conditions were necessary and sufficient for knowledge. A guiding pre sumption at this time was that an adequate theory of knowledge must pro vide a list of such conditions for knowing. As Timothy Williamson writes, 4 Bertrand Russell unintentionally provided a counterexample to the traditional analysis of knowledge in 1948. In Human Knowledge: Its Scope and Limits, he describes a scenario in which a subject forms an (accidentally) justified true belief that it is two o’clock on the basis of looking at a clock that stopped twelve hours ago. Russell takes this as a counterexample to the proposal that knowledge is true belief, but he did not seem to notice that the agent’s belief is also justified.
16
|
What’s the Point of Knowledge?
“the received idea is that we can conceptualize the factors whose conjunc tion with belief is necessary and sufficient for knowledge independently of knowledge” (2000: 3). In a similar vein, Alan Millar says: It has been widely assumed over the last fifty years or so that epistemology, so far as it addresses the nature of knowledge, should issue in a concep tual analysis of knowledge-that—a statement of necessary and sufficient conditions for the truth of instances of the schema X knows that p in terms that do not implicate the concept of knowledge. (2007: 197)
Philosophers therefore searched for an “anti-Gettier condition”—a con dition that must be added to true belief, or justified true belief, to have knowledge. Were this found, it would presumably give us a complete anal ysis of knowledge. For better or worse, all attempts to analyze knowledge in this way have succumbed to a pattern of counterexamples. This point is well known, so I will not survey the many failed attempts to analyze knowledge in this way.5 Instead, I will suppose (not entirely uncontroversially) that no satis factory reductive analysis of knowledge is currently available. What, then, are the prospects for achieving an analysis of knowledge? If Williamson is right, the prospects are dim. In Knowledge and Its Limits, Williamson considers several reasons why we might suppose that knowledge must be analyzed into conceptually prior components (e.g., belief, truth, etc.), but he finds them all wanting. Further, there are two powerful inductive arguments against reductive conceptual analysis. First, the intractability of the traditional project provides us with a good reason to think that whatever the next proposed analysis might be, sooner or later a counterexample will emerge. Second, this research program has been unsuccessful in areas of philosophy outside of epistemology. Indeed, philosophers have been unable to provide convincing analyses of almost any philosophically interesting concept. As Stephen Stich claims, “Ac? commonsense concept that has been studied has turned out to be analyzable into a set of necessary and sufficient conditions” (1992: 250). Similarly, Jerry Fodor says that “the number of concepts whose analyses have so far been determined continues to hover stubbornly around none” (2003: 6). And William Lycan reminds us that “no effort of analytic philosophy to
5 Others have done that adequately. See Shope (1983), Fogelin (1994), and Lycan (2006) for a review of this literature.
METHODOLOGIES IN EPISTEMOLOGY
|
17
provide strictly necessary and sufficient conditions for a philosophically interesting concept has ever succeeded” (2006: 150). At present, there is no convincing reason to expect knowledge to be analyzable into logically necessary and sufficient conditions. A growing number of philosophers are therefore distancing themselves from the tradi tional and voluminous attempts to analyze knowledge this way. Instead of assuming that traditional analyses were basically on the right track, which is a thought Wittgenstein believed “shackled philosophical investigations” (1964:26-27), philosophers are increasingly skeptical about reductive con ceptual analysis. Williamson calls it a “degenerating research programme” (2000: 31) and Craig suggests we set aside this method in favor of an “al ternative angle” (1990: 2). This book is an attempt to develop a rigorous way of doing episte mology outside the paradigm of the traditional view. Unlike traditional analysis, the goal of function-first epistemology is not to enumerate the necessary and sufficient conditions of our epistemic concepts. It is dan gerous to presume that our concepts must be analyzed in this way, since this can prove too restrictive in our theorizing. As a function-first epistemologist, I am interested in identifying what might be called the “core” of our epistemic concepts and practices. This allows me to specify conditions that hold only for the most part, but not always. An advantage of this approach is that, unlike reductive conceptual analysis, we need not ignore important characteristics of our epistemic concepts. On the traditional approach, a theory will be rejected in the face of any counterexample, irrespective of how rare or odd that counterexample might be. For example, if one were to devise a strange case involving a subject who knows that p without believing that p, then belief would simply vanish from our theory of knowledge. It would vanish despite the fact that people generally believe what they know.6 We function-first epistemologists, however, do not give such enormous power to these al leged “counterexamples.” Rather, we allow features that are extremely important, but not obviously necessary, to play a crucial role in our epis temological theorizing. This is a theoretical asset because philosophers should be open to the possibility that explaining the nature and value of our epistemic concepts, norms, and practices will partly come from cer tain contingent characteristics. And yet this idea seems unthinkable on the traditional approach.
61 take this example from Craig (1990: 14).
18
|
What’s the Point of Knowledge?
While a function-first epistemologist will not presume that our epi stemic concepts are analyzable into necessary and sufficient conditions, this methodology is still perfectly compatible with that goal. Reflecting on the role of some epistemic concept, and the conditions that would govern its application, might ultimately issue in necessary and suffi cient conditions. Who knows? But regardless of how things turn out, a benefit of this approach is that it would explain why a certain epistemic concept has the features it does have. We should simply regard it as an open question whether an account of our epistemic concepts will fit comfortably into a list of necessary and sufficient conditions (see Craig 1990: 5). It is surprising that many scholars have misinterpreted Craig on this issue. Unlike Williamson, Craig does not think our worries about tradi tional analysis are sufficient to mandate the rejection of this approach. Rather, he takes the problems facing traditional analyses to justify con sidering alternative approaches. But some commentators have mistak enly interpreted Craig as proposing that epistemologists should “eschew” (McBain 2004: 193) or “abandon” (Jones 1997: 435) traditional anal ysis, or that “the project of analysis is hopeless and should be replaced” (Ichikawa and Jenkins 2017). This reading is too strong, for Craig nowhere recommends that we abandon the traditional way of analyzing knowledge. He simply thinks we should consider other options. To be charitable, let’s temporarily suppose that knowledge can be analyzed in the traditional way. Even if this were possible, there is still a good reason to think traditional analyses are inadequate. Craig makes this point:
Let us suppose, however optimistically, that the problem of the analysis of the everyday meaning of ‘know’ had both been shown to exist and subse quently solved, so that agreed necessary and sufficient conditions for the as cription of knowledge were now on the table. That would be a considerable technical achievement and no doubt a long round of hearty applause would be in order, but I hope that philosophers would not regard it as a terminus, as many writers make one feel they would. I should like it to be seen as a prolegomenon to a further inquiry: why has a concept demarcated by those conditions enjoyed such widespread use? (1990: 2)
In other words, even if the project of reductive analysis were successful, it would leave unanswered some significant questions in our epistemological theorizing—questions that a function-first approach tackles head on.
METHODOLOGIES IN EPISTEMOLOGY
|
19
Panayot Butchvarov considers this the most characteristic symptom of the inadequacy of traditional accounts of knowledge. He says that “an ade quate account of the concept of knowledge must display its essential place in the conceptual framework through which we would most perspicuously understand ourselves, our life, and the world in which we live” (1970:2526). Such an account must address the question of how, and why, if we lacked the concept of knowledge, we would introduce such a concept. Yet traditional accounts of knowledge, even if correct, fail to show why knowledge should have the features it has (e.g., truth, justification, and be lief), how these are related to one another, and why these must be present for knowledge. There is something deeply unsatisfying with an account that leaves these questions unanswered. I hope to explain why our concept of knowledge has the features it does and why this concept plays a crucial role in our fife.
1.3 Knowledge-First Epistemology A function-first epistemologist will not analyze knowledge in the tradi tional way, but the traditional approach still shapes her research program. To illustrate this idea, recall the central hypothesis of this book (which I’ll defend in the next chapter): the primary function of the concept of knowl edge is to identify reliable informants. This presupposes that true belief is more basic than knowledge and must be explicable independently of the concept of knowledge. But this very starting point is controversial. In Knowledge and Its Limits, Williamson argues that it is mistaken to stipulate that inquirers want “true beliefs about our environment, as though this were somehow more basic than our need for knowledge of the environment” (2000: 31). Williamson is not necessarily proposing an alternative methodology to mine, but our views have sufficiently different starting points to make it worth considering his view here. In particular, my account is predicated on the conceptual development of true belief to ward knowledge, whereas Williamson’s “knowledge-first” account leaves no room for such conceptual development in logical space. Knowledge, says Williamson, is primitive: it cannot be analyzed in terms of more basic components like truth, belief, and justification. Epistemologists should instead take knowledge as basic and use it to understand justification, belief, and other things. Thus, Williamson would not characterize an in quirer as someone who wants a true belief as to whether p, but rather as someone who wants to know whether p. Characterizing inquiry this way
20
|
What’s the Point of Knowledge?
is incompatible with my approach. If the concept of knowledge were em ployed in selecting the conditions of a reliable informant, as Williamson claims, this would cripple my account.7 Craig acknowledges a similar worry. He asks, “Shouldn’t being a good informant be explained in terms of knowing whether p, rather than the other way around?” (1990: 94). In other words, doesn’t one need to know before qualifying as a good informant? Knowing seems primary, whereas being a good informant seems to require that one be a knower plus some other factors.8 The objection that true belief is not conceptually prior to knowledge, and that we cannot understand knowledge in terms of true belief, is an im portant one. If this were correct, then my proposal would be constructed on a shaky foundation. I find Williamson’s proposal fascinating, but an ad equate discussion of his view would take us too far afield. This is because his claim that knowing is a simple and general factive mental state is not based on any straightforward, knockdown argument. Rather, it relies on an overall package of related theories along with several considerations against various grounds for resistance. For this reason, I will not provide a direct reply to Williamson’s proposal. But this need not bring our in vestigation to a halt. Having acknowledged this dissenting voice, we may show it to be doubtful by developing a plausible alternative. I will argue that by starting with the idea of an inquirer who wants true beliefs, we are ultimately led to a comprehensive account of the nature and value of knowledge. This approach has at least two advantages over Williamson’s. First, Williamson presents his view as arising from the ashes of the current and inadequate approaches to knowledge. It is because he thinks these approaches run aground on important considerations that he is motivated to try his radical alternative. This means we can remove some of the mo tivation for Williamson’s knowledge-first approach by showing that epis temological progress is possible without treating knowledge as a s mental state that cannot be understood in terms of more basic epistemic notions. As I will argue, knowledge is best understood in terms of the idea of a good informant, and we get at the nature of knowledge by reflecting on the criteria for being a good informant. A second advantage of my approach 7 In some ways, function-first epistemology is compatible with the idea that knowledge is primitive; for instance, we could try to explain the roles and value of the (primitive) concept of knowledge. However, it is difficult to see how Williamson could opt for anything less than a rejection of the view I develop, since I treat true belief as more basic than knowledge. 8 Versions of this objection can be found in Schmitt (1992), Feldman (1997), and Fricker (2015).
METHODOLOGIES IN EPISTEMOLOGY
|
21
is that, unlike Williamson’s knowledge-first account, function-first episte mology does not leave unanswered why the concept of knowledge would have arisen in the first place. This marks a deep divide between Williamson and me. For Williamson, the explanation why we have a concept of knowl edge will derive from the properties of knowledge itself. In contrast, I am skeptical about the idea that knowledge exists independently of our epi stemic practices. In short, Williamson is a staunch antipragmatist, whereas I will defend a pragmatist view of knowledge. While I do not treat knowledge as an unexplained explainer, I agree with Williamson that knowledge is indispensable—or at least deeply important—for humans. In chapter 5,1 will defend the value of knowledge by drawing on Williamson’s work on the relations between knowledge, assertion, and practical reasoning. I hope to show that function-first epis temology has more in common with knowledge-first epistemology than it might at first seem.
1.4 Reverse Engineering Epistemic Evaluations Another approach is to “reverse-engineer” our epistemic evaluations. This strategy initially brackets speculation about what our epistemically evalua tive practices axe for and instead looks at what these practices actually are. In other words, we start by looking at how people actually use epistemi cally evaluative terms and then we draw conclusions about the function of epistemic evaluation from facts about usage. J. L. Austin might be regarded as the forerunner of this approach. Perhaps more than anyone, Austin believed we should begin our study of knowledge by attending to the way the word “knows” is ordinarily used.9 Sinan Dogramaci (2012) is a more recent proponent of this methodology; he investigates the use of words like “rational” and “unreasonable” in ordinary epistemic evaluations. On Dogramaci’s view, these evaluations have an overall tendency to influence our audience to follow the endorsed belief-forming rules, which promotes coordination and makes testimony trustworthy. I do not want to exaggerate the differences between function-first epis temology and the method of reverse engineering epistemic evaluations. Indeed, there is a sense in which reverse engineering may be regarded as part of a broader, more holistic process that I am calling function-first epistemology. I will say more about this shortly. For now, I simply want
9 Austin will come to play a key role in this book, especially in chapters 7 and 8.
22
|
What’s the Point of Knowledge?
to point out that function-first epistemology and reverse engineering have different starting points. Someone who wishes to “reverse-engineer” ep istemic evaluation will first look at some aspect of our actual evaluative practice and then try to infer how this practice benefits us. A function-first epistemologist, in contrast, will reverse the direction of investigation: we start with some prima facie plausible hypothesis about the role of some epistemically evaluative aspect of our language, and then we examine the extent to which this matches our everyday notion. Why the different starting points? Isn’t it more reasonable to first look at our ordinary practices of epistemic evaluation and then hypothesize about the function of those practices? Indeed, you might doubt whether it is even possible to formulate a plausible hypothesis about the function of some epistemic concept without first thinking about how people actually use epistemically evaluative terms. This raises an important challenge: why should we take function “first” if we can simply reverse-engineer our way to the purpose of epistemic evaluation? While reverse engineering certainly has its merits, this approach faces some important limitations. First, anyone who attempts to reverse-engineer epistemic evaluations must assume that we can glean the point of our epistemic concepts from our linguistic practice featuring the words that express those concepts. But this strategy is doubtful for familiar reasons. Consider the following comparison:
Why do we say that things are on fire? Well, sometimes to scare people, sometimes to encourage people to escape the burning building, sometimes to celebrate (if we are resistance fighters having just succeeded in putting the enemy’s property on fire). (Kappel 2010: 71)
Something similar may hold for our use of the words “knows,” “understands,” and cognate expressions. As Kappel points out, the set of aims manifested when saying that someone knows something may tell us little about the point of the concept of knowledge because we may use it for diverse purposes once we have it. Suppose the concept of knowledge is used to flag good informants. Even if this were correct, we might also ascribe knowledge for a variety of other purposes. For instance, I might try to comfort a friend who is experiencing hardship by saying, “I know you’ll get through this.” I might say this even though I realize I do not know this. My use of “knows” in this case is intended to provide reassurance, not to identify a reliable informant. Thus, we cannot make any simple inferences
METHODOLOGIES IN EPISTEMOLOGY
|
23
from the way words are used in ordinary language to conclusions about the point of the concept in question. This is because the ways in which “knows” is used are too various. We can, however, reflect on the deeper reasons why humans might need a concept to help share information and solve coordination problems; then we can compare our use of “knows” with this hypothesis and ask whether this concept can be used for addi tional purposes. This is precisely what the function-first epistemologist suggests we do. You might think this objection backfires on my view. If people ascribe knowledge for many reasons, why should we think the concept of knowl edge has any sort of distinctive function? Perhaps function-first episte mology rests on a fundamental error by assuming there is anything like the point of the concept of knowledge. Obviously we can use things—including words and concepts—for many purposes. It is important, however, to distinguish what the function of something is as opposed to what it may function as. A hammer, for ex ample, can be used as a paperweight or a weapon even though the point of a hammer is to drive nails into objects. Tools can always be used for diverse purposes once we have them. But to say the point of a hammer is to drive nails into objects is to say that hammers have their current shape in order to achieve this end. Indeed, the very existence of hammers likely depends on the fact that they help to satisfy this particular human need. This explanation is not undermined by the fact that a hammer can be used in many ways, or for diverse ends. The point is that hammers likely wouldn’t be able to serve these other purposes if they did not func tion as a means to drive nails into objects. It is the fact that hammers are designed for this function that explains why they have a certain weight and shape, and these qualities explain why hammers may also be used as paperweights and weapons—not the other way around. Something similar can be said about the concept of knowledge. This concept exists because it helps us satisfy certain basic human needs. As I will argue in this book, the point of the concept of knowledge is to iden tify reliable informants to members of our community. This does not pre sume that some actual human beings intentionally designed the concept of knowledge for this purpose. I am giving an abductive argument that shows we can reconstruct familiar epistemological practices on the basis of our hypothesis about the purpose of the concept of knowledge. This provides us with good evidence that we have the concept because it satisfies the relevant need. Once this concept enters our conceptual repertoires, how ever, we may use it in diverse ways. But it is doubtful that it would serve
24
|
What’s the Point of Knowledge?
any of these other purposes if it did not satisfy its primary function. When I speak of the “point “purpose,” “role,” or “function” of some epistemic concept, it is this sense of purpose that I have in mind. I will elaborate on this idea in chapter 5. This brings us to another limitation of some current methodologies in epistemology. Methodologies that rest heavily on linguistic usage or in tuitive judgments about cases, such as reverse engineering and traditional conceptual analysis, will only end up describing our actual practices, not characterizing those that would be best for us. These descriptive approaches might be useful for other goals, such as explaining aspects of human cognition. But what is unsatisfying about these approaches is they simply assume the propriety of our ordinary practices of epistemic evaluation; they do nothing to say why we should think these practices are worthy of our endorsement. In contrast, function-first epistemology goes beyond the project of merely describing our current epistemic practices and concepts: it allows us to engage in the normative project of evaluating how well or poorly our epistemic practices actually satisfy our needs and goals. We might think of this as conceptual re-engineering. To re-engineer an epistemic concept, we must first consider what this concept does for us. In this way, the method of reverse engineering is part of a broader nor mative project. Nevertheless, this normative project will require us to take function “first” in the sense that we must assess our current practices in light of the roles that epistemic evaluation is supposed to serve. Only then can we adequately aim to improve our practices. Still, a function-first epistemologist should acknowledge that epistemo logical theorizing must be constrained by our folk epistemology—viz., by how we ordinarily think and speak of knowing, understanding, and so forth. Our aim is to construct a concept that functions in the manner suggested by our hypothesis and that matches our intuitions. After hypothesizing why we have a certain epistemic concept, we must compare whether the ordi nary situations in which this concept is applied actually cohere with the putative role of this concept. In this way, my methodology allows for epi stemic normativity without being indifferent to how our results tally with ordinary language, our judgments about cases, and our actual epistemic practices. Our hypothesis is measured against our folk epistemology, al though these considerations enter our theorizing at a later stage. But if theorizing is constrained by our actual practices only at this later stage, will our initial hypothesis be too speculative? One might worry that nothing prevents us from engaging in wild conjecture about the purpose of epistemic evaluation.
METHODOLOGIES IN EPISTEMOLOGY
|
25
It would of course be unwise (and perhaps impossible) to completely ignore aspects of our folk epistemology when formulating our initial hy pothesis. For our hypothesis to be even remotely plausible, it must be in formed by our preexisting judgments—otherwise we might hypothesize that knowledge is clam chowder. Thus, it would be a mistake to think we can formulate a plausible hypothesis about the role of some epistemic con cept wholly independently of our folk epistemology. We should think of our inquiry as balancing considerations about the perceived role of some epistemic concept with our folk epistemic judgments. That said, these pre existing judgments do not play as central a role in a function-first episte mology as they do in other methodologies, such as reverse engineering, traditional conceptual analysis, and (as I’ll argue in the next section) epis temological naturalism. I use the label “function-first” to highlight the rel ative importance of functional considerations in our theorizing, but I do not assume these are the only relevant considerations. We should refine our understanding of a concept’s function in light of other constraints, treating the whole process as broadly abductive (see Henderson and Horgan 2015). For this reason, function-first epistemology and reverse engineering might complement each other in certain ways. We may appeal to additional constraints that will prevent us from slip ping into mere conjecture. In particular, our hypothesis must be compat ible with certain facts about human life, such as facts about our physical environment, our social organization, our cognitive capacities, and the basic aims and interests humans typically have. These facts about humans and their circumstances will then give rise to a certain conceptual need that is supposed to be satisfied by the purpose described in our hypothesis. These remarks are fairly abstract, so let’s get more concrete. Why do humans think and speak of “knowing”? According to Craig, with whom I agree, the concept of knowledge is used primarily to identify reliable informants. This hypothesis isn’t just wild speculation. It is based on the following plausible assumptions.10 First, humans need true beliefs about their environment in order to successfully guide their actions. After all, it is undeniable that without true beliefs, it would be much harder to find food, shelter, and achieve our goals. Second, we require sources of infor mation that will lead to true beliefs. Third, the easiest and most efficient way to acquire a true belief is often to ask someone reliable. Only a very small part of the information we receive comes to us firsthand, since others
10 See Craig (1990: 11).
26
|
What’s the Point of Knowledge?
are often in a position to access more truths than we ourselves are. Fourth, on almost any issue some people will be more likely than others to provide us with accurate information. My doctor, for example, is a better source of medical information than my car mechanic. Beyond cases of special ized expertise, people vary in terms of their reliability in myriad ordinary domains. For example, if you are looking at the television while I am in the kitchen, then you are in a better position to see what just happened in the movie. Bernard Williams calls this “purely positional advantage” (2002: 42), an example of which is looking at the right place at the right time. For all these reasons, we may presume that any community will have an interest in evaluating sources of information. The concept of knowledge is believed to arise in connection with this interest. We may use these highly plausible considerations to set constraints on our initial hypothesis about the function of the concept of knowledge. I will discuss these points in more detail in the next chapter.
1.5 Epistemological Naturalism Function-first epistemology, as I have characterized it, is a kind of con ceptual inquiry: it studies our epistemic concepts. According to many philosophers, however, the idea that philosophy consists in an investigation of our concepts is deeply mistaken. Michael DePaul, for example, says, “I am not concerned with questions about the value of concepts . . . my focus will instead be on the value of the things that fall in the extensions of concepts” (2009: 114-115). Likewise, Williamson claims that “the subject matter of much philosophy is not conceptual in any distinctive sense. Many epistemologists study knowledge, not just the ordinary con cept of knowledge” (2007: 211). And Hilary Komblith, the foremost defender of this idea in epistemology, says, “the subject matter of epis temology is knowledge itself, not our concept of knowledge” (2002: 1). But what exactly is the difference between our concept of knowledge and knowledge itself? This distinction is not often remarked upon, and many epistemologists slide between the two claims.11 Perhaps this is because many theorists assume that investigating one’s concept of knowledge is the same as investigating knowledge itself (the aforementioned authors11
111 have already done this several times, and Ichikawa and Jenkins (2017) point out that Williamson does this in Knowledge and Its Limits.
METHODOLOGIES IN EPISTEMOLOGY
|_
27
aside). Or perhaps many believe that reflecting on the concept of knowl edge will reveal the nature of knowledge. Komblith most clearly articulates why the subject matter of epistemo logical investigation is not (or should not be) the concept of knowledge but rather the phenomenon of knowledge. I will therefore focus on his view. Kornblith’s central claim is that knowledge is a natural kind and so it should be studied like any other natural phenomenon. In an illuminating comparison, he writes:
What we are doing, as I see it, is much like the rock collector who gathers samples of some interesting kind of stone for the purpose of figuring out what it is that the samples have in common. We begin, often enough, with obvious cases, even if we do not yet understand what it is that provides the theoretical unity to the kind we wish to examine. Understanding what that theoretical unity is is the object of our study, and it is to be found by careful examination of the phenomenon, that is, something outside of us, not our concept of the phenomenon, something inside of us. In short, I see the in vestigation of knowledge, and philosophical investigation generally, on the model of investigations of natural kinds. (2002: 10-11) This model of scientific inquiry has important ramifications for philo sophical methodology. Once we stop analyzing the concept of knowledge in favor of investigating knowledge itself, it is no longer appropriate to rely on some common methodological practices in epistemology; in par ticular, we should not compare accounts of knowledge with our intuitions about various imaginary cases, nor should we consider whether we would be inclined to say that someone does or does not have knowledge in var ious circumstances. After all, it would not be a good idea for a rock col lector to investigate rocks by examining his concept of rocks, since his concept may be incomplete or build in mistakes. Proper scientific proce dure suggests that we should look at natural objects themselves (rocks, gold, water, etc.), not our concepts of these things. Likewise, instead of looking at thought experiment judgments and linguistic considerations, the proper way to investigate knowledge, according to Komblith, is empirically. This type of epistemological naturalism raises a bevy of interesting questions. What are natural kinds? Why think knowledge is a natural kind? What is the proper role of intuitive judgments in philosophical inquiry? Can we learn anything useful about knowledge by studying our concept of knowledge? Komblith has answers to all these questions. In what follows,
28
|
What’s the Point of Knowledge?
I will highlight only the most significant points of agreement and disagree ment between our views. A function-first epistemologist will agree with Komblith that intuitive judgments have been given too central a role in epistemological theorizing. Instead of resting on our judgments about the extension of the word “know,” a function-first epistemologist will start with some prima facie plausible hypothesis about what the concept of knowledge does for us. This is not to suggest that our untutored judgments have no role to play in episte mological investigation. On my approach, ordinary epistemic judgments are used to test the merits of one’s hypothesis about the role of the con cept in question; on Komblith’s naturalistic approach, our intuitions about knowledge are used to locate paradigmatic cases that serve as the objects of study.12 Neither view can do without intuitions entirely. I suspect, how ever, that intuitive judgments play afar less significant role in Komblith’s epistemology than my own. As mentioned previously, function-first epis temology is heavily anchored by the everyday concept we are looking to explicate. In contrast, Komblith allows that knowledge might have turned out to be significantly different from what we intuitively thought. Still, function-first epistemology resembles Komblith’s naturalism in at least two ways. First, both approaches appeal to an evolutionary story about how knowledge is the product of the needs and interests of creatures in their environments. Second, the notion of reliability plays a central role in both our theories of knowledge. Following Craig, I maintain that the concept of knowledge is used to flag reliable informants. Komblith says knowledge is reliably produced true belief. Despite of these similarities, however, there is a much deeper disagreement between us. According to Komblith, the proper subject matter of epistemological investigation is not the concept of knowledge but rather knowledge itself which is a natural kind. For those interested in the nature of knowledge, he claims, “there may be no reason at all to care about our concept of knowledge” (Komblith 2011: 50). If knowledge were a natural kind, then function-first epistemology would be an inadequate method to investigate it. To illustrate this idea, con sider the difference between water and the concept of water. It seems clear that the concept of water serves an important purpose, namely satisfying 12 Komblith considers the case of an early chemist who wishes to determine what various acids have in common, but his samples of would-be acids include shoes, a ship, sealing wax, and his pet dog (2005: 428-429). Clearly this investigator is not engaged in the same process as an early chemist who has good samples of acids, even if some of them are mistaken. Thus, some reliance on intuition is required even to study natural phenomena.
METHODOLOGIES IN EPISTEMOLOGY
|
29
our need to think and talk about water. But it would be a mistake to inquire into the nature of water by asking what work this concept does for us, what function it serves. As Frederick Schmitt (1992) points out, this method would lead one to conclude that being identifiable by human beings for purposes of drinking and bathing is essential to the nature of water, and of course this is false. Why, then, should we expect to understand the nature of knowledge by inquiring into the function of the concept of knowledge in human life? Couldn’t it just be that knowledge is a natural phenomenon, like gold and water, and the purpose of the concept of knowledge is to think and talk about it? Many theorists have simply taken for granted that we should investigate our epistemic concepts without explaining why our investigation isn’t better aimed at knowledge itself. Indeed, many theorists don’t even explain what this difference amounts to. Knowledge, like many resources, can be acquired, used for diverse purposes, and lost. But there are important differences between knowledge and natural resources like water, rocks, gold, and atoms.13 As Jennifer Nagel (2014) observes, knowledge has a much closer connection to us than resources like water or gold, since the continued existence of knowledge depends on the existence of knowers. Unlike water and gold, knowledge would not continue to exist if sentient life were wiped out. Admittedly, this doesn’t conclusively show that knowledge is not a natural kind, since there may be natural kinds that would not exist if sentient fife were wiped out (e.g., biological species or certain mental states). Nagel’s point is just that knowledge, unlike gold or water, must always belong to someone. Why does Kornblith think knowledge is a natural kind? He says our best scientific theories will guide us as to which natural kinds there are, and knowledge plays a central causal role in certain scientific theories, specifically those of cognitive ethology. As he writes,
Gold is a natural kind if a proper chemistry is committed to it; genes are a natural kind if a proper biology is committed to them. Similarly, I argue, current work on animal behavior (cognitive ethology) makes use of a no tion of knowledge; knowledge plays an essential explanatory role in these theories. (2005: 400)
Knowledge, according to Kornblith, is needed to explain the behavior and psyche of cognitive agents. Drawing on cognitive ethology, he argues
13 BonJour (2006), Millar (2007), and Pemu (2009) compellingly argue that knowledge is not a natural kind.
30
|
What’s the Point of Knowledge?
that knowledge plays an ineliminable role in explaining successful behav ior of animals, such as piping plovers and dolphins. Unlike Williamson, Komblith does not say that knowledge plays a vital explanatory role at the level of individual action—mere belief would suffice for that. Rather, knowledge is essential for explaining a behavioral capacity that contributes to the survival of a species in its environment (Komblith 2002: 57). This is because false beliefs misrepresent the environment and would not typi cally be conducive to survival. There are a variety of problems with this argument. Komblith’s entire case for taking ethologists at their word rests on the alleged explanatory and predictive power of imputing knowledge to nonhuman animals, but Komblith doesn’t establish the legitimacy of this claim (see Roth 2003). Putting this point aside, we might wonder why cognitive ethology is the best science involving knowledge. According to this area of science, it might be true that “knowledge” picks out reliably produced true beliefs. But why not plunge into a different field? Jonathan Weinberg (2006: 30) suggests psychology or informatics, to name a couple. In either case, one potentially gets a different analysis of knowledge. Indeed, one’s theory of knowledge might look very different depending on the area of science to which one looks for guidance, yet cognitive ethology is the only area of research to which Komblith pays attention. Presuming we get a dif ferent analysis of knowledge in other fields, this creates a problem for Komblith’s claim that knowledge is a natural kind. The idea that knowledge is not a natural kind is widely endorsed. According to Quassim Cassam, “There are too many disanalogies between knowledge and genuine natural kinds for this to be plausible” (2007: 127). Likewise, Stewart Cohen says the investigation of knowledge is “not a straightforwardly empirical matter” (2005: 67). But if knowledge is not a natural kind, what is it? The most plausible alternative is that knowledge is a social or artifactual kind. In other words, knowledge is not something we find in the world, like rocks or water, but rather something we impose on it. I will now say a few words about what artifactual kinds are, but what follows is nothing like a full theory of such kinds. It is widely agreed that artifactual kinds reflect our needs and interests. Accordingly, the identity conditions for them are intertwined with human intentions. A natural kind, in contrast, has nothing to do with human purposes, intentions, interests, or decisions. Consider the case of a sports utility vehicle (SUV), which is an example I take from Komblith (2006). Our culture divides vehicles into a number of different kinds—station wagons, SUVs, and so forth. A culture with the very same vehicles might
METHODOLOGIES IN EPISTEMOLOGY
|
31
have categorized them differently, but this would merely reveal that their interests and concerns are different from ours. There is no preexisting fact in nature about how to categorize these items. In this way, being an SUV is a property whose very existence depends upon certain social customs or conventions, like marriage and money. It is not something we discover in the natural world; humans create it for certain purposes. More generally, what determines whether or not something is a social or artifactual kind pertains to whether that thing fulfills certain purposes.14 Of course natural kinds like gold can also play social roles, but the point is that knowledge, unlike gold, exists to play a certain social role (or roles). While gold is able to play certain social roles because of its natural properties, knowledge has the properties it has because of the social role it plays. In other words, knowledge’s existence depends on its ability to re alize certain human interests. This is not to say that social kinds consist of sets of human intentions and purposes, as Kristoffer Ahlstrom (2008:113) reminds us. Rather, it simply means that what determines whether or not something is an instance of a particular artifactual kind depends on whether that thing answers to a specific set of human intentions. As I’ll be arguing in this book, this is precisely what provides theoretical unity to our notion of knowledge. Instances of knowledge owe their membership to the fact that they fulfill a certain purpose, namely the identification of reliable informants. The term “knowledge” picks out a kind with an important and interesting theoretical unity to it, but it is not unified by some natural phe nomenon. Rather, knowledge turns out to be a type of social phenomenon. On such a view, the gap between knowledge and the concept of knowl edge gets blurred, if not eviscerated. Just as it makes no sense to wonder whether our culture’s concept of an SUV properly answers to the true nature of SUVs, it makes little sense to distinguish the attempt to be come clearer about our concept of knowledge from the attempt to become clearer about knowledge as such. There is no sharp contrast here. We get at the nature of knowledge by investigating the concept precisely because knowledge is a socially constructed kind that acquires its content in virtue of its function. Thus, I will largely ignore the distinction between analyses of our epistemic concepts and theories of the things themselves. While this distinction clearly holds in many cases, it is not clear in the case of knowl edge whether one can draw an interesting distinction between an adequate
14 For a detailed suggestion along these lines, see Ahlstrom (2008) and Thomasson (2003).
32
|
What’s the Point of Knowledge?
analysis of the concept and an adequate theory of the thing in question.15 As A. J. Ayer writes,
It is . . . indifferent whether, in this manner of philosophizing, we represent ourselves as dealing with words or as dealing with facts. For our inquiry into the use of words can equally be regarded as an inquiry into the nature of the facts which they describe. (1956: 29) That said, Komblith’s appeal to cognitive ethology raises an important question: must a theory of knowledge account for (nonhuman) animal knowledge? There is a large literature on animal cognition, and the experts in this field often speak of animals knowing things. It seems there may well be a notion of animal knowledge that is used by cognitive ethologists for their purposes, but there is also an everyday notion of knowledge that answers to general human interests—and it is this latter notion that interests me and most other philosophers. I will return to this issue in chapter 5.
1.6 Concluding Remarks This chapter has outlined function-first epistemology, compared it with some alternative approaches, and highlighted some of the benefits of this methodology. The guiding idea is that we can shed light on the nature and value of knowledge (and other epistemic concepts) by reflecting on its function or purposes. I compared this approach to four alternatives in order to highlight some of the distinctive features of function-first epistemology. For example, this approach is perfectly compatible with the goal of reduc tive conceptual analysis, and yet it does not presuppose that knowledge must be analyzed in the traditional way. It also creates room for genuine epistemic normativity by allowing us to evaluate our current practices in light of the roles that epistemic evaluation is supposed to serve. Along the way, I attempted to answer some potential objections (e.g., why should we care about the concept of knowledge rather than the phenomenon of knowledge?) as well as identify issues that will be discussed later in the book (e.g., why think knowledge has just one function?). This book is not just about metaepistemology. I plan to use the method of function-first epistemology to make progress on a number of important issues; for instance, I hope to increase our understanding of the semantics
15 See DePaul (2009: 124—125) for similar remarks.
METHODOLOGIES IN EPISTEMOLOGY
|
33
of knowledge ascriptions, the Gettier problem, skepticism, and the rela tionship between knowledge, assertion, and practical reasoning. I also want to cast light on issues in social epistemology such as the division of epistemic labor and group knowledge. In order to make good on these promises, however, I must now set aside the job description of functionfirst epistemology and move on to the job itself. What is the point of epi stemic evaluation? More specifically, why do humans think and speak of knowledge?
34
|
What’s the Point of Knowledge?
chapter 2 | The Point of Knowledge
It would be useful for those who believe that the concept of knowledge has an important role to play in epistemology to do more than, I believe, has already been done to explain just what that role is. —HILARY KORNBLITH (2OO5: 431)
the heart of this book is a simple but powerful idea: the point of the concept of knowledge is to identify reliable informants.11 will often refer to this as “our central hypothesis.” This hypothesis is rooted in the idea that epistemic evaluation helps us satisfy some basic human needs. In particular, all humans have a need for enough truths (and not too many falsehoods) to help us satisfy other basic needs, such as our need for water, food, and shelter. Put differently, true beliefs have survival value. This is not to suggest that seeking the truth is always motivated by practical concerns or that false beliefs are never useful. False beliefs are sometimes biologically or psychologically adapt ive; further, we may want to know some things simply to satisfy our nat ural curiosity. My point is just that we clearly require true beliefs in order to successfully get by in the world. How else would we reliably find our way home?1
1 Craig describes these informants as “good” rather than “reliable,” but reliability is clearly what he has in mind. For example, he says a knower is “someone with a very high degree of reliability, someone who is very likely to be right—for he must be acceptable even to a very demanding inquirer” (1990: 91). A number of theorists who have taken up Craig’s view prefer to speak in terms of reliability (e.g., Williams 1973, Lackey 2012, Pritchard 2012, and McKenna 2013). I will use “good informant” and “reliable informant” interchangeably. A good informant is reliable in a counterfactual sense; viz., a good informant as to whether/? believes that/; ifp is true, and believes that not-/; if not-/; is true, across those possible worlds that are relevant to the predicament of the inquirer (Fricker 1998: 162). I will elaborate on this point in chapters 2 and 3.
This need for truth (and the avoidance of error) encourages us to re alize the benefits of pooling and sharing information. As a lot recent work in evolutionary theory and the epistemology of testimony makes clear, humans are heavily epistemically dependent on each other. Thus, it is to our collective advantage that we pool and transmit information. This prac tice of sharing information gives rise to another need: we must distinguish reliable informants from unreliable informants. With respect to almost any issue, people vary in terms of their reliability as sources of information. For example, if you want to figure out why your cat is sick, you should ask a veterinarian and not a street psychic. If you want to know whether the Yankees clinched a playoff spot, you should ask a baseball fan. To build up our stock of accurate information, we must mark out those people on whom we should rely from those we should not. The concept of knowledge emerges in connection with these basic needs. Our starting point is the idea that the concept of knowledge functions as a marker for reliable informants, which is an important role in the general economy of our concepts. I will try to convince you that this hypothesis has wide-reaching implications.
2.1 Reliable Informants as Knowers Earlier statements of this idea can be found in the writings of Ernest Sosa and Bernard Williams. In his 1974 paper “How Do You Know?” Sosa says that knowledge has a “social aspect” and that certain departures from the traditional account [of knowledge] may make better sense if we reflect that the honorific term “knowledgeable” is to be applied only to those who are reliable sources of information, surely an important category for a language-using social species. (1974: 118)
Similar remarks may be found in Sosa’s later paper “Knowledge in Context, Skepticism in Doubt.” He writes, We are social animals. One’s linguistic and conceptual repertoire is heavily influenced by one’s society. The society will tend to adopt concepts useful to it. A concept of epistemic justification that measures the pertinent virtues or faculties of the subject relative to the normal for the community will be useful to the community. The community will hence tend to adopt such a concept. (1988: 152-153)
36
|
What’s the Point of Knowledge?
Although Sosa here refers to epistemic justification, it is reasonable to think the concept of knowledge will be similarly influenced by one’s society. In Problems of the Self Bernard Williams suggests a similar approach to theorizing about knowledge. He denounces traditional analyses of knowledge as “mistaken” and calls the idea that knowledge is justified true belief a “deep prejudice” in philosophy (1973:146). This mistaken view is encouraged by what Williams dubs the “examiner situation,” which is “the situation in which I know that p is true, this other man has asserted that p is true, and I ask the question of whether this other man really knows it, or merely believes it” (1973: 146). Gettier cases are familiar examples of the examiner situation. In cases of this sort, we are represented as checking on someone else’s credentials for something about which we already know. But this is not our standard situation with respect to knowledge. The cen tral focus of epistemic evaluation is the activity of inquiry. Whereas the ex aminer situation is concerned with whether some potential knower really qualifies as such, the actual business of inquiry involves an inquirer who does not know whether p but wants to. As Williams remarks, our standard situation with regard to knowledge (in relation to other per sons) is rather that of trying to find somebody who knows what we don’t know; that is, to find somebody who is a source of reliable information about something. (1973: 146, emphasis mine)
Here we find a clear statement of the important connection between knowledge and identifying people who are reliable sources of information. However, neither Sosa nor Williams develops this idea in detail, so I will concentrate on Craig’s proposal in Knowledge and the State of Nature. There are two key players in Craig’s story: The inquirer: someone who is trying to find out whether or not p is true. The reliable informant: someone who provides the information the inquirer seeks.
It is by focusing on the notion of a reliable informant that Craig hopes to overcome the inadequacies of traditional analyses and to answer some puzzles about knowledge. Who qualifies as a reliable informant? What criteria must she meet? First and foremost, she is someone who tells the truth on the relevant matter. If I am inquiring as to whether p, a reliable informant is someone
THE POINT OF KNOWLEDGE
|
37
who, at a minimum, tells me whether p. This will typically involve the informant holding at least a true belief as to whether p\ viz., either p and the informant believes that p or not-p and she believes that not-p. Reliable informants will not in general tell the truth unless they hold a true belief about it (I’ll return to this point in §2.3). This comes very close to two nec essary conditions on the traditional analysis of knowledge (truth and be lief), but these requirements are couched in terms of a disjunctive condition because we are concerned with the process of inquiry. When inquiring into a certain matter it is an open question as to whether p, whereas the tradi tional theory of knowledge is couched in terms of individual propositions (i.e., S knows that p). For the sake of brevity, we can skip many of the details of Craig’s pro posal in order to focus on the main features of a reHable informant. Craig enumerates the following general properties of what he calls a “good in formant” (1990: 85):
(a) The informant tells the inquirer the truth on the question. (b) The informant is detectable by the inquirer (via some property X) as likely to be right about p. (c) The informant is accessible to the inquirer here and now. (d) The channels of communication between the inquirer and the in formant are open. (e) The informant is as likely to be right about p as the inquirer’s concerns require.
These general properties characterize a “prototypical” good informant. An inquirer who is looking to figure out whether p will typically seek an in formant who meets these conditions. A problem surfaces at this stage. It is not part of our familiar concept of knowledge that a knower has a recognizable property X indicative of true belief, nor must she be willing or able to convey her information. Rather, we can easily imagine cases involving agents who know whether p even though they are unwilling or unable to tell anyone what they know. Further, it seems perfectly conceivable to describe situations in which someone knows whether p even though we cannot detect any property that correlates with having a true belief as to whether p. For example, Steven Shapin’s A Social History of Truth (1995) provides a historical illustration of how many knowledgeable people were not regarded as good informants because they were perceived to lack credibihty. In particular, being a seventeenth-century “gentleman” in England was considered a highly
38
|
What’s the Point of Knowledge?
reliable indicator property for telling the truth, whereas being female or nongentle was a negative indicator property. Of course, the corrupting sort of social pressure on norms surrounding credibility did not actually de prive nongentle men and women of knowledge, but it did disqualify them from being recognizable as reliable informants. In other words, our familiar concept of knowledge does not feature an ything like conditions (b), (c), or (d). This creates a large gap between the concept of a reliable informant and the concept of a knower, which is a prima facie problem for Craig’s proposal. Given the gap that has opened up between the concept we have arrived at by considering the practical sit uation of an inquirer, on the one hand, and our ordinary judgments about knowledge, on the other hand, must we conclude that our central hypoth esis is inadequate?
2.2 Knowledge Objectivized These gaps between a “knower” and a “reliable informant” do not spell doom for our hypothesis. To understand why, we must look at the next part of Craig’s proposal. In his book, Craig attempts to bridge many ap parent gaps between his notion of a reliable informant and the concept of knowledge by appealing to a process he calls objectivization. What is objectivization? It is the backbone of the genealogical narrative that traces the development of the concept of knowledge from the state of nature to our present situation. Jonathan Dancy pithily characterizes this process as follows: “A concept is objectivized if it becomes progressively less tied to the particular concerns of the user” (1992: 394). But how exactly does this process work and why does it come about? Craig provides a “state of nature” explanation for our conceptual and linguistic practices concerning knowledge. A state of nature explanation proceeds by identifying certain basic human needs and then claims that cer tain human practices (e.g., conceptual, linguistic, moral, or institutional) are appropriate responses to these basic needs. To make things simple, we may divide Craig’s state of nature narrative into two broad stages. First, we are asked to imagine an individual inquirer with the problem of having to determine the truth as to whether p. This inquirer will presumably seek an informant who has the general properties of a reliable informant outlined above, since the inquirer is only looking for an informant who is reliable enough for his own purposes. In other words, the inquirer will seek out informants who are accurate, available right now, and willing to answer
THE POINT OF KNOWLEDGE
|
39
the inquirer’s specific question. This roughly characterizes a reliable in formant “at its most subjective” (Craig 1990: 84). But this is only half the story. As Aristotle observes, humans are not isolated inquirers. We tend to form social groups and, as a result, inquirers will almost inevitably belong to a community that involves a number of individuals with a shared problem of having to figure out the truth on di verse topics. In such a community, we tend to be responsive to the needs of others to some degree.2 This takes us to the second part of Craig’s genea logical narrative, where we “objectivize” the notion of a reliable informant in response to certain pragmatic pressures. Now imagine a more complex situation in which a community of inquirers with diverse interests and different cognitive abilities collabo rate to achieve their goals.3 In such a community, other people are often better situated than oneself to acquire true beliefs, so we must hope that others will recommend good informants to us, including when someone recommends himself. For example, Tom up in the tree has a better view of the landscape and is therefore better informed about whether there are predators nearby. Further, we recognize that we do not always need an informant here and now, although we might at some point in the fu ture. It is therefore in our shared interest to store useful information when it is available, since we do not always know when, why, or under what circumstances it might be needed. These considerations put pressure to ward the formation of so-called objectivized concepts. An “objectivized good informant” is someone who is a good informant for others, not just me. We humans come to assess the adequacy of informants for people and purposes beyond our own immediate concerns because our epistemic needs are more socially and temporally extended than desiring informa tion for oneself here and now. In other words, the concept of knowledge derives from our need to identify good informants in general.4 Although each of us needs reliable information to successfully guide our own actions, we also have a shared need to pool and transfer information in order to make it easily accessible. We therefore must identify individuals who are reliable enough for a va riety of individuals with a wide range of interests, projects, and purposes. And yet we have no idea what the needs or interests are of the many people
2 Even if I am not so inclined, I still need to appreciate their point of view if I am to be successful at getting them to help me (Craig 1990: 88). 3 See Craig (1990: 87-88). 4 See Craig (1990: 90-94).
40
|
What’s the Point of Knowledge?
who might require information. Thus, to ensure that informants are reH able enough for many people with diverse interests, a practice develops whereby the epistemic standard is set fairly high. According to Craig, this will
edge us towards the idea of someone who is a good informant as to whether p whatever the particular circumstances of the inquirer, whatever rewards and penalties hang over him and whatever his attitude to them. That means someone with a very high degree of reliability, someone who is very likely to be right—for he must be acceptable even to a very demanding inquirer. (1990: 91)
In short, there will be a general level of stringency for knowledge ascriptions (I will elaborate on this idea in the next two chapters). While someone who is barely reHable or even fairly reliable may be a perfectly acceptable informant for certain purposes, that person is not necessarily a knower. This is because the epistemic standard is higher for knowers than it is for merely acceptable informants. In saying that someone knows, we certify that person as epistemically positioned such that others can freely draw on his or her information. Put differently, the concept of knowledge is used to certify an informant’s befief or information to members of the epistemic community at large.5 This is not to say that flagging reHable informants is a conscious goal of our use of “know.” Most people probably do not reafize that we participate in this activity and achieve this goal by classifying others as knowers. I am simply claiming that by reflecting on the point of epistemic evaluation and the way the word “know” is used, this goal can be plausibly attributed to us. Further, to say the concept of knowledge serves this purpose is not to say this is the content of the concept.
2.3 Closing the Gaps between a Knower and a Reliable Informant By appeaHng to objectivization, we may narrow significant gaps between the concept of knowledge and the concept of a reHable informant. In this section, I will consider three of the largest gaps between these concepts.
5 An “epistemic community” is, roughly, a group of individuals who share information through testimony. I will elaborate on this notion in the next two chapters.
THE POINT OF KNOWLEDGE
|
41
First, what should we say about cases in which a person’s reliability on an issue cannot be identified by anyone! Will such a person qualify as a knower? Second, what should we say about cases in which an informant is unwilling to tell anyone her information? Christoph Kelp (2011) asks us to imagine a priest under the seal of confession who refuses to divulge information about his confessors’ sins. Such a person could certainly have knowledge, but he is not a good informant because the channels of com munication are not open. The same could be said for many cases of pro fessional secrecy; for instance, doctors, lawyers, and CIA operatives might take certain secrets to the grave. Third, what should we say about good informants who do not count as knowers because they do not believe the relevant proposition(s)? For example, Jennifer Lackey (2007a) describes the case of a creationist schoolteacher who learns the theory of evolution in order to teach it to her students. As a result of her highly competent teaching, her students come to know various truths about evolution; how ever, the schoolteacher does not herself know these various things because she does not believe them. These examples all create bridging problems for our central hypothesis. Why must a good informant be detectable, willing to share her informa tion, or believe what she tells her inquirer? Many have rejected Craig’s hy pothesis because they do not think these questions have adequate answers. I will try to answer them. First, the principle of objectivization will “dilute” the detectability re quirement in the following way: the detectability of a knower’s reliability must only be detectable in principle to someone. To say that a good in formant must be publicly detectable is to say only that “her reliability must be objectively ascertainable by some possible means or other, even if not by any means that any actual person happens to have at her disposal” (Neta 2006: 267). You might think we should just drop the detectability requirement alto gether. According to Dancy, “the recognizability requirement is supposed to vanish” (1992: 394); likewise, Kappel says this requirement “is abandoned as the concept [of a good informant] evolves towards our mature concept of knowledge” (2010: 86). However, we should hesitate to eliminate this condition entirely. After all, we cannot reliably designate individuals with properties that are not detectable by us, since identifying them would be akin to detecting these properties (Berry 2011: 24). Further, there would be no practical advantage to a strategy that included people who are in principle undetectable. Objectivizing the concept of the good informant is supposed to transform this concept in response to a community’s practical
42
|
What’s the Point of Knowledge?
needs; but, as Kelp (2011: 59) observes, it is difficult to see how a practical need could arise to extend the applicability of the concept of the reHable informant to people whose reliabiHty is not in principle detectable by an yone. Thus, we should expect the detectability requirement to leave its mark on the concept of knowledge. What about cases in which an informant is unwilling to tell anyone her information? A knower who will not convey her information may still be a reliable informant (in the objectivized sense). The issue is simply that the channels of communication are not open. By objectivizing the notion of a reliable informant, however, we remove the requirement that the channels of communication must be open between the inquirer and the informant. All we need to say is that the informant would be a good source of information if the channels of communication were open. The priest who knows his confessor’s sins certainly has the ability to provide us with that information—he is simply unwiHing to tell us. Still, he is a potential source of reliable information. To use another example, imagine a priest who does not speak our language. This person might be unable to provide us with the relevant information, but he still has the ability to tell someone what he knows. An informant need not actually convey his information or exercise this ability in order to possess it. This allows us to bridge another gap between a good informant and a knower. What about the creationist schoolteacher, who is a highly reliable in formant and yet does not seem to know the relevant propositions because she doesn’t beHeve them? We can dodge the worry raised by this example by including a belief condition as an essential property of the objectivized reHable informant. Why would we require someone to beHeve that p as a condition of knowing that pl A function-first epistemologist has two answers at hand. First, informants will not in general tell the truth un less they themselves hold a true beHef about it. As Craig points out, cases involving people who do not hold true beliefs and yet somehow manage to insincerely give true “information” are rare, and people who do this regularly are probably nonexistent (1990: 12). In most cases, a reliable informant will not only beHeve the information he gives, but also he will give it because he believes it. Inquirers will thus tend to seek informants that either beHeve that p when p, or believe that not-/? when not-p. Second, the fact that someone beHeves p is often a reHable and relatively accessible indication that the believer possesses good evidence. Usually people who do not beHeve p lack good evidence that p; and many people who out right beHeve p do so because they have good evidence. Further, it is often easier to tell whether someone beHeves what she asserts than it is to find
THE POINT OF KNOWLEDGE
|
43
out whether she has good evidence for it. As Steven Reynolds (2002: 156) points out,
we can reliably tell whether someone believes what they are saying by their manner of speaking and by their facial expression. A convincingly told lie is a difficult achievement. Finding out in some other way whether they have evidence that p may require a much more difficult, perhaps practically im possible, investigation of their relevant epistemic history. Thus, people will use this indirect but generally reliable indication of whether the informant’s testimony is based on good evidence. For this reason, we may come to regard belief as a requirement for knowledge. It is by objectivizing the concept of a good informant that we are led closer to our familiar concept of knowledge. With respect to the general properties of the concept of a good informant (see §2.1), objectivization will abstract away from the individual-relative features of these properties. In particular, it will leave (a) unaffected, it will relax conditions (b), (c), and (d), and it will turn (e) into a strong reliability condition of the form (e*) The informant is highly likely to be right about p.
This eliminates the problematic individual-relativity that created a gap between Craig’s initial hypothesis and the concept of knowledge.6 Identifying good informants becomes “socially directed.” We are not just evaluating whether someone is reliable enough for our own individual purposes, but also whether that person is reHable enough for other poten tial inquirers. Knowledge attributions function as a kind of general cer tification for informants who are in a sufficiently high-quality epistemic position to allow inquirers to draw on their information for a wide variety of purposes. What about cases in which a knower does not function as an informant for anybody? If identifying good informants is socially directed, does this leave mysterious why isolated individuals can still have knowledge? Let’s call this the “Robinson Crusoe” problem. Cases of isolated individuals are not too worrying. Even if nobody else is around, it is important that we have a practice whereby people can declare themselves to be reHable sources of information, since these individuals
61 borrow this way of framing the issue from Kelp (2011).
44
|
What’s the Point of Knowledge?
will often be the only ones in a position to tell whether they are reliable or not (Craig 1990: 63). Further, an isolated individual might eventually be a source of information for someone else, or even for oneself at some future time. A knower must therefore meet a sufficiently high-quality epistemic position such that a wide range of potential inquirers could in principle rely on this person’s information, even if nobody actually does ever seek such an informant. These facts explain why we might want to say that someone knows whether p even though, as it happens, that person does not actually function as an informant for anybody. It seems, then, that we can avoid the mismatch between our initial hy pothesis and our ordinary judgments about knowledge by adopting the idea that our familiar concept of knowledge is what remains after objec tivization. To say that someone knows whether p is to certify her as an in formant who satisfies the objectivized standard. Thus, when I henceforth speak of “reHable informants” or “good informants,” I will be using these terms in the objectivized sense. When explaining this view to others, I often encounter the following objection: why would we use the word “know” to flag reliable informants rather than simply call them “reliable informants”? Isn’t this an unnec essary prohferation of words and/or concepts? Mikkel Gerken and Greg Lynch have mentioned this point in conversation, and Patrick Rysiew raises it in print: we have at our disposal other epistemic terms we can and do use in picking out good informants—‘trustworthy,’ ‘reliable,’ ‘always right,’ etc. But given that there are plenty of terms available for picking out informants of one or another degree of reliability, what’s special about ‘knowis)’! (2012: 278).
This is a natural objection. It is true that there are a variety of terms of epistemological appraisal that are designed to flag informants that are reHable to various degrees, or in various ways. However, the concept of knowledge is used to identify a particular type of reliable informant, namely people who satisfy the objectivized standard. I am using “re liable informant” as a theoretically fertile philosophical notion, not a concept that has lay currency. We might count many people as “good” or “reliable” informants (in the nonobjectivized sense) because they satisfy some positive epistemological status, but they do not qualify as knowers unless they are good enough for the community at large. When someone fails to meet this standard, we may fall back on more guarded expressions of evaluation (e.g., “I am pretty sure that,” “He’s usually
THE POINT OF KNOWLEDGE
|
45
right about this sort of thing,” and so forth). It is useful for us to have a range of predicates that express a variety of evaluations of the epistemic standing of others, but knowledge is an especially important epistemic status. This is because knowers must be reliable enough for the commu nity in general. As Craig puts it, “knowledge lies at the end of the road of objectivization” (1990: 91).
2.4 The Road to Skepticism How far does this road take us? One might worry the road of objectiviza tion will lead all the way to skepticism. Objectivization is supposed to result in a concept of knowledge that serves to identify reliable informants to members of our community. To play this role, objectivization will push us toward accepting fairly de manding epistemic standards. But why not expect our informants to be maximally reHable? After all, an infallible informant would certainly be reHable enough for any potential inquirer. The problem, however, is that the more demanding these standards are, the closer we are to skepticism. As Frederick Schmitt says, in recommending good informants we will not always know the purposes to which the information will be put, and we must therefore set the degree of correlation with truth of the property X as high as would be required for any purpose—to a probability of 1. (1992: 558, emphasis mine)
Jonathan Dancy and Richard Feldman raise a similar worry. Dancy writes, “Craig suggests that because we never know how much may hang on getting correct information, we screw the reHability requirement up more or less as far as it will go” (1992: 395). Feldman illustrates the skeptical worry by example: By setting them [the standards] high we will set them too high to please everyone. If you ask me for someone who can tell you about the best route to a certain place, and I restrict myself to informants who are very likely to be right, then I might have to tell you that I can recommend no informant at all. And you might rightly be annoyed if you later learn that I neglected to tell you of someone who was merely likely (but not very likely) to be right about that. Setting the standards high will fail to capture the needs and wishes of some. (1997: 211)
46
|
What’s the Point of Knowledge?
If these interpretations are correct, then objectivization threatens to un dermine our pragmatic account of knowledge. As Feldman points out, it is difficult to see how demanding infallibility from fallible creatures will serve our purposes in everyday fife. By insisting upon the highest level of justification, we would exclude a vast number of informants on whom many people could reasonably rely for their purposes. A thorough discussion of this issue will have to wait until chapter 8. In that chapter, I argue that we are held back from skepticism because de manding infallibility from our informants would undermine the practical requirements that comprise the purpose of our concept of knowledge.7 The point of this concept is to flag informants who can provide us with reliable information, which furthers our goals of pooling and sharing information. But our informants need not be reliable enough for any inquirer whatso ever, including arbitrarily demanding ones. By setting the standard too high we would frustrate our communal epistemic practices, which is anti thetical to the point of epistemic evaluation. Thus, objectivization need not push us into the skeptic’s arms. What about Feldman’s example? He claims that by restricting our selves to informants who are very likely to be right, we cannot recommend informants for ordinary matters, such as who can tell you the best route to a certain destination. I think we can escape Feldman’s worry by appealing to the fact that informants tend to be cooperating members of our species. As such, they often react not just to the inquirer’s question but also to the presumed pur pose of asking it.8 Suppose, for example, that I want directions to Madison Square Garden. When asking for directions, I would typically say some thing like: “Do you know how to get to Madison Square Garden?” or “Do you know the way to Madison Square Garden?” And it is likely that my informant will provide me with directions to Madison Square Garden even if she is unsure about whether or not the route she describes is strictly speaking the best one. My informant will rightly suspect that all I really want are reliable directions. But what if I really am determined to find someone who knows the best route to a certain place (as mentioned in Feldman’s example)? Imagine that my wife is pregnant and expected to give birth within the next few days. Suppose we live quite far from the nearest hospital, that there has been a lot of road construction recently, and that high volumes of traffic
7 Craig (1990: 117) makes a similar point. 8 This would fall under Grice’s (1989) cooperative principle.
THE POINT OF KNOWLEDGE
|
47
are common in our area. Given these conditions, I might insist on knowing the best route to the hospital. I will therefore seek an informant who is very likely to be right about that. But this is precisely what Craig’s ac count predicts: I should not say that I know the best route unless I re gard myself as very likely to be right. Further, we can expect the same result in a variety of situations in which the stakes are higher than usual. Consider the example of a taxi driver in a major city. When traveling by taxi, people generally have a more pressing interest to travel by the best route for a variety of reasons—i.e., to save money, to not feel cheated or duped, and because taxi drivers are expected to know the best routes to local destinations. For these reasons, our expectations for a reliable in formant will be higher than usual. Feldman might grant everything I have said but still maintain there are cases in which an inquirer wants to know the best route even for everyday, trivial matters (not just when the stakes are higher than usual). In these cases, Feldman says we cannot properly recommend an informant and the inquirer might rightly be annoyed if he later learns that we neglected to tell him of someone who was merely likely (but not very likely) to be right. But I do not find this compelling. Even if the inquirer were initially an noyed at me for having not told him about someone who was merely likely (but not very likely) to be right, I have a perfectly good response at hand, namely: “You wanted someone who knew the best route and I wasn’t sure he knew that, but I knew he could tell you how to get there.” This leads us to an important distinction. Informants, unlike mere sources of information, are often cooperating members of our species who are capable of empathizing with an inquirer. As Paul Grice (1989) observes, our fellow humans are capable of reacting not just to our literal questions but also to the presumed purpose of asking them. An informant might therefore give an inquirer useful information that was not explic itly requested. I believe this thought can go a long way toward assuaging Feldman’s concern. I will illustrate this point with one final example. Imagine that my brother and I are driving along a quiet country road, when I suddenly experience a sharp pain in my chest. We decide that I need to get to a hospital—and fast. My (visibly worried) brother notices a passerby and asks her: “Do you know the best route to the hospital?!” Our potential informant can now respond in several ways, depending on how she chooses to interpret my brother’s question. She might provide us with useful directions to the nearest hospital that she knows to be closed. She might also tell us the best route to an open hospital that is much far ther away than the nearest open one. More likely, however, is that our
48
|
What’s the Point of Knowledge?
informant will recognize that my brother and I want to get to the nearest open hospital. Even though my brother’s utterance does not explicitly con tain these scope restrictions, all parties to the exchange assume these con textually appropriate restrictions are in force. What if our informant is unsure about whether she knows the best way to the hospital? In this case, she will likely recognize it is important for us to get to a hospital quickly, so she will tell us how to get there. We can escape Feldman’s worry by appealing to what Craig (1990: 36) calls “the special psychology of teamwork in a community, something that is involved in the use of informants but not in the use of sources of informa tion” (I’ll elaborate on this in the next section). This explains why some body might appropriately recommend herself as a knower even though she is not very likely to be right about which route is strictly speaking the best. She will recommend herself as an informant precisely because she understands the point of my brother’s question.
2.5 Informants versus Sources of Information As just mentioned, there is an important distinction between informants and mere sources of information (see Craig 1990: 35). An encyclopedia, for example, is not an informant but it is a good source of information. The same can be said about thermometers, libraries, films, fossils, tree rings, pamphlets, a cloudy sky, and so on. Even people can be treated as sources of information without being informants. For example, when my wife enters our house wearing a wet raincoat, I may treat her as a source of information about the weather even though I would not regard her as an informant on the issue. This distinction matters because knowledge seems tied to informants rather than sources of information. We all agree that books do not know anything even though they can be great sources of in formation. Likewise, a thermometer does not know what the room temper ature is, but someone can come to know this by looking at a thermometer. It therefore seems appropriate to link knowledge to informants rather than sources of information. What is the purpose of this distinction? Why should we have felt it was worth drawing? It cannot merely have to do with a distinction between agents and nonagents because an agent can be a source of information in both ways, as illustrated above. According to Craig, this distinction is worth marking be cause informants have two distinctive features. First, they are more conven ient than mere sources of information. For example, to extract information
THE POINT OF KNOWLEDGE
|
49
about the age of trees, you must have specialized information about the con nection between tree rings and age. In contrast, the convenience of informants has to do with the fact we do not typically require specialized knowledge to extract information from them. We can learn about various things by simply asking someone who is knowledgeable. The second distinguishing feature of informants is they can participate in the special psychology of teamwork in a community, as I highlighted in the previous section. Feldman maintains that we cannot adequately distinguish informants from sources of information by appealing to convenience. To support this idea, he notes that computerized encyclopedias are extremely convenient sources of information and yet they are not informants (Feldman 1997: 209). But this objection misfires for two reasons. First, neither convenience nor coop eration should be considered sufficient conditions for drawing this distinc tion. Rather, the point is that informants (rather than sources of information) tend to satisfy these conditions. Second, we must remember that the con cept of knowledge entered our cognitive economy long before computerized encyclopedias and other technological advancements. This is important be cause we are seeking to explain why our concept of knowledge has its cur rent shape (or, less metaphorically, what conditions govern its application), and this explanation will inevitably appeal to various practical pressures that humans have tended to encounter. But computerized encyclopedias and other technological advancements only emerged at a very recent point in our his tory, so the concept of knowledge would not be shaped by any such factors. Rather, our ancestors were far more dependent on other people for informa tion precisely because convenient information technology did not yet exist. It is for this reason that we might expect the concept of knowledge to be deeply connected to informants rather than highly convenient electronic sources of information.9 What Feldman needs to show is, first, that sources of informa tion are in general at least as convenient as informants and, second, these sources are ones that have existed for a long enough period to shape our con cept of knowledge. But he does not argue for these claims. Let’s turn to the second distinguishing feature of informants. Unlike sources of information, informants tend to cooperate with others and respond to their presumed interests. This introduces an ethical dimension as well as an epi stemic one.10 If an informant is a cooperating member of our species, then she
9 It is possible that we will eventually regard information technology as “knowledgeable.” Work in extended cognition might reveal that we are already moving in that direction. 101 take it for granted that individuals do tend to cooperate and share information, even with strangers (see Richerson, Boyd, and Henrich 2003). Perhaps we do this because, as Williams
50
|
What’s the Point of Knowledge?
is someone who will empathize with our situation and share a common pur pose. She doesn’t just provide information—she can understand why we want the information we seek (as illustrated by my case about driving to the hos pital). Unlike mere sources of information, informants may be actively helpful. They can “know what the inquirer is up to” (Craig 1990: 36). This is because communication that aims at transferring information takes place against a background of cooperation and mutual expectations. Miranda Fricker elucidates this idea by invoking the notion of trust. As individuals who tend to cooperate with each other to satisfy our basic practical and epistemic needs, inquirers “do not merely rely on their good informants, relating to them as more or less complex epistemic instruments; rather they riuri them” (Fricker 2012:255). What is involved in die notion of trust? To elu cidate this notion, Fricker asks us to consider an example in which Immanuel Kant’s neighbors rely on Kant’s regular habits as a clock for their own behav ior. In this case, Kant is surely functioning as a good source of information for his neighbors. But imagine that one day Kant sleeps late and, as a result, his neighbors’ morning schedule goes awry. Were this to happen, the neighbors might be disappointed or annoyed, but it would be inappropriate for them to feel betrayed or let down in any ethically loaded sense. In this relation there is only reliance, not trust, and so the ethical aspect of cooperation is missing. A crucial difference between informants and mere sources of information is the trust that binds them. Flicker’s proposal is essentially this: when the inquirer asks a potential informant whether p, he signals his trusting epistemic dependence on her; and in responding so as to meet the inquirer’s epistemic need, the informant signals her acceptance of respon sibility to honor (and so not betray) the trust invested in her. Thus a deal of testimonial trust is swiftly, implicitly, and perhaps very lightly struck. In effect, the informant thereby accepts responsibility for her inquirer’s belief as to whether p; and this reveals how profoundly intersubjective the busi ness of testimony is—not merely a matter of an informant’s belief being transmitted to another party, but moreover of responsibility taken for the other person’s epistemic status. (2012: 258)
This is not to suggest it is necessary for informants to take on this respon sibility. Rather, it is due to an implicit cooperative maxim that informants
(2002) argues, humans regard accuracy and sincerity as virtues that are intrinsically good or valuable. Or perhaps we are simply built to collaborate, as some empirical evidence suggests (Tomasello and Carpenter 2007). I leave this question open.
THE POINT OF KNOWLEDGE
|
51
may be regarded as accepting responsibility for an inquirer’s belief when the informant responds so as to appear to meet the inquirer’s need for information.
2.6 Knowledge, Speculation, and the State of Nature So far I have followed Craig by tracing the development of the concept of knowledge from a more primitive notion of a reHable informant, which arose in a hypothetical state of nature. We are asked to imagine a com munity of inquirers who rely on each other for information and who are similar to us in terms of their general needs and interests, yet they do not yet have the concept of knowledge. But doesn’t all this rely on dubious quasi-historical postulations? Why think this just-so story can tell us any thing about the development of our actual concept of knowledge? This is perhaps the most common objection to function-first epistemology.11 Are we just doing “a priori anthropology”? According to many scholars, Craig’s genealogical narrative is the weakest and most problematic aspect of his view. For example, Bernard Williams (2002), Martin Kusch (2009), and Alex Gelfert (2011) maintain that hypothetical genealogies cannot tell us how and why the concept of knowledge actually developed through the course of human evolution. Instead they think we should explore the development of this concept during the course of real human history. Jonathan Weinberg worries that the state-of-nature metaphor is fundamentally history-oriented and thus is “hostage to arguments about the actual past origins” of our epistemic practices (2006: 34). And EHzabeth Fricker (2015) says the very idea of an epistemic state of nature is incoherent. This is because inquirers cannot share a language and use it to intentionally solicit and distribute informa tion without already possessing the concept of knowledge. Intentionally sharing information involves the speech act of assertion, and by asserting p we tend to present ourselves as knowing that p. Thus, it is incoherent to think there could be a community of inquirers who use language to share information without already possessing the concept of knowledge. Fortunately, we may jettison all the historical baggage that Craig has been accused of carrying. A function-first epistemologist need not make any essential reference to prehistory, nor must we trace the genealogical11
11 See Fricker (1998), Williams (2002), Weinberg (2006), Kusch (2011), Gelfert (2011), and Gerken (2017) for versions of this worry.
52
|
What’s the Point of Knowledge?
development of our current concept of knowledge from a more primitive concept. As Craig says, his account need not presuppose that the wholly egocentric, “subjectivised” thought from which it began actually exists or existed, any more than a Hobbsian account of the State needs the corresponding presupposition about the war of every man against every man. The argument is only that if it exists, at any time, or in any individual, it will develop in the direction of objectivization. Therefore there will be objectivized concepts, whether things started that way or not. (1990: 84)
In other words, the principle of objectivization invites us to think of concepts as having started out a certain way without actually presupposing that they began this way at some point in human history. As Craig puts it, the genealogical component of his view is a “nonloadbearing frill” (2007: 193). His strategy is to invite us to see how a concept coming under a certain sort of social pressure would differ from another, related concept that does not come under that pressure. Thus, the central claim is not about human prehistory but about human nature. In his own retrospective assessment of Knowledge and the State ofNature, Craig puts the matter as follows: the circumstances that favor the formation of the concept of knowledge still exist. . . otherwise I would have no support for my thesis that the method reveals the core of the concept as it is to be found now. (2007: 191)
What, then, is the point of the genealogical story? It is to make intelli gible why language-using humans like us, living in social communities, would develop a concept of knowledge in the first place. To explain our contemporary conceptual equipment, however, we only need to make claims about actual facts as they are now.12 As mentioned in §1.4, these very general facts about the human situation include the following: we need true beliefs about our environment to successfully guide our actions; reliable sources of information will lead to true beliefs; asking a good informant will often be the easiest way to acquire a true belief; on any issue some people will be more likely than others to provide a true belief. These firm starting points prevent us from becoming too speculative when constructing our theory. 12 Kappel (2010) makes a similar point.
THE POINT OF KNOWLEDGE
|
53
2.7 Does the Concept of Knowledge Have a Single Purpose? This book’s central hypothesis is that the purpose of the concept of knowl edge is to flag reliable informants (in the objectivized sense). But why think this concept has just one purpose? Perhaps the concept expressed by the word “know” serves a variety of communicative functions.13 As it happens, I do think this concept serves a variety of purposes. I am a pluralist about the functional roles of the concept of knowledge. Here I part company with Craig, who claims the “sole” purpose of the con cept of knowledge is to identify good informants (1990: 18). On my view, certifying good informants is a central (common, important) purpose of this concept, but not the only purpose. In chapter 5,1 will consider a va riety of proposals about the roles of the concept of knowledge. Moreover, I will attempt to give these putative roles a unified treatment. My aim is to show that we think and speak of knowing for many reasons, but identifying reliable informants is the most fundamental of them. What makes it the most fundamental? First, this hypothesis provides the best explanation for the range of data, including our intuitive judgments about cases, linguistic data about the use of “knows,” and the plausible theoretical considerations about human needs and interests that led us to our hypothesis in the first place. Second, the “informant-flagging function” can be used to explain how and why the concept of knowledge serves a variety of other functions, whereas those alternative functions do not better explain why the concept of knowledge is used to identify good informants. In this way, my hypoth esis is more explanatorily fundamental than its competitors. It for these reasons that identifying reliable informants is the primary point of the con cept of knowledge. Thus, I will often speak of this as the role (purpose, function) of the concept of knowledge, even though this concept might be used in ways that go beyond its central point. I will return to this issue in chapter 5.
2.8 Concluding Remarks There is an important connection between evaluative concepts and the purposes they serve. Thus, when analyzing an evaluative concept, it is
13 Kappel (2010), Kelp (2011), Rysiew (2012), Beebe (2012), and Gerken (2017) make this objection.
54
|
What’s the Point of Knowledge?
important to keep in mind the purposes in which the concept has its roots. As David Henderson (2011) reminds us, evaluative concepts arise in con nection with what is conducive to the successful pursuit of some project. The shape of the relevant concept will therefore be conditioned by the needs arising in pursuit of the project to which an evaluative concept is connected. If the purposes of any evaluative concept should constrain one’s un derstanding of that concept, then epistemically evaluative concepts like “knowledge” must also be understood in terms of the practices these concepts serves to regulate. Following Craig, I believe we information gatherers are especially concerned with the project of evaluating sources of information for ourselves and other members of our community. This practice is important, and perhaps necessary, if we are to pool and share reliable information across time and space. Consequently, we seek out informants who are in a sufficiently good epistemic position to provide us with information on which we may reasonably act. But how good an epi stemic position is that? How reliable must an informant be to qualify as a knower? I have claimed that a knower must be reliable enough for many people with diverse interests; however, this is admittedly quite vague. In the next two chapters, I will try to sharpen this idea in order to better elu cidate the shape of my proposal. What follows will increasingly deviate from the view Craig defends in Knowledge and the State of Nature, but I think of myself as developing rather than supplanting his core idea.
THE POINT OF KNOWLEDGE
|
55
chapter 3 | The Value of Fallible Knowledge
The human intellect and senses are, indeed, inherently fallible and delusive, but not inveterately so. Machines are inherently liable to break down, but good machines don’t (often). It is futile to embark on a “theory of knowledge” which denies this liability: such theories constantly end up by admitting the liability after all, and denying the existence of “knowledge.” —J. L. AUSTIN (1946: 98)
human fallibility is rampant. We often make mistakes and we can rarely guarantee the truth of our beliefs. Indeed, almost no belief can ever be ra tionally supported or justified in a conclusive way—a way that removes all possible doubt as to the truth of the belief. And yet we know a lot. It is for this reason that fallibilism about knowledge is widely endorsed. As I’ll understand it, “fallibilism” is the view that one can know that a particular claim is true even though one’s justification for that claim is less than fully conclusive. Put differently, the level of justification requisite for knowing that p does not guarantee the truth of one’s belief that p. This view is attractive because it avoids the alleged skeptical implications of the infallible conception of knowledge. Infallibilists think that for any belief to be knowledge it must be conclu sively justified. Knowledge, on this view, requires the highest possible de gree of justification. But if this were true, we would know almost nothing.1 We are not infallible creatures.1
1 The relationship between infallibilism and skepticism is not straightforward. The widely accepted view is that infallibilist theories of knowledge are “doomed to a skeptical conclusion” (Cohen 1988: 91). But some theorists deny that infallibilism leads to skepticism (e.g., Dretske 1981,
Suppose fallibilism is true. Is fallible knowledge valuable! Many epistemologists have argued that an adequate theory of knowledge must make sense of the supposed value of knowledge. This idea goes back to Plato’s Meno. Assuming knowledge is fallible, we must explain why fal lible knowledge is valuable. In recent work, however, Laurence BonJour appeals to value-related considerations to argue that “this whole ‘fallibi list’ view of knowledge is mistaken.” He writes: There simply is no well-defined, intellectually significant concept of knowl edge fitting this general conception: none that can be genuinely found in common sense or indeed can be constructed or stipulated in a satisfactory way. The supposed weak [fallible] concept of knowledge in question is, I am suggesting, a philosophical myth. (BonJour 2010: 57)
Why does BonJour claim there is no well-defined, intellectually significant concept of fallible knowledge? He gives two reasons. First, he says we cannot specify what level of justification is required for fallible knowledge in a nonarbitrary way. Second, and more significantly, he says that even if we could identify the precise level of justification needed for fallible knowledge, we could not say why any such level would have the signifi cance that we intuitively take knowledge to have. He therefore concludes that such a concept is a myth. These concerns give rise to the threshold problem for fallibilist accounts of knowledge. This is the problem of how to provide a plausible account of what fixes the threshold (level, degree) of justification (evidence, prob ability, warrant, supporting ground) for knowledge in a way that makes sense of its perceived value.2 The threshold problem is a worry for all fallibilist theories of knowl edge. Although fallibilism is widely accepted, it is surprisingly difficult to state the level of justification required for fallible knowledge in a clear and nonarbitrary way. Perhaps for this reason epistemologists have been largely silent about how strong the justificatory component of knowledge must be. Williamson 2000, Davis 2007, Neta 2011). I doubt these view ultimately succeed in warding off skepticism, but I will not argue for that claim here. For a convincing argument that infallibilists can avoid skepticism only at the cost of problematic commitments concerning evidence and evidential support, see Brown (2018).
2 To keep things simple, I’ll speak in terms of the level of justification required for knowledge, rather than in terms of evidence, probability, warrant, or supporting grounds. I do not deny that these terms can fruitfully be given distinct senses, but these distinctions do not matter for my purposes. I use “justification” generically, encompassing both internalist and externalist conceptions.
THE VALUE OF FALLIBLE KNOWLEDGE
|
57
As BonJour writes, “it is fair to say that nothing like a precise specification of this [level of justification] has ever been seriously suggested, let alone more widely endorsed” (2010: 61). Stephen Hetherington says this is “a serious problem about the nature of knowledge, one upon which there is scant epistemological comment” (2006: 41). My goal in this chapter is to solve the threshold problem. I will argue there is a nonconclusive level of justification that can play the role required by the fallibilist conception of knowledge. I’ll start by outlining the two main challenges that BonJour raises against fallibilism, as well as some of the alleged benefits of rejecting fallibilism. I will then reply to these challenges and show that the perceived benefits of rejecting fallibilism are illusory. As will become apparent, my argument is motivated by the idea that the point of the concept of knowledge is to flag reliable informants.
3.1 Challenges for Fallibilism Let’s start by assuming that knowledge is something like (nonaccidental) justified true belief, which is an account that BonJour considers “broadly correct, albeit obviously in need of much additional clarification and spec ification” (2010: 58). With this sketch of knowledge in place, BonJour outlines four traditional assumptions about knowledge (2010: 58-59). First, it is a “supremely valuable and desirable cognitive state” indica tive of “full cognitive success.” A satisfactory concept of knowledge must therefore make sense of this supposed supreme value. Second, knowledge is not a matter of degree: either one has it or one does not, with no room for gradations. Third, epistemic justification is a matter of degree. Fourth, epistemic justification is tied to likelihood or probability of truth, with stronger degrees of justification making the truth of the relevant belief cor respondingly more likely or probable. It is controversial whether all of these assumptions are correct. For ex ample, several philosophers have recently argued that knowledge is not the “epistemic summum bonum” as BonJour calls it. Rather, it has been suggested that understanding, not knowledge, is the most desirable and distinctively valuable epistemic state.3 Also, some philosophers have defended the idea that knowledge comes in degrees (e.g., Hetherington
3 Kvanvig (2003) and Pritchard (2009a) defend the value of understanding. I will discuss the comparative value of understanding and knowledge in chapter 9.
58
|
What’s the Point of Knowledge?
2001). But I will grant all of BonJour’s assumptions for the sake of argument. With these assumptions in place, BonJour distinguishes two conceptions of knowledge: the Cartesian (or “strong”) conception and the fallible (or “weak”) conception. More specifically, The Cartesian Conception Knowledge requires the highest possible degree of justification; viz., justification that is conclusive, which guarantees the truth of the claim that is believed.
The Fallible Conception Knowledge requires a degree of justification that falls short of conclu sive justification but is still fairly strong; viz., justification that makes the belief very likely to be true, but does not guarantee it.
According to many philosophers, the standard for knowledge required by the Cartesian conception is impossible to satisfy in the majority of circumstances (bracketing simple a priori claims and some aspects of im mediate experience). The Cartesian conception therefore entails that we have very little knowledge. This consequence is often thought to be an unacceptable form of skepticism contrary to common sense, which has led many epistemologists to adopt the fallible conception of knowledge. According to Harvey Siegel, “we are all fallibilists now” (1997: 164). But the fallible conception of knowledge is not without problems. This view is committed to the claim that there is some specific level of justifi cation that is less than conclusive but that nonetheless alters our cognitive situation in an important way. Imagine that we are inquiring as to whether p is true. It is easy to understand how increasingly high levels of justifi cation for the claim that p improves our cognitive situation by making it more likely that our belief is true; but the idea that there is some specific threshold (below certainty) at which our cognitive situation transforms from not-knowing to knowing seems quite peculiar. BonJour calls this the “magic” level of justification. Before this level is attained, one merely has a belief that is more and more likely to be true. At the “magic” level one suddenly and miraculously has knowledge. We can put this point differently by thinking of probability as measured by the use of numbers in the interval [0, 1] on the number fine. A prob ability of 0 means the claim is guaranteed to be false and a probability of 1 means the claim is guaranteed to be true. Framed this way, we can
THE VALUE OF FALLIBLE KNOWLEDGE
|
59
ask: how probable must your belief be to qualify as knowledge? There are two obstacles to answering this question. First, any point lower than 1 seems arbitrary. Why pick that point precisely? Second, it is unclear why achieving some specific level less than 1 would make an important difference in our cognitive situation (i.e., by taking us from not-knowing to knowing). BonJour admits that it might be “unreasonable to demand a precise nu merical specification of a level of probability and perhaps not even entirely clear that numerical probabilities are the right way to think of degrees of justification.” But presuming there is a definite concept of knowledge that denotes an “exalted cognitive state,” it is surely not good enough to say merely, as is commonly said, that the level of justification in question is “strong” or “high” or “adequate” or enough to make it “highly likely” that the belief in question is true, for nothing this vague is enough to specify a definite level of justification and a corresponding definite concept of knowledge. (BonJour 2010: 60)
We can summarize BonJour’s two challenges in the following way:
The Conceptual Challenge What specific degree of justification is required to achieve the exalted cognitive state of knowledge? The Value Challenge Why does achieving this specific level of justification make such a dif ference and why is this difference so important?
BonJour considers the value challenge to be more serious because he thinks it is “impossible to see what could give any level of justification that is short of being conclusive the kind of special significance that the weak [fallible] conception requires it to have” (2010: 61). We are therefore led to reject fallibilism by reflecting on the value of knowledge. On the fallibilist picture, it is clear that further increases in jus tification do not cease to matter even once knowledge has been obtained. Presuming that the level of justification is less than conclusive, additional increases in justification will still be cognitively valuable in exactly the same way as earlier increases in justification, before the threshold for knowledge was reached. Why, then, is it so important whether or not
60
|
What’s the Point of Knowledge?
this “magic” level is reached? Why is this level of justification special? BonJour thinks there is no plausible answer to this question, no account of why any nonconclusive level of justification would have such a special status. Furthermore, BonJour claims the fallible conception of knowledge gives rise to two additional problems: the Gettier problem and the lottery paradox. If we abandon the fallible conception, however, these problems allegedly disappear. Edmund Gettier (1963) showed that there are cases in which an agent’s true belief is strongly justified and yet intuitively does not qualify as knowledge. These “Gettier cases” arise because the level of justification required for knowledge is presumed to be less than conclusive, which allows luck to undermine knowledge. For example, Smith might have a strongly justified true belief that Jones owns a Ford, yet Smith’s be lief is true in a way that intuitively is merely accidental in relation to his justification. If knowledge required conclusive justification, however, there would be no room for luck and thus Gettier cases would pose no problem. There is a further problem for the fallibilist, as BonJour notes. Even if we found some new requirement to make knowledge Gettier-proof, the addition of such a condition would not yield any insight into why knowledge should require such a condition. Suppose, for example, that knowledge were nondefeasible justified true belief. On this view, a person S knows p only if there is no true proposition, d, such that if S were to believe d, S would no longer be justified in believing p. What would the rationale for such a condition be, apart from merely avoiding counterexamples? Without any rationale for why knowledge should re quire some elusive anti-Gettier condition, the significance of the con cept of knowledge remains a mystery. We can avoid these problems by denying the assumption that knowledge requires less than conclusive justification. BonJour also claims that fallibilism is an essential ingredient of the socalled lottery paradox.4 The lottery paradox consists of a tension between two thoughts: on the one hand, we are unwilling to assert that we know a specific lottery ticket will lose on the basis of statistical probability alone (e.g., there are a million tickets and only one winner); on the other hand,
4 See Hawthorne (2004) for a discussion of the lottery paradox.
THE VALUE OF FALLIBLE KNOWLEDGE
|
61
we are perfectly willing to assert that we know things which entail that our ticket will lose. For example, when planning my summer vacation I might purchase a local travel guide because I know that I will not have enough money for a Norwegian cruise. If, however, I did have enough money, I would go on such a cruise. Do I know that I will not win the lottery? It seems that I do not know this. So how do I know that I will not have enough money for a Norwegian cruise? The problem generalizes. I am perfectly willing to assert that I know the outcome of a sports game on the basis of reading about the result in a newspaper, and yet the newspaper has a higher probability of containing a misprint than my lottery ticket has of winning. Ruling out knowledge in the lottery case thus seems to impugn a much wider range of ordinary, common-sense knowledge.5 For instance, it seems acceptable for me to claim to know that I will visit my mother tomorrow (call this “p”) even though I do not know that I will not die before then (call this ‘V’). Yet p entails q\ so it seems I cannot know p without knowing q. However, I do not know q\ therefore, I do not know p. Similar examples abound. By ruling out knowledge in the lottery case we end up threatening a wide range of ordinary knowledge. Can fallibilists solve the lottery problem? BonJour says the “deepest and most serious objection” to any proposed additional requirement to rule out lottery cases is the lack of any plausible intuitive rationale for this requirement (2010: 68). We have no clear understanding of why knowl edge should involve such a requirement. If we endorse infallibilism, how ever, then the lottery example constitutes a simple and decisive reductio against fallibilism. If less than conclusive justification were enough for knowledge, then we would have knowledge in the lottery case. But we intuitively do not. BonJour therefore recommends that we abandon the fallibilist concep tion of knowledge. He then tentatively argues that the correct conception of knowledge is some version of the Cartesian view, which holds that one knows that p only if one’s justification guarantees p. This view is said to have several advantages over fallibilism. In addition to avoiding the Gettier and lottery problems, we also get straightforward answers to BonJour’s two challenges. We have a clear answer to his conceptual challenge, which asks us to specify the degree of justification required for knowledge. The answer is we must have conclusive justification. We also get a plausible
5 Jonathan Vogel (1990) highlights this point.
62
|
What’s the Point of Knowledge?
answer to his value challenge, which asks why achieving knowledge-level justification is so intellectually important and valuable. Presupposing that truth is the aim of belief, it makes sense to say that knowledge requires conclusive justification because that would guarantee that the belief’s aim is achieved. Complete cognitive success is achieved only when our jus tification is sufficient to estabfish that our belief is true, which makes it natural to identify the situation of such completeness as the most valuable cognitive state, namely knowledge. Fallibilists like myself thus face multiple challenges. First, we must offer a clearer specification of the threshold for knowledge. Second, we must explain why this threshold is significant. Third, we must show that the perceived benefits of infallibilism vis-a-vis the Gettier and lottery problems are illusory. I will develop a version of fallibilism that answers these objections—a task that BonJour considers “impossible” (2010: 76). More specifically, I will respond to BonJour’s first challenge by arguing that common sense tacitly involves a concept of fallible knowledge that accounts for the “magic” level of justification to a reasonable degree of approximation. I will also suggest that BonJour demands more precision than is achievable (or necessary). Subsequently, I will explain why the “magic” level of justification is valuable and why the perceived benefits of rejecting fallibilism are illusory.
3.2 Vagueness One might try to dismiss BonJour’s critique of fallibilism on the grounds that it merely rehearses the familiar problem of vagueness. To illustrate why, consider the classic example of the predicate “is bald.” Clearly someone with 10,000 hairs on his head is not bald and clearly someone with zero hairs on his head is bald. But at what point does a person go from being “not-bald” to “bald”? Must there be less than 100 hairs on his head? Less than 99 hairs? 98? And so on. It is reasonable to think there is no precise point at which someone goes from being not-bald to bald: the property “bald” is vague.6 And since this form of sorites-style reasoning leads to absurd conclusions (e.g., there are no bald people or, mutatis mutandis, everyone is bald), any such argument is assumed to be problematic.
6 Williamson (1994) is an exception.
THE VALUE OF FALLIBLE KNOWLEDGE
|
63
The analogue with justification in terms of probabilities runs something like this:
(1) A probability of .01 is not sufficient for knowledge-level justification. (2) If a probability of .01 is not sufficient for knowledge-level justifica tion, then a probability of .02 is not sufficient for knowledge-level justification. (3) If a probability of 0.98 is not sufficient for knowledge-level justi fication, then a probability of 0.99 is not sufficient for knowledge level justification.
On the one hand, if we presuppose that a slight increase in probability can never make the difference between not-knowing and knowing, then this reasoning will illustrate that there is no such thing as knowledge (since someone with conclusive justification would still lack knowledge). This conclusion is highly counterintuitive. Even the skeptic thinks conclusive justification is necessary for knowledge; she simply claims that humans rarely, if ever, meet this standard. On the other hand, if we suppose that a slight difference in probability can take one from not-knowing to knowing only when it makes the difference between nonconclusive and conclusive justification (i.e., .99 to 1), then the Cartesian view is correct. Either way, fallibilism is false. This might seem like a problem for fallibilism, but this conclusion is not especially new or worrying (fit merely relies on the familiar problem of vagueness. Assuming this style of reasoning were compelling, the loss of fallibilism would be the least of our worries. Nevertheless, I take BonJour’s challenge to fallibilism seriously. It is surely not good enough to say, as fallibilists often say, that the level of nonconclusive justification required for knowledge is “high” or “strong” or makes one “highly likely” to be right. More clarification is needed if we are to have a cognitively significant concept of knowledge that fits the fallible conception. My own solution to this problem will draw on Craig’s insight about the role of the concept of knowledge. I should point out, however, that Craig is also guilty of being insufficiently precise on this issue. He says a knower is “someone with a very high degree of reliability, someone who is very likely to be right—for he must be acceptable even to a very demanding inquirer” (1990: 91, emphasis mine). While this is true as far as it goes, more needs to be said to make this idea plausible.
64
|
What’s the Point of Knowledge?
3.3 A Reasonable Degree of Precision Although I take BonJour’s challenge seriously, we should question some presuppositions he makes. First, I think BonJour demands more than a reasonable degree of precision from fallibilists. To illustrate why, consider the concept of an “expert.” We might ask, what is the relevant standard to qualify as an “expert” craftsman? How many faultless constructions must she make? How many different objects must she know how to craft? In other words, what is the “magic” level of skill required for expertise? There seems to be no clear answer to this question. There is no obvious and seemingly nonarbitrary point at which one who exercises a skill becomes an expert at it. Does this suggest there are no experts? Of course not. Does it suggest that in order to qualify as an expert one must be maximally skilled such that one could not possibly be more skilled? That also seems false (and perhaps incoherent). Why, then, should we expect knowledge to be any different? The comparison between an expert and a knower can be strengthened. Notice that both “expert” and “knower” are evaluative concepts that serve important roles in our sociolinguistic interactions. More specifically, both concepts are used to flag people who exhibit a particularly high degree of reliability in some domain. An expert craftsman successfully creates the products of her craft in a reliable manner due to her abilities (or at least she could in the appropriate conditions). Similarly, a knower is someone who successfully reaches true belief in a reliable manner due to his cognitive abilities. The comparison between expertise and knowledge is therefore quite apt.7 Indeed, Plato’s term techne is often translated as “knowledge,” but Gentzler (1995) and LaBarge (1997) suggest that it is perhaps better translated as “expertise.” This puts some pressure on BonJour’s antifallibilist argument. His objections to fallibilism seem to work equally well against other familiar concepts, such as the concept of an expert. It is doubtful, however, that we should accept infallibilism about expertise—the view that an expert must be maximally skilled such that one never makes any mistakes and could not improve whatsoever. It is more plausible to think that experts are highly skilled and yet it is logically possible for them to exhibit more skill. So where does BonJour’s argument go wrong? The comparison be tween knowers and experts suggests that he demands too much precision
7 This analogy is not perfect. It seems natural to say that one person can be more of an expert than another, but many people assume that knowledge does not come in degrees.
THE VALUE OF FALLIBLE KNOWLEDGE
|
65
from our evaluative concepts. How much is too much? This is a difficult question to which I can provide no pithy answer, but I will develop an an swer to this question throughout the next two sections. Contrary to what BonJour thinks, I believe there is a way to specify (to a reasonable measure of approximation) the required threshold for knowledge. BonJour doubts this is possible, but only because he mistakenly expects a theory of knowl edge to provide more precision than the subject matter allows. We should follow Aristotle’s advice in the Nicomachean Ethics, where he cautions us against seeking more precision than is achievable for the nature of our inquiry (Book 1.3).
3.4 The Qualitative Model of Justification What, then, would be a “reasonable measure of approximation” for specifying the threshold for knowledge? To answer this question, we must reject another presupposition of BonJour’s argument. He thinks of justifi cation in terms of gradual increase, which presumes a continuous quanti tative model of epistemic justification. This view renders mysterious why the particular point that knowledge marks out is more significant than a point just before or after it. On these grounds, BonJour concludes there is no adequate answer to his challenge, since it is implausible that a gradual increase in justification could yield a qualitatively different state, namely knowledge. But we can reject the quantitative model presupposed by BonJour and instead treat justification qualitatively, with potentially significant differences between adjacent levels of justification. Indeed, if knowledge is qualitatively different from lesser epistemic states, then it makes perfect sense to think that justification might be better understood qualitatively rather than quantitatively. By conceiving of justification in this way, we can expose a crucial flaw in BonJour’s argument. Earl Conee and Richard Feldman endorse the qualitative model of jus tification in their book Evidentialism. They recommend the legal standard for conviction in criminal cases as sufficient for knowledge-level justifi cation, which is proof beyond a reasonable doubt (2004: 296). This level of justification is weaker than, say, a mathematical proof, but it is stronger than a good reason to believe: it calls for “a strong reason to believe.” For a belief to meet this standard, one must have “strong reasons in support of it, no undefeated epistemic reasons to doubt it, and no undefeated epistemic reason to believe that one’s evidence for it is unreliable” (2004: 296).
66
|
What’s the Point of Knowledge?
Two features of this account are important for my purposes. First, Conee and Feldman do not require conclusive justification for knowl edge, and so they defend a version of fallibilism. Second, this account leaves open exactly how strong one’s reasons must be in order to qualify as strong enough. As they say, “There is no conspicuously correct fact of this matter” (2004: 296). In this way, they provide a nice outline of the framework of epistemic justification within which I will operate. They do not conceive of epistemic justification in terms of gradual increase but rather in terms of qualitatively discrete categories: “good reason to be lieve,” “strong reason to believe,” “deductive certainty,” and so forth. This follows the work of Roderick Chisholm, who, in Theory of Knowledge, mentions the following grades of justification: acceptable, reasonable, be yond reasonable doubt, evident, and certain (1977: 9-10). In what follows, I aim to specify the nonconclusive level of justification that is required for knowledge, as well as to explain why this level of jus tification can account for the value we attribute to knowledge. I motivate this view by appealing to our hypothesis that the concept of knowledge is used to identify reliable informants in the objectivized sense.
3.5 The Reliable Informant Standard for Knowledge As argued in the previous chapter, the concept of knowledge derives from our need to identify reliable informants to others. Each of us needs reliable information to successfully guide our actions, but we also have a shared need to pool and transfer information to make it easily accessible. We therefore have a need to identify individuals who are reliable not just for ourselves here and now, but also for a variety of individuals with a wide range of interests, projects, and purposes. However, we often do not know the practical needs or interests of agents to whom an informant might be useful, so we need to ensure that informants are reliable enough for many people with diverse interests. This motivates the following view of knowl edge: the level of justification needed for knowledge is that which puts the agent in a strong enough epistemic position for her to fittingly serve as a reliable source of actionable information for members of her epistemic community. This is admittedly somewhat vague. You might also worry that it merely pushes the question of the level of justification required for knowledge back to the question of the level of reliability required to be a source of actionable information. To more clearly articulate this idea, I will use the
THE VALUE OF FALLIBLE KNOWLEDGE
|
67
framework of relevant alternatives. Craig does not describe his view in this way, so what follows is my own elaboration on his core hypothesis. I pro pose the following level of justification for knowledge: The Reliable Informant Standard for Knowledge To know that p, an agent must be in a strong enough epistemic position with respect to p to eliminate all of the not-p possibilities that are rel evant alternatives to members of the epistemic community that might draw on the agent’s information.
This proposal raises a number of questions. Who are the members of the “epistemic community”? Does it include all possible inquirers, or all living inquirers, or something else? And which alternatives are “relevant” to members of the epistemic community? One thing is clear: including an unrestricted class of people in the epi stemic community would have disastrous results. Skepticism looms if we think that a knower must be in a strong enough epistemic position to sat isfy the demands of any inquirer whatsoever. Sometimes the stakes for an individual are so high that only the strongest of epistemic positions will meet their demands. To avoid skeptical worries, we should think of the epistemic community as roughly comprised of anyone who might actually draw on the relevant information, where “might” tracks the notion of a possibility that could rea sonably be expected to occur. Several scholars have suggested something along these lines, including J. L. Austin (1946), David Henderson (2009), Krista Lawlor (2013), and Stephen Grimm (2015b). This use of “might” does not track mere logical possibility. There are certain alternatives that are not the kind we normally take to be likely counterpossibilities to what the subject knows, and these are the possibilities we need not rule out. For example, suppose I believe the Red Sox won the baseball game last night because someone reliable told me so. It is widely accepted that reliable testimony typically suffices for knowledge, so it is plausible to say I know the Red Sox won. Nevertheless, it is logically possible that someone with life-or-death stakes might need to know whether the Red Sox won last night. A person with such incredibly high stakes would almost certainly regard my secondhand information as insufficiently reliable to act on. But the possibility that such a person “might” appeal to my belief that the Red Sox won is not a possibility that could reasonably be expected to occur. The mere fact that such an inquirer could possibly exist does not in gen eral impugn my ability to acquire knowledge through testimony. If we are
68
|
What’s the Point of Knowledge?
committed fallibilists, we do not expect people to meet the highest pos sible demands to qualify as knowers. Another way to put this idea, following Grimm (2015b), is to say the threshold for knowledge will reach a level high enough to respect the “typical” or “normal” stakes of people in the epistemic community. Thus, the epistemic community is narrower than all possible inquirers. It is comprised of individuals we can reasonably expect to draw on our in formation. A knower must therefore be reliable enough for anyone who may reasonably seek to rely on his or her testimony. This might plausibly include some nonactual inquirers (i.e., reasonable inquirers who happen not to exist but we can easily imagine they might). Moreover, the interests of some actual individuals will not affect the reHable informant standard, since there are arguably some real skeptical epistemologists who think that deceiving demons are relevant to every contingent proposition.8 This helps to clarify which alternatives are relevant to members of the epistemic community: a knower must rule out those that are fitting or rea sonable to the members of the epistemic community.9 Determining the set of relevant alternatives is not an exact science. Our ordinary practices are not tidy enough to provide a simple, general account that will tell us by virtue of what a certain possibility does (or does not) need to be ruled out. But it is clear that we have a rough idea of what normally counts as having done enough to establish the propriety of a knowledge claim.10 People develop a conception of how reliable an informant must be to count as a knower because the practice of epistemic evaluation is one that we all grow up into, and which has continued for some time. Sociafization and accul turation make us proficient at distinguishing the possibifities that must be efiminated in order to have knowledge from those that typically do not. As a result, we can say tolerably well, in particular cases, what does and what does not need to be eliminated in order to appropriately credit someone with knowledge. For example, suppose I claim there is milk in the fridge. It would not be reasonable to criticize or blame me for faifing to rule out the possibility that I am really just a brain in a vat, but I would be open to
8 See Henderson (2011) for a relevant discussion. 9 Lawlor defends a similar “reasonable person standard” for knowledge. An alternative is reasonable “if a reasonable person would want the alternative eliminated before saying she knew p” (2015: 164). I will discuss Lawlor’s view in chapter 5. 10 As Craig (1990: 114) puts it, “the application of the concept [of knowledge] is left to judgement, and ... in human adults of a certain level of intelligence and experience judgment produces a workable degree of uniformity.”
THE VALUE OF FALLIBLE KNOWLEDGE
|
69
reasonable criticism if I failed to take into account the far more everyday possibility that someone finished the milk. There is still vagueness here. We could try to precisify the “reliable in formant” standard by looking at each knowledge claim on a case-by-case basis and attempt to state, for each proposition in each situation, what needs to be done to establish the propriety of a knowledge claim. Mark Kaplan (2008: 353) mentions this strategy, but he rightly does not endorse it. This is not a task that we really want to accomplish, nor is it one that we must accomplish to have an adequate theory of knowledge. Even if we did go through this excruciating process, there would not be unan imous agreement about every possibility that must be ruled out. Some possibilities will be clearly relevant, other will be outrageous, and many will fall somewhere in the middle. Relevance can be a somewhat vague notion. All we must presume is that such judgments will (or would) coin cide sufficiently to give us what Patrick Rysiew calls “a set of ‘core’ not-/? alternatives” (2001: 489). Without this assumption, it would be mysterious why people are adept at determining what a speaker means in uttering, “S knows that /?.” BonJour says fallibilists have not provided (and could not provide) a reasonably clear answer to the question “How much justification is re quired to know?” Does my account provide a reasonable answer to this question? BonJour might not be satisfied; he might claim my account is too imprecise. But I have argued that he demands more precision than is both achievable and necessary. We are not forced to give a precise an swer because our subject matter does not allow it. As Bertrand Russell cautions us, ‘Knowledge’ is not a precise conception. ... A very precise definition, therefore, should not be sought, since any such definition must be more or less misleading. (1912: 134)
Our everyday thinking about possibilities and likelihoods does not occur in precise quantitative terms. Some imprecision is therefore inherent in our concept of knowledge. But despite the imprecision involved, our ordinary practice of epistemic evaluation works well in the course of everyday fife. Moreover, it does not prevent us from saying something useful about when an agent has done enough to qualify as a knower. In short, the reliable informant standard provides us with a promising way to think about the threshold for fallible knowledge.
70
|
What’s the Point of Knowledge?
3.6 The Value of Fallible Knowledge This account not only specifies the threshold for fallible knowledge, it also indicates why this standard is intellectually significant. It is for our collective benefit that we assess the reliability of informants not just for ourselves but for others, too, since this allows us to store reliable infor mation while it is available, without knowing when, why, or under what circumstances it might be needed. We all benefit from pooling and sharing information because we occupy different epistemic positions and our own epistemic position is often not as good as we would like. To achieve the goal of sharing useful information, we must be able to identify informants who meet a sufficiently high standard—that way our various sources of information are certified as reliable enough for individuals with diverse interests. This social picture of knowledge explains how we store, retrieve, and transmit reliable information across time and space, which contributes to human survival and prosperity. This point can be made another way. As inquirers, we seek reliable in formation on which to base our beliefs and actions. One of the basic aims of belief is to store correct information, for the rational person wants his actions to be guided by true beliefs. However, the process of inquiry is potentially open-ended. Consider an inquirer who wants to answer the fol lowing question: Is the bird in my garden a goldfinch? There are a number of possible views about what it takes to know this fact. Must I see that the bird has a red face and know that goldfinches have red faces? Must I dis tinguish it from woodpeckers, which also have red faces? Must I eliminate the possibility that its face is merely painted red? Must I prove that I am not slightly colorblind? Must I know that I am not a brain in a vat? Inquiry can be thought of as a process of ruling out various possibilities. However, it is always possible to continue one’s inquiry—to seek further confirma tion of one’s view. The reliable informant standard is cognitively valuable because it provides a general point at which people can reasonably termi nate inquiry. This “magic” level of justification thereby satisfies an impor tant role in epistemic evaluation: it signals the point of legitimate inquiry closure. The same cannot be said for belief, justification, reliability, etc.1111
11 There may be exceptional cases in which speakers are willing to ascribe knowledge to somebody and yet it is reasonable to check further (see Brown 2008: 1444—1445 and Reed 2010: 228-229). However, it is uncontroversial that “S knows that p” typically conveys that there is no need for further investigation.
THE VALUE OF FALLIBLE KNOWLEDGE
|
71
I will return to this idea in chapter 5, where I discuss the idea that inquiry reasonable ends at knowledge.
3.7 The Cartesian View The Cartesian conception of knowledge is often dismissed too quickly because of its perceived skeptical consequences. It might be true that our concept of knowledge as widely unrecognized skeptical implications, so we should not simply presume that any theory with such consequences is automatically false. Indeed, our pervasive and persistent nonskeptical beliefs are unstable to some extent, since skeptical pressure sometimes leads us to withdraw or weaken our knowledge claims. This makes the Cartesian view more tempting. I will consider the skeptic’s challenge more seriously in chapter 8. At the moment, I will simply say it would be misleading to overemphasize these fleeting moments of doubt that favor skepticism. It is a defeasible criterion of theoretical adequacy that our theory of knowledge does not strip us of everything (or almost everything) we think we know. One of the advantages of fallibilism is that does not need any extra explanation about why common sense could be so radically mistaken. However, a full-scale assessment of the merits of fallibilism over infallibilism will not be offered here. Rather, I will simply demonstrate that on several important issues my proposal is more plausible than infallibilism. Assuming infallibilism is true, why do people so frequently claim to know things? BonJour provides a number of explanations for why what seems obviously true (that we have knowledge) is in fact false.12 One ex planation is that people falsely attribute knowledge due to simple epis temological error: we reasonably regard our justification for a belief as conclusive when deeper philosophical insight reveals it is not. This seems implausible, however, because people often come to terms with their own fallibility. We all recognize that a well-supported belief may yet turn out to be false, but this fact does not lead us to conclude that nobody knows anything. It therefore seems mistaken to impute such an error to us. Moreover, this explanation commits us to systematic irrationality. Shortly after recognizing that we are fallible, we go right back to saying that we know many things.
12 See BonJour (2010: 71-74).
72
|
What’s the Point of Knowledge?
BonJour provides a second explanation for why we typically (but mis takenly) ascribe knowledge in everyday life. He suggests that people may exaggerate when they ascribe knowledge to themselves and to others. Following Butchvarov (1970), BonJour includes the concept of knowl edge in a class of concepts that are often used in a way that involves exag geration. This class of concepts pertains to evaluative standards or ideals that have “serious practical significance,” an example of which is romantic love. According to BonJour, It is obvious that romantic love represents a kind of ideally perfect situation with respect to a certain sort of interpersonal relation and also that the appli cability of this concept has a range of practical implications that are usually viewed as highly desirable. This creates a strong pressure to attribute the state of being in love to oneself and to others even when the strong (though rather elusive) requirements for being in such a state are not clearly met— and even when they are clearly not met. (2010: 72)
But this explanation is implausible. An infallible concept of knowledge would not be more closely tied to evaluative standards that have “se rious practical significance.” In everyday life, people do not require their informants to be in a strong enough epistemic position to rule out radical skeptical scenarios. It makes no practical difference to our inquiries whether or not a radical skeptical scenario actually obtains be cause, as far as we can tell, there is no perceivable difference between these two scenarios. If the concept of knowledge is connected to epi stemic standards that have serious practical significance (as I think it is), we should expect downward pressure on the standards required for knowledge.
3.8 The Gettier Problem What about the alleged payoffs of infallibilism vis-a-vis the Gettier problem and the lottery paradox? Let’s start with the Gettier problem. Many philosophers are persuaded that an agent in a Gettier scenario does not have knowledge, and conse quently much ink has been spilled over how to accommodate these cases. By demanding conclusive justification, infallibilism allegedly secures the correct intuitive verdict in Gettier cases. If the luckless fall guy in a Gettier scenario does not know the relevant proposition, then rejecting fallibilism
THE VALUE OF FALLIBLE KNOWLEDGE
|
73
is one way to make this problem disappear: the Gettier victim does not have knowledge because he lacks adequate justification. But the infallibilist might not escape the Gettier problem entirely. As William Lycan (2006: 152-153) points out, this problem resurfaces for infallibilism in a different shape. Any infallibilist will presumably admit that knowing is at least a regulative ideal, so some cognitive conditions will come closer than others to approximating this ideal. BonJour accepts this view. He says we falsely ascribe knowledge because “a case may seem so close to an ideal standard as to make it seem unreasonably or need lessly fussy to insist on the difference.”13 If knowing is a regulative idea, however, then a person who has overwhelmingly strong evidence should be counted as just-about-knowing or knowing-for-all-practical-purposes, even if people rarely (or never) know anything. However, this holds only so long as the subject is not in a Gettier case. A Gettier victim does not “just-about“ know or “as-good-as” know. Rather, Gettier victims simply do not know. This difference remains to be explained by the infallibilist. Even if we set aside this objection, the infallibilist “solution” to the Gettier problem comes at a high price. To vindicate our intuitions about these problematic cases, the infallibilist imputes to us widespread error in the vast majority of cases in which the word “know” is used in daily life. Infallibilism therefore gets the wrong intuitive verdicts across a much wider range of cases than fallibilism. After all, we happily and appropri ately credit knowledge to people in ordinary fife without demanding they be infallible. Further, there are good reasons to play down the importance of the Gettier problem. In fact, I think there is something distinctively uninter esting about the project of trying to solve the Gettier problem.14 I’ll now briefly gesture as to why. We begin by identifying a deep prejudice in philosophy that is encouraged by Gettier scenarios, a prejudice that I mentioned in §2.1. The prejudice is to concentrate on a particular situation that epistemologists are fond of, which Williams calls the “examiner situation”: [This is] the situation in which I know that p is true, this other man has asserted that p is true, and I ask the question of whether this other man really knows it, or merely believes it. (Williams 1973: 146)
13 See BonJour (2010: 73). Unger (1975) also defends this view. 14 Lycan (2006) calls this “the Gettier problem problem.”
74
|
What’s the Point of Knowledge?
To illustrate, consider Gettier’s own example in which we are told that Smith has strong evidence for his true belief that the man who will get the job has 10 coins in his pocket (call this proposition “p”). The situation is described such that we (the evaluators) know that p is true, that Smith has asserted that p is true, and we are to decide whether Smith really knows that p. In cases of this sort, we are represented as checking on someone else’s credentials for something about which we already know. But this is not our standard situation with respect to knowledge. The central focus of epistemic evaluation is the activity of inquiry. Our inquiries generally aim at truth, so it is apt that our knowledge ascriptions gauge the progress of inquiry. The examiner situation is concerned with whether some potential knower really qualifies as such, but in the actual business of inquiry, one who needs to find a good informant does not yet know whether p. Will solving the Gettier problem do anything to advance or clarify the proper conduct of inquiry? Mark Kaplan (1985) provides a compelling argument for why solving the Gettier problem will do no such thing.15 If you are a responsible inquirer engaged in an inquiry, then you will care fully consider and evaluate the evidence and then conclude whether or not the weight of your evidence supports p. Now imagine that you are deter mining whether someone else knows whether p. From your perspective, the situation looks different from that of the Gettier victim: you are in a po sition to tell whether the agent’s belief is justified and yet false, or justified and true and yet arrived at by a false premise, or Gettiered in some other way. You can distinguish between various states of affairs that the Gettier victim cannot. How important is our ability to make these discriminations? This discriminatory capacity does not seem important or relevant to the proper conduct of inquiry. If I can tell that your belief that p is perfectly justified and arrived at by the best methods of inquiry, and yet unluckily based on a false belief, then I must already know that p. Why, then, does it matter to me whether or not the word “know” is appropriately ascribed to you in this case? What difference does answering this question make to my interest in figuring out whether or not pl It is not as if I am assessing your own performance as an investigator. It has already been granted that your evidence was admirably gathered and that you did what any reason able investigator would have done. Your performance was exemplary and thus the propriety of your conduct as an inquirer is not at issue, as Kaplan reminds us (1985: 357). This makes it difficult to see what possibly hangs
15 However, I do not endorse Kaplan’s conclusion that the concept of knowledge does not provide a useful goal for our inquiries.
THE VALUE OF FALLIBLE KNOWLEDGE
|
75
on solving the Gettier problem from the point of view of inquiry. For the purpose of understanding and advancing rational inquiry, we have little reason to worry about Gettier cases. In short, there is a discontinuity between Gettier situations and the practices or situations that knowledge ascriptions serve to regulate. Suppose we encounter a Gettier case in the course of our daily experience. What practical purposes could be served by determining whether Smith (really) knows that the person who has 10 coins will get the job (or only believes it)? Why should we need or care to know whether Smith knows any such thing? The anomalousness of this question should raise doubts about the significance of our answers to it. Does my account leave mysterious why many people do not regard Gettier victims as knowers? After all, Gettier victims seem to qualify as reliable informants in the objectivized sense. They gather their evidence in admirable ways and they do what any reasonable investigator would have done to arrive at a true belief. So why do many people deny knowledge to Gettier victims? I think we have a good reason to deny knowledge in Gettier cases. The discovery that the correlation between one’s justification and being right about p was accidental or lucky will affect our attitude toward the informant’s reliability as to whether p, and it will affect it quite apart from how it affects our willingness to rely on him for similar issues on future occasions. The luckless fall guy in a Gettier case is not stripped of credit, nor do we cast doubt on his trustworthiness in general; but, as Craig notes, relying on such an informant would produce the retrospective feeling of having run a risk, of having done some thing that one would not have done had one been just a little better informed at the time, rather like finding that the person who has just driven you 50 miles down a busy motorway without incident hasn’t passed the driving
test. (1990: 49)
A Gettier victim can luckily arrive at a justified true belief in a number of ways. When Smith infers that the man who will be hired has 10 coins in his pocket, his reasoning proceeds via a false premise (i.e., that Jones is the man who will be hired). Although he uses a reliable method to ac quire a true belief, that method played no part, or no appropriate part, in his success. He was not led to a true belief by the features of the method that make it generally reliable. But it is in the interest of a truth-seeking inquirer to want true beliefs that are not accidental relative to the method used by one’s informant. It certainly cannot be recommended as a policy
76
|
What’s the Point of Knowledge?
that we should rely on reasoning that proceeds via a false premise. In such circumstances it will be lucky that one’s reasoning culminates in a true conclusion.16 Of course not all lucky believers arrive at the truth via a false premise. Consider the case of Henry and his young son, who are driving along an unknown country road in an area that is (unbeknownst to them) teeming with papier mache bam facades.17 Imagine that Henry is periodically identifying objects in his visual range for the sake of his young son’s edifi cation. As luck would have it, Henry points to the only real bam and says, “That’s a bam.” In this case, Henry does not obviously reason via a false premise, and yet many people believe that he does not know the object is a bam.18 What seems obvious is that Henry is lucky to hit upon the troth. It is this fact that affects our attitude toward his status as a good informant, even if the method he used is generally reliable. The Gettier victim’s reliability is understood counterfactually. These agents seem unreliable because they could have easily been wrong had things been slightly different (e.g., Jones might not have had 10 coins in his pocket; Henry might have easily identified a fake bam). There are a number of ways in which a Gettier victim might luckily arrive at a justified true belief. In any such case, how ever, the informant will seem reliable in one sense (i.e., he uses a reliable method) but unreliable in another sense (i.e., he could have easily been wrong; he was not led to a true belief by the features of the method that make it generally reliable; and so on). Thus, Gettier cases do not pose a significant problem for my account.
3.9 The Lottery Puzzle What about the lottery puzzle? According to BonJour, the lottery case constitutes a simple and decisive reductio against the fallible conception of knowledge. In the case of a lottery, we judge that nobody knows she will lose on the basis of probability, despite the fact that we have incredibly
16 For this reason, those writers (e.g., Lehrer 1965 and Harman 1973) who sought to solve the Gettier problem by adding a “no false lemma” condition to the traditional justified true belief analysis certainly had a substantial point in their favor. 17 See Goldman (1976: 772). 18 The intuition that Henry does not know that he sees a bam has been denied by several philosophers, including Lycan (2006: 158), Millikan (1984), and Turri (2011: 8). Gendler and Hawthorne (2005) argue that the intuition is unstable, and DeRose says this judgment is not “clearly enough correct for it to be usable as the premiss of a good argument” (2009: 49).
THE VALUE OF FALLIBLE KNOWLEDGE
|
77
good justification to believe she will lose (e.g., we are aware the odds of winning are one in a million). Thus, it seems that conclusive justification is necessary for knowledge in the lottery case. This threatens much of what we claim to know because we often take people to know on the basis of nonconclusive justification. But if less than conclusive justification were enough for knowledge, we would presumably have knowledge in lottery cases; however, it is intuitively clear that we do not. Fallibilism therefore issues the wrong verdict in this case. No matter how probable it is that somebody will lose, nothing short of conclusive justification will do. The fallibilist may handle the lottery puzzle in one of two ways. One ap proach is to argue that we mistakenly deny knowledge in lottery cases (and hopefully offer some explanation for why this is so). Baron Reed (2010) believes that denying knowledge in the lottery case will lead to skepticism, so he argues that we do know, for each losing ticket, that it will lose. This approach is troublesome for several reasons. If I know that my ticket will lose, why buy it? And if I know that I will lose, why not sell my ticket for a penny? Furthermore, wouldn’t parity of reasoning allow me to know that every other ticket will lose? Finally, why bother checking the newspaper for the result of the draw? The claim that I know that my ticket will lose has un desirable consequences. It seems we simply do not know that we will lose. A better approach is to maintain that we do lack knowledge in lottery cases and try to show why this does not impugn a wide range of common sense knowledge attributions. This approach is more attractive because it does not commit us to the counterintuitive consequences mentioned in the previous paragraph, nor does it commit us to infallibilism. The challenge is to somehow distinguish between cases in which the object of knowl edge is a proposition like “My lottery ticket will lose” and cases in which the object of knowledge is a proposition like “I will visit my mother to morrow” (which, as mentioned in §3.1, seems to entail that I know that I will not die before then). Is there any principled way to distinguish these types of cases? Here I will avail myself of a solution recommended by Stewart Cohen (1988). I have little to add to Cohen’s proposal, but I will connect his view with some of my earlier remarks about fallibilism. Whether or not you find Cohen’s solution to the lottery puzzle compelling, I regard my version of fallibilism as more plausible than infallibilism for the reasons thus far provided. Cohen argues that we lack knowledge in the lottery case as a result of the statistical nature of our reasons. The statistical nature of the reasons involved in typical lottery cases make the chance of error salient, and the
78
|
What’s the Point of Knowledge?
salience of error sets the epistemic standard to the level of conclusive justification (or, if you are an invariantist, it sets the standard for appro priately asserting that one knows to this very high level). However, this solution still embraces fallibilism: we can know that p even though there is a chance of error. When the chance of error is salient, we do not count the belief as knowledge; when the chance of error is not salient, we may rightly attribute knowledge. The chance of error is not salient in ordinary cases where we appropriately say that we know that somebody has lost a lottery because our reasons consist of testimony, a newspaper, or some thing else that is not explicitly statistical in nature. Does this difference really exist? Ordinarily I may come to know that I have lost a lottery by hearing an announcement on television or reading the results in a newspaper. Surely, however, facts about testimony or news paper reports constitute reasons (evidence) only in conjunction with facts about the reliability of these sources, which is less than perfect. Aren’t the reasons more accurately described as, say, the Daily Star reports that q and that this newspaper is n/m reliable (where n < m)? If the reasons are statistical in this way, then the chance of error should be salient, according to Cohen’s explanation. Consequently, we should be reluctant to attribute knowledge. Cohen’s answer is that “we do not normally think of reasons in this way” (1988: 107). Our attention is not focused on the explicit statis tical probability that the newspaper is wrong. Insofar as we think about the newspaper case, we might imagine a scenario where the informa tion about the winning ticket is transmitted to the newspaper in a typical way: the reporter witnesses the drawing and tells the editor, who then prints it. Of course newspapers are not perfectly reliable, and there is al ways the chance of a misprint. The point is that the chance of error is not salient in the imagined scenario because our reasons are not described in terms of the newspaper’s high, but less than perfect, reliability. If we did consider the various ways in which newspapers make mistakes (e.g., misprints), we -would begin to wonder whether we really knew that our ticket had lost. But we do not normally think of the case in this way (Cohen 1988: 107-108). Why are the standards set in this way? Why are we prevented from knowing in the lottery case? According to BonJour, the “deepest and most serious objection” to any proposed solution to the lottery paradox is the lack of any plausible intuitive rationale for why knowledge would have the proposed requirement (2010: 68). However, it is a general fact about humans that when error possibilities are salient to us, we will deny, or at
THE VALUE OF FALLIBLE KNOWLEDGE
|
79
least feel inclined to deny, that we know. Although we are fallible, we tend to acknowledge our fallibility only when we have some concrete reason in the context to think that we might be mistaken. One way error possibilities can become salient is when my reasons are explicitly probabi listic. This does not prevent us from acknowledging in the abstract that our knowledge ascriptions are fallible and yet sometimes true. As Austin says, “being aware that you may be mistaken doesn’t mean merely being aware that you are a fallible human being: it means that you have some concrete reason to suppose that you may be mistaken in this case” (1946: 98). Thus, the explanatory power of this solution is not limited to lottery cases.
3.10 Concluding Remarks Fallibilists have devoted too little attention to identifying the threshold for knowledge and explaining why this level of epistemic justification is sup posed to matter. This chapter has argued there is a plausible way to specify, to a reasonable degree of approximation, the nonconclusive level of jus tification that is required for knowledge. I have provided a value-based explanation that locates the importance of fallible knowledge in broadly practical, coordinative concerns. I also argued that the perceived benefits of rejecting fallibilism are illusory. If the story I have told is roughly correct, you might think it supports some version of epistemic invariantism. According to the invariantist, the threshold for knowledge is fixed because the truth conditions of knowl edge claims do not vary across contexts. This contrasts with the recently popular idea that the truth conditions of knowledge claims are in some way context sensitive. At least two authors have taken Craig’s hypothesis to support some version of invariantism: Christoph Kelp says this proposal will lead us to a “high-standards” version of classical invariantism, while Martin Kusch thinks the threshold for knowledge will be more ordinary.19 My own view is that the reliable informant standard naturally approximates the level envisioned by classical (nonskeptical) invariantism, but nothing I have said rules out variable standards for knowledge. Who counts as a re liable source of information might change as a result of shifting standards across contexts. This question will occupy us in the next chapter.
19 See Kelp (2011: 65) and Kusch (2011:11).
80
|
What’s the Point of Knowledge?
chapter 4 | Impure Knowledge
Where the assent one way or the other is of no importance to the interest of anyone ... ’tis not strange that the mind should give itself up to the common opinion, or render itself to the first comer. —JOHN LOCKE (1975: 717)
is required for knowledge? In the last chapter, I argued that a knower must be sufficiently reliable to serve as a source of actionable information for the members of his or her epistemic com munity. This highlights an important connection between knowledge and action: if an individual is epistemically positioned with respect to p so as to be a good source of actionable information, then she meets the general standard to know that p. But this communal standard for knowledge is not good enough for every practical reasoning situation. Someone with a particularly important reason for wanting to get the right belief on a question is going to demand more justification than usual. For example, when a mistake would lead to the loss of human life, it seems that only an exceptionally well-justified belief will count as good enough for knowledge. On these grounds, a group of so-called impurists have defended the following thesis: the level of justification required for knowledge is partly fixed by one’s practical reasoning situation. Put differently, the standards for knowledge are sensi tive to what is at stake, or how important it is to someone that a particular belief is true.1 Impurists claim that as the stakes for an individual go up, it how much justification
1 “Impurists” typically focus on what is at stake for an individual, but other practical factors might make for the presence or absence of knowledge (see Anderson 2015). Nevertheless, I will concentrate on the more specific view about stakes—viz., how important it is to S that p be true.
becomes more difficult to know because the individual faces increasingly onerous requirements to act. We thus face a puzzle. On the one hand, I have claimed that a knower must meet the community’s standard for knowledge, which I have called the “reliable informant standard” (see §3.5). On the other hand, impurists say the threshold for knowledge is partly determined by an individual’s prac tical reasoning situation. But how can there be both a general standard for knowledge and a fluctuating standard that is partly fixed by an individual’s interests? This chapter will seek to reconcile these ideas. I will argue that a communal standard for knowledge is required to promote a deep kind of coordination in our basic epistemic practices, yet an individual’s stakes may trump the communal standard when it is not sufficiently high for the relevant practical reasoning situation.
4.1 Impurism and the Threshold Problem According to impurists, whether an agent knows some proposition depends not only on truth-conducive factors but also on stakes. The threshold for knowledge is thus partly fixed by one’s practical reasoning situation. John Hawthorne and Jason Stanley (2008), Jeremy Fantl and Matthew McGrath (2009), and Stephen Grimm (2011) defend this view. As Fantl and McGrath write, We hope it is clear how a pragmatist account can help here [with the threshold problem]. How probable mustp be for p to be known? It must be probable enough to be properly put to work as a basis for belief and action. (2009: 26)
The impurist solution exploits the sufficiency direction of the knowledge norm of practical reasoning, which states that the minimum level of justi fication (or probability) required forknowledge is that which is enough for the subject to properly rely on the proposition in her practical reasoning.2 In “Impurism, Practical Reasoning, and the Threshold Problem,” Jessica Brown considers but rejects the impurist solution to the threshold problem for knowledge. I will start by summarizing her argument in order to highlight where, on her view, standard formulations of impurism go 2 Fantl and McGrath (2009) and Stanley and Hawthorne (2008) have defended the sufficiency direction of the knowledge norm of practical reasoning; Brown (2008) and Gerken (2015) have criticized it. I will not enter that debate here, but I discuss it more in chapter 5.
82
|
What’s the Point of Knowledge?
wrong. My goal in this chapter is to defend a new version of impurism that is importantly different from the type that Brown considers. This new view, which I call communal impurism, is motivated by the “informant flagging” function at the center of this book. I believe this view will solve the threshold problem without succumbing to previous criticisms—but I’ll say more about this later. Brown distinguishes two ways to implement the impurist strategy, nei ther of which she finds plausible. They are The Unity Approach The threshold for a subject to know any proposition whatsoever at a given time is determined by a privileged practical reasoning situation she then faces, most plausibly the highest stakes she is then in.
The Relevance Approach The threshold for a subject to know a proposition at a given time is de termined by the practical reasoning situation she is then in to which that particular proposition is relevant.
Brown attributes the unity approach to Fantl and McGrath (2009) and the relevance approach to Hawthorne and Stanley (2008). I’ll now say a bit more about each view, as well as articulate Brown’s objections to them.
4.2 The Unity Approach According to the unity approach, there is “a single privileged practical reasoning situation facing a subject at a time [that sets] a general standard for that subject to know any proposition whatsoever at that time” (Brown 2014: 181, emphasis mine). This general standard is most plausibly set by the highest-stakes practical reasoning situation faced by the subject (I’ll explain why in the next paragraph). For example, imagine that Keith and Rachel stop at the bank to deposit their paychecks. They have an impending bill that must be paid and little money currently in their ac count, so it is very important that they deposit their paychecks by Saturday. Keith remarks that two weeks ago he was at the bank on Saturday and it was open, but Rachel points out that banks do change their hours (DeRose 1992: 913). If the highest practical reasoning situation that Keith faces at 4:30 p.m. on Friday is the decision whether to wait in line at the bank or return tomorrow, then the stakes in this context set the epistemic standard
IMPURE KNOWLEDGE
|
83
required for Keith to know any proposition at that time. For instance, the same (strong) epistemic position would be required for Keith to know that there are daffodils in the local park (see Brown 2014: 181). Why would the highest-stakes practical reasoning situation for a sub ject set the general threshold for knowledge? The rationale is straightfor ward: if the privileged practical reasoning situation were identified with a lower-stakes scenario, this would undercut the sufficiency direction of the knowledge norm of practical reasoning. It would no longer be true that if a subject knows a proposition, then her epistemic position is good enough to rely on that proposition for practical reasoning. This is a problem for impurism because the sufficiency direction is required for one of the most important arguments for impurism (see Fantl and McGrath 2009: ch. 3). If the threshold for knowledge were not set by the highest-stakes prac tical reasoning situation, then Keith could know that the bank is open on Saturday (because he satisfies the threshold set by a lower-stakes situation) even though he could not rely on that proposition in his practical reasoning about whether to come back on Saturday.3 Thus, on the unity approach the general standard for knowledge must be set by the highest-stakes practical reasoning situation faced by the subject. Brown’s central objection to the unity approach is that it has unwel come skeptical consequences. A subject in a high-stakes situation at time t is at risk of knowing very few propositions at t because the epistemic standard set by the subject’s high-stakes context at t would also set the threshold for her to know any proposition whatsoever at t. As a result, an yone who is in a high-stakes situation will not only find it more difficult to know propositions related to her practical reasoning environment, but also to know any proposition whatsoever at that time. I will call this the problem of skeptical consequences. The impurist might try to confine this skeptical problem to a few points in time; however, Brown argues that this strategy is unavailable because (a) “we are frequently in high-stakes practical reasoning situations” and (b) making high-stakes decisions “often takes a considerable period of time” (2014: 184). In support of (a), Brown points out that everyday fife often confronts us with high-stakes decisions. Her examples include “whether or not to enter or end a marriage, to have children, to take a certain 3 As Brown (2014: 183) points out, setting the general threshold for knowledge lower than the highest-stakes situation would also undermine Fantl and McGrath’s (2009: 22) solution to the problem of concessive knowledge attributions.
84
|
What’s the Point of Knowledge?
job, to move house, to invest one’s life savings in certain stocks, and so on” (2014: 184), all of which are plausible scenarios where the threshold for relying on a proposition in one’s practical reasoning is quite high. In support of (b), Brown uses the example of deciding whether or not to end a marriage, which may take months or even years. Brown acknowledges that such a person would not be consciously considering this decision at every second of every day over several months or years, but she claims that “a lack of conscious reflection does not show that one is not in a decision situation” (2014: 184). Brown doesn’t elaborate much on this point, but I will grant it for discussion. Thus, the unity approach implies that much of what we take ourselves to know is not knowledge at all. Brown rightly takes this to be a problem atic skeptical result. She therefore recommends that impurists reject the unity approach.
4.3 The Relevance Approach To avoid the skeptical implications of the unity approach, one might turn to the relevance approach. According to this view, “the strength of ep istemic position required for a subject to know a proposition at a time depends on what practical reasoning situations she then faces to which that proposition is relevant” (Brown 2014: 186). For example, the fact that Keith is in a high-stakes situation with respect to the proposition that the bank is open on Saturday means that he must be in a very strong epistemic position with respect to this proposition in order to know it. However, the fact that Keith is in this high-stakes situation does not affect the strength of epistemic position required for him to know other unre lated propositions. Although Keith must be in a very strong epistemic po sition to know that the bank is open on Saturday, this does not entail that he must be in an equally strong epistemic position to know that there are daffodils in the local park. The relevance approach faces two problems. I will call the first the problem of easy knowledge, which Brown summarizes as follows: The [problem] arises from propositions which are relevant only to actions with very low stakes. We may worry that the relevance approach would allow that, in such situations, one can know a proposition on the basis of an extremely weak epistemic position. (2014: 187)
IMPURE KNOWLEDGE
|
85
With Brown, I believe it is problematic to allow that a subject can know a proposition on the basis of a very weak epistemic position simply because the specific stakes are so low.4 Another worry for the relevance approach is what Brown calls “the problem of practically irrelevant propositions”: The second problem arises from the fact that, for a subject S at a time t, there may be some propositions which are irrelevant to any of S’s actual practical decisions at t. (Brown 2014: 188)
To illustrate this point, Brown asks us to consider the proposition that Mars has two moons. It is true that for me right now this proposition is irrelevant to any practical decisions I face. Granted, it might be relevant at some time to someone else (or to me at some future time); but, as Brown says, the fact that the fact that the proposition concerning Mars is practically rele vant for, say, a NASA scientist next year “is of no help in determining the standards for me now to know that proposition since, for impurists, it is the practical reasoning situation faced by a subject at a time which determines whether she then knows” (2014: 188). The relevance approach thus fails to fix the level of justification re quired to know those propositions that are irrelevant to any practical reasoning situation a subject then faces. This is a problem because we have a vast range of beliefs that are, at any given time, irrelevant to any practical reasoning situation we face. As a result, the relevance approach not merely fails to give guidance for judgment about such cases, it also leaves knowledge indeterminate (Brown 2014: 189). This is a significant theoretical cost. So far my discussion has focused mainly on Brown’s argument because it nicely sets up the threshold problem and illustrates why, on the basis of her objections, two current forms of impurism do not satisfactorily an swer the threshold problem. In the next section, I will propose a third way to implement the impurist strategy, which I call communal impurism. What distinguishes this version of impurism from both the unity and rel evance approaches is that, on my view, the threshold for knowledge is not necessarily set by the subject’s practical reasoning situation. Rather, the practical reasoning situation faced by other inquirers (even potential inquirers) largely determines the level of justification required to know.
4 See Russell and Doris (2008) for a similar objection.
86
|
What’s the Point of Knowledge?
As I will argue, this view does not succumb to the criticisms Brown raises against the unity and relevance approaches. This is largely because the impurists that Brown has in mind are swbjecr-centered impurists, whereas my view is not one that she considers. If we adopt communal impurism, however, we can set the level of justification required for knowledge in a nonarbitrary way, thereby solving the threshold problem.
4.4 Communal Impurism As inquirers we want information on which to base our beliefs and actions. Drawing on the work of Craig, I have argued that this basic practical con cern gives rise to a practice whereby we certify others as “knowers” if they meet a general standard of reliability. More specifically, an agent knows that p if she is epistemically positioned with respect to p so as to be a good source of actionable information, which means that one may take it from her that p. By reflecting on the epistemic needs of a broad class of people, we find a principled basis for our common judgments about whether a subject knows. But we also recognize that this general standard is not high enough for every practical reasoning situation. Compare the following pair of bank cases that were originally developed by Keith DeRose (1992):
Low Stakes Keith and his wife Rachel are driving home on a Friday afternoon. They plan to stop at the bank on the way home to deposit their pay checks. It is not important that they do so, as they have no impending bills. But as they drive past the bank, they notice that the lines inside are very long, as they often are on Friday afternoons. Realizing that it isn’t very important that their paychecks are deposited right away, Keith says, “I know the bank will be open tomorrow, since I was there just two weeks ago on Saturday morning. So we can deposit our paychecks tomorrow morning.”
High Stakes Keith and his wife Rachel are driving home on a Friday afternoon. They plan to stop at the bank on the way home to deposit their paychecks. Since they have an impending bill coming due, and very little in their account, it is very important that they deposit their paychecks by Saturday. Keith notes that he was at the bank two weeks before on a
IMPURE KNOWLEDGE
|
87
Saturday morning, and it was open. But as Rachel points out, banks do change their hours. Keith says, “I guess you’re right. I don’t know that the bank will be open tomorrow.” In the low-stakes context, Keith’s claim to know that the bank will be open seems true; meanwhile, it also seems true that in the high-stakes context Keith doesn’t know that the bank will be open. On my account, this is because in the former context Keith is reliable enough for most people to base their beliefs and to act on his information, but in the latter context his epistemic position is not sufficiently strong to tell against some error pos sibility that reasonably seems significant in his practical reasoning situa tion. This illustrates that the individual’s stakes may trump the communal standard when it is not sufficiently high for the relevant practical reasoning situation. Any particular epistemic standard will be too low for some people and too high for others. For this reason, one might worry that my view posits a puzzling asymmetry with respect to whose stakes affect the relevant epi stemic standard. On the one hand, I claim that the threshold for knowledge cannot drop below the communal standard because it would cease to serve the function of identifying reliable informants for the community at large. On the other hand, I claim that individual stakes can trump the communal standard when the stakes for an individual are higher than normal (as illus trated in the bank case above). But you might wonder why an individual’s interests would ordinarily get swamped by those of the community except when the individual’s stakes are exceptionally high. We can explain this seemingly puzzling asymmetry. On the one hand, the necessity of pooling and sharing actionable information creates a need for a communal threshold for knowledge. This is because sharing useful information requires that we identify informants who are reliable enough for a variety of individuals with diverse interests. To ensure that informants are sufficiently reliable for this purpose, a practice develops of setting the standard fairly high. This communal threshold for knowledge must be high enough to ensure that anyone who meets it will be suffi ciently reliable for most practical reasoning situations. (And this standard cannot be too high because that would make knowledge less than widely available, which would prevent knowledge from playing its role in prac tical reasoning.) Moreover, the standard cannot drop below this threshold because to do so would be inconsistent with the function of the concept of knowledge, namely, to identify people who are epistemically positioned to render information that is fitting to the community’s interests. Thus,
88
|
What’s the Point of Knowledge?
whatever practical reasoning situations a subject faces, one must meet a general minimum level of strength of epistemic position to know a propo sition. This is the communal threshold. On the other hand, the communal threshold is not high enough for all practical purposes. Sometimes people demand informants who satisfy very high standards, as shown by the high-stakes bank case. Although Keith meets the communal threshold for knowledge in both the low- and highstakes cases, his evidence is not strong enough for his practical reasoning situation in the high-stakes scenario. And since an agent knows that p only if his epistemic position is strong enough to tell against some error possi bility that reasonably seems significant in his practical reasoning situation, Keith fails to know. When we recognize that serious consequences (and not necessarily our own) turn on having true information, the communal threshold will be inadequate. In this way, my view remains a version of impurism. Keith can know that the bank is open in one context and yet that same proposition fails to be known by Keith in another context, even though the strength of his epistemic position is the same in both cases. Similarly, communal impurism permits a proposition to be known by one person, while that same proposition can fail to be known by another person with the same strength of epistemic position at the same time. Allow me to summarize this view. Communal impurists claim that the threshold for knowledge is sufficiently high to allow us to pool and share reliable information with a wide range of people that have diverse interests and projects. However, some practical reasoning situations require us to crank the epistemic standard up at least another notch. Thus, meeting the communal threshold for knowledge is necessary but not always sufficient to know. This account offers a plausible diagnosis of the commonly held intuitions about low-stakes and high-stakes cases. Someone who meets the communal threshold for knowledge will usually qualify as a knower, but she will not qualify as a knower if she isn’t reliable enough to meet the more demanding expectations in a high-stakes practical reasoning sit uation. Unsurprisingly, we should not recommend an informant to an in quirer if we are aware that the inquirer’s purposes are particularly pressing and so the informant will fall short of the inquirer’s heightened demands. Thus, we do not attribute knowledge in high-stakes scenarios unless a very high epistemic standard is met. An insensitive invariantist will reject this view. According to insensitive invariantism, what counts as being in a sufficiently good epistemic posi tion to know some proposition does not vary—is not sensitive to—any individual’s stakes or practical interests at the time in question (whether
IMPURE KNOWLEDGE
|
89
it be those of the subject, the attributor, or the evaluator of a knowledge claim). Thus, the insensitive invariantist might argue that the threshold for knowledge firmly settles at a level high enough to meet the community’s standard, while the alleged context-sensitivity of our knowledge ascriptions ought to be dealt with at the level of pragmatics. I will return to this idea in chapter 7. My aim in this chapter is not to settle the issue of whether we should endorse insensitive invariantism or some rival view. Rather, my aim is to outline a new version of impurism that (a) is motivated by plausible assumptions about the purpose of the concept of knowledge, (b) does not succumb to Brown’s objections against standard formulations of impurism, and (c) offers a plausible solution to the threshold problem. Having outlined communal impurism, let’s now re-examine Brown’s objections.
4.5 Revisiting Brown’s Objections The problem of skeptical consequences is no problem for communal impurism. We invite unwelcome skeptical consequences if a subject’s high-stakes situation at t sets the threshold for her to know any proposi tion whatsoever at t; but the communal impurist is not committed to the idea that a single practical reasoning situation facing a subject at a time sets a general standard for that subject to know any proposition whatso ever at that time. Rather, the threshold for a subject to know a proposi tion must be at least as high as practical reasoning typically requires, and it may be temporarily heightened if an inquirer’s practical reasoning sit uation calls for it. Whether or not the threshold for knowledge is higher than the communal standard will be determined by the practical rea soning situation to which that particular proposition is relevant. Thus, it does not follow that anyone who is in a high-stakes practical reasoning situation will find it difficult to know any proposition whatsoever at that time. (It is theoretically possible for an entire community to be in a high-stakes reasoning situation with respect to some proposition, but that would only make it difficult for the community to know the truth of that proposition.) What about the problem of easy knowledge and the problem of prac tically irrelevant propositions! The first problem arises if a subject can know a proposition on the basis of very weak justification simply because that proposition is relevant to only low-stakes situations the subject faces.
90
|
What’s the Point of Knowledge?
The second problem arises if the level of justification required to know a proposition remains unfixed—thereby leaving knowledge indeterminate— because the proposition is irrelevant to any practical reasoning situation the subject faces. Does communal impurism make knowledge too easy or leave knowledge indeterminate? For communal impurists, the threshold for knowing a proposition does not drop to an implausibly low level (or remain indeterminate) simply be cause the proposition is relevant only to low-stakes situations (or no prac tical reasoning situation at all) faced by the subject. Whatever practical reasoning situations a subject faces, there is still a general minimum level of strength of epistemic position required for a subject to know a proposi tion. This is the communal threshold for knowledge. While I think this account is right as far as it goes, it leaves two issues unresolved. First, the problem of easy knowledge will resurface at the level of the community. According to communal impurism, the epistemic standard for knowledge is set not just by the subject’s practical reasoning situation but also—and perhaps principally—by the practical interests of members of the subject’s epistemic community. In other words, the practical reasoning situations faced by other inquirers largely determines the level of justifi cation required to know. However, we can imagine a community in which not much is at stake for anyone in the truth of certain propositions. For example, imagine a community whose members are not very interested in baseball results. Communal impurism appears to entail that in this com munity, one can know that the Red Sox won on the basis of very weak justification. This is a problem because knowledge requires good evidence even with respect to issues in which an entire community might have little stake (i.e., relative to which the practical costs of being mistaken are basi cally nil). Let’s call this the problem of communally low stakes. Second, the problem of practically irrelevant propositions also resurfaces at the level of the community. According to Brown, impurism fails to fix the threshold for knowing propositions that are irrelevant to any practical reasoning situation faced by a subject. Communal impurists like Grimm (2015b), Henderson (2009), and myself (2013) have tried to solve this problem by appealing to the relevance of certain propositions to some possible actions that are not currently under consideration. In particular, we have argued that the epistemic standards for knowledge are sensitive to the concerns not just of the subject (or attributor) of knowledge but also to members of the subject’s (or attributor’s) epistemic community, including merely potential inquirers (who are like us but are interested in
IMPURE KNOWLEDGE
|
91
the proposition in question). However, some propositions are so trivial that nobody in any community will likely care about their truth; for example, it is doubtful that anyone does or would care about the number of grains of sand currently on Waikiki Beach. This information will not likely be rele vant to any action in which someone has stake. (This is not to suggest that seeking the truth is always motivated by practical concerns; we may want to know something simply because we are curious. Nevertheless, it is un likely that anyone would, or should, be interested in the number of grains of sand merely for its own sake.) Thus, communal impurism seems to provide no guidance in cases where a proposition is of no practical interest to anyone. Consequently, it will be indeterminate what level of justification is required to know these propositions, so it will be indeterminate whether anyone knows such things. This is a problem because even with respect to questions that nobody would reasonably care about, good evidence is still required for knowl edge. Let’s call this the problem of communally irrelevant propositions. When confronted with these two new objections, the communal impurist might simply bite the bullet. She might be willing to do this because her view is already on much better footing than the (subject-sensitive) versions of impurism that Brown considers. If communal impurists are right that a knower’s epistemic position must be strong enough to meet the community’s general minimum threshold, then knowledge would not drop to an implausibly low level simply because some subject does not have much stake in the issue. For instance, although I might not care who won the 1990 World Series, it is plausible that someone cares about it. Communal impurists thus greatly shrink the amount of easy knowledge in the world, which gives their view an advantage over other versions of impurism. Communal impurism also shrinks the amount of indeterminacy about knowledge. This is because the threshold for knowing that p is not typ ically set by the relevance of p to some subject’s practical reasoning situation, but rather by the relevance of p to a wide range of possible actions that are not currently under consideration (i.e., to the actions of other people or possibly to oneself at some future time). Thus, communal impurism can provide a fixed threshold for knowing many propositions, including those irrelevant to any practical reasoning situation faced by a given subject. The only propositions that do not have a fixed threshold for knowledge would be ones that we cannot imagine anyone reasonably appealing to. In this way, the communal impurist mitigates another one of Brown’s concerns.
92
|
What’s the Point of Knowledge?
But instead of just blunting Brown’s objections, we should try to ad dress them more completely. This requires us to deal with the problem of communally low stakes and the problem of communally irrelevant propositions. In what follows, I will sketch an answer to these problems. Getting clear on these issues will not only further distinguish my view from other communal impurists, such as Grimm (2015b) and Henderson (2009), but will also provide an opportunity to fully address the threshold problem and move the discussion forward.
4.6 Refinements Let’s start with the problem of communally low stakes. Are communal impurists committed to the view that people in a community can know p (i.e., the Red Sox won) on the basis of very weak justification simply be cause no one in that community has much stake in whether p? I argue they are not. Even if I five in a community that does not care much about base ball results, it is plausible that someone cares about them. What matters for the communal impurist is whether it is reasonable to expect someone to care about whether p, including people outside of one’s immediate community. A lot turns on what we mean by a “community,” so I’ll try to be more precise. On my account, an agent knows that p only if he or she is in a strong enough epistemic position with respect to p to eliminate all of the not-p possibilities that are relevant alternatives to “members of the epistemic community that might draw on the agent’s information.” As mentioned in §3.5, this proposal raises difficult questions about who counts as a member of one’s epistemic community. Taking my cue from Grimm (2015b), I claimed we should think of the epistemic community as roughly comprised of anyone who might actually draw on one’s infor mation, where “might” tracks the notion of a possibility that could rea sonably be expected to occur. On this view, an “epistemic community” is much broader than, say, a group of people living in roughly the same area (i.e., a tribe, village, town, etc.). It is also broader than a group of people who share a language, a religion, and so forth. Although there are varieties of social, geographical, linguistic, and religious communities (among others), I think there is a sense in which we are all part of the same epi stemic community. Let me elaborate. As inquirers, we all belong to a community of inquiry—a collection of people motivated by the shared need for reliable information to guide
IMPURE KNOWLEDGE
|
93
our actions and satisfy our natural curiosity. Our reflective understanding of knowledge is that it is based on needs so central to human life that it can be abstractly modeled in a way that, while it might be culturally elaborated in particular ways, is culturally invariant. (I will defend this idea at length in chapter 6.) The human interests that ground our concept of knowledge are so basic and so important that this conclusion is ines capable. This is why we have epistemic standards that must be met not only by our linguistic, social, or religious communities, but also by our epistemic community, which is far more expansive. We all have a need for reliable information, yet human interests are plastic and unpredictable, so we require informants to meet a sufficiently high standard before we certify their information to others. Even if no one in my town cares about baseball results, there are certainly people in my “epistemic community” who might reasonably care. Thus, our standards for how reliable an in formant must be to count as a knower will be sensitive to a diverse range of interests and concerns, including those of people outside my linguistic, social, and religious communities. This explains why communal impurists are not committed to the view that it is easier for, say, the members of a certain village to know that p even when no one in that community has much stake in whether p. However, appealing to the existence of an epistemic community does not fully resolve the worry. Even if we expand our notion of a community to include the vast majority of people, it is still plausible that certain truths are so trivial that nobody would reasonably care about them. Thus, we still encounter the problem of communally irrelevant propositions. Grimm (2015b) attempts to solve this problem. He argues that because humans have diverse practical interests, there might be people with an interest in whatever random topic is at issue. If this claim were correct, it would allow us to solve both the problem of communally low stakes and the problem of communally irrelevant propositions. Unfortunately, this idea is implausible. Consider the following passage from Grimm: I think the right answer is that, given how plastic and unpredictable our practical concerns can be, there is always a story one might tell about why a topic might be of interest to someone, no matter how trivial or insignificant it might seem on the whole. Thus even if getting to the truth about the 323rd number of the Wichita, Kansas phone directory might be (and presumably is) something that you and I could not care less about, it is easy enough to imagine someone who might care about this, if only because he or she wants to phone the person up. More generally, the very idea that there are
94
|
What’s the Point of Knowledge?
certain topics that are necessarily trivial or unimportant, from a practical point of view, as opposed to just contingently trivial or unimportant, seems like a mistake. (Grimm 2015b: 127-128)
What Grimm says in this passage is only partially right. Our practical concerns are indeed plastic and unpredictable, so a vast range of seem ingly trivial truths might nevertheless be important to someone. Assuming our judgments about knowledge are sensitive to the practical concerns of others, this will push the threshold for knowledge high enough to respect these concerns. But even if we are sensitive to the plasticity and unpre dictability of people’s practical concerns, there is not always a (reason able) story to tell about why a trivial fact might be of interest to someone. Take my earlier example: it is doubtful that anyone does or would care about the number of grains of sand currently on Waikiki Beach. Contrary to Grimm’s point, it is not easy to imagine someone who might (in the relevant sense) care about this fact. Rather, it seems like a fact in which nobody will likely have any stake. Thus, it is implausible that for any random topic at issue, there might reasonably be someone with an interest in that topic. There are certain topics whose truth nobody could reasonably care about. So even though our judgments about knowledge are sensitive to the practical concerns of others, the stakes of some third party will not push the threshold for knowl edge to a sufficiently high level in the case of many trivial propositions. For this reason, appearing to the plasticity and unpredictability of our prac tical concerns will not fully answer these objections. I want to propose a different solution, one that will allow us to address the problem of communally low stakes and the problem of communally irrelevant propositions in the same way. This solution appeals to the intui tive idea that “like cases should be treated alike.” Although this principle is more commonly invoked in ethical and legal issues, it also bears on episte mological issues. To clarify this idea, I’ll start with an example. I have absolutely no interest in the number of active artificial satel lites currently orbiting Mars. Nevertheless, it is reasonable to suppose that someone does (or will) have an interest in this question, such as a NASA scientist. Thus, to know the number of active artificial satellites currently orbiting Mars, I must meet a sufficiently high standard. Now let’s also suppose, perhaps somewhat unrealistically, that we cannot imagine an yone reasonably caring about the number of inactive artificial satellites orbiting Mars (assume NASA only has use for active satellites). If we cannot reasonably imagine anyone caring about these inactive satellites,
IMPURE KNOWLEDGE
|
95
then it seems to follow (on the impurist view) that it is easier to know the number of inactive satellites orbiting Mars than to know the number of ac tive satellites orbiting Mars. This is an absurd result. Whatever level of jus tification is required to know the number of active satellites orbiting Mars, a similar level of justification is presumably required to know the number of inactive satellites orbiting Mars. How can we explain this? A reasonable explanation is that we treat like cases alike. Thus, if a cer tain level of perceptual justification and technological sophistication is re quired to know the number of active satellites orbiting Mars, then a similar level of perceptual justification and technological sophistication will, ce teris paribus, be required to know the number of inactive satellites orbiting Mars. Put more generally, if a certain level of justification (whether per ceptual, inferential, mathematical, etc., or some combination) is required to know that p, then a similar level of justification will be required to know similar truths. This is analogous to a standard way of arguing in ethics, namely, if we should do X in circumstances Cp and circumstances C2 are relevantly similar to Cp then we should do X in C2. Further, this prin ciple seems to hold for the various ways of knowing, such as testimony, memory, and reason (inference). Consider the following case of testimonial knowledge. Suppose I want to know whether the phrase “There’s daggers in men’s smiles” is from Hamlet or Macbeth. Let’s also stipulate that I have no internet access, so I am unable to look up the answer at home. Fortunately, my neighbor, Jess, is a renowned expert on the works of William Shakespeare. I decide to ask Jess and she tells me with unwavering confidence that the phrase appears in act 2, scene 3 of Macbeth. Jess is right, and I unhesitatingly form the corresponding true belief. It should provoke no controversy to say this is a case of testimonial knowledge. As an expert on the subject, Jess clearly meets the level of justification required to know the answer to my question. Further, I can acquire knowledge by trusting her testimony. Now let’s modify the example. It is easy to imagine why someone might (in the relevant sense) care about the correct reference for a famous phrase like “There’s daggers in men’s smiles,” but it’s difficult to imagine why someone might care about the correct reference for an inconsequential phrase like “I thank you, doctor” (Macbeth, act 4, scene 3). This particular phrase seems so unimportant that it is unlikely anyone will care much, if at all, about it. But if we cannot reasonably imagine someone caring about the source of this phrase, then we cannot fix the threshold for knowing this truth by appealing to its importance to some possible actions. This leads
96
|
What’s the Point of Knowledge?
us back into the problem of communally low stakes and the problem of communally irrelevant propositions. If, on the one hand, we cannot im agine someone having much stake in this issue, then communal impurism seems to imply that it is more difficult to know that “There’s daggers in men’s smiles” is from Macbeth than to know that “I thank you, doctor” is from Macbeth. On the other hand, if we cannot imagine anyone having any stake in this issue, then communal impurism will fail to fix the level of justification required to know this truth. As a result, it will leave knowl edge of this truth indeterminate. Both consequences are theoretical costs. To resolve this issue, we must explain why good evidence is required not only to know truths that might reasonably matter to someone, but also why good evidence is required to know truths that nobody has much, if any, stake in. As mentioned above, an intuitively plausible explanation is that we treat like cases alike.5 Thus, whatever level of justification is re quired to know that one quote is from Macbeth, a similar level of justifi cation will, ceteris paribus, be required to know that another quote is from Macbeth. More generally, if a level of justification n is required to know that p, then a similar level of justification is required to know truths that are sufficiently similar to p. What counts as a “sufficiently similar” truth (or which cases qualify as “alike”) is a difficult conceptual issue. I doubt we can articulate a set of gen eral criteria to draw this distinction in a principled way. Nevertheless, it is reasonable to suppose that our judgments about these cases will sufficiently coincide for us to say tolerably well, in particular cases, which truths re quire similar levels of justification. This is illustrated by the two examples discussed above. We intuitively recognize that if a certain threshold must be met to know the number of active satellites orbiting Mars, then, ceteris paribus, a similar threshold must be met to know the number of inactive satellites orbiting Mars. In other words, we recognize these cases are suf ficiently similar to demand similar thresholds of justification. This is also true for the case of testimonial knowledge mentioned above. Why do we treat like cases alike? Is there some deeper explanation for why this principle seems to inform our judgments about knowledge? I’ll conclude by sketching one possible answer, namely, that using comparable epistemic standards has some utility, whereas applying widely divergent
5 The policy of treating like cases alike might threaten to leach into absolutely all cases via the transitivity of “x is similar to y.” However, we can resist this by proposing a two-tiered policy: only add those cases that are similar to a base case (i.e., to a case captured by the original communal impurist principles).
IMPURE KNOWLEDGE
|
97
thresholds in similar contexts would be a confusing and inefficient epi stemic strategy. To illustrate this, let’s revisit the sand on Waikiki Beach. As I’ve argued, it is doubtful that anyone does (or would) care about the number of grains of sand currently on Waikiki Beach. I, for one, cannot imagine anyone rea sonably appealing to the proposition “Waikiki Beach has n grains of sand” in their practical reasoning. However, it is not unreasonable to imagine an arenophile (someone who collects sand as a hobby) who might care about counting grains of sand in certain contexts. For example, Dilbert the arenophile might care that he has found n grains of gypsum in a small heap of sand, since gypsum is a rare type of sand. In order to know he has n grains of gypsum, Dilbert must count them carefully. Put differently, he must meet a certain minimum threshold to know there are n grains of sand. Further, our ability to imagine someone with Dilbert’s practical interests explains why, on the communal impurist view, one must satisfy a suffi ciently high minimum threshold to know there are n grains of gypsum. And if the thresholds for knowledge are fairly standardized across contexts (bracketing cases in which someone has especially high stakes), then we should expect agents to be in a similarly strong epistemic position to know, for any similarly sized heap of sand, there are n grains in it. (The heap of sand must be “similarly sized” because it becomes more difficult to count grains of sand as the number goes up.) But why expect the thresholds to be symmetrical in this way? We gain clarity on this question by considering an alternative view, namely, the possibility of widely diverging standards for knowledge across similar cases. When we reflect on this possibility, we can recognize how epistemically inefficient such a practice would be. First, it introduces opportunities for miscommunication because, on this view, we would be relying on each individual to judge whether or not a given proposition is one in which an inquirer might reasonably have stake, but people’s judgments about this might not coincide. As a result, there will be diverging thresholds for knowledge, which makes epistemic evaluation less effective. Second, the use of widely diverging standards in similar cases would place additional cognitive burdens on each member of the epistemic community, since they must keep track of the various thresholds for knowledge. But this might stretch our cognitive load beyond its capacities. A community-wide epi stemic system that worked this way would be inefficient because widely diverging standards would make it more difficult to pool and share poten tially useful information, and informational exchange might get bogged down. Third, there seems to be something deeply inconsistent about an
98
|
What’s the Point of Knowledge?
epistemic system that allows two informants, in roughly similar situations, to both merit the honorific label “knower,” even though one of them is in a strong epistemic position and the other is in a very weak epistemic position. If this account were correct, we would demand far more justifi cation to know the number of active satellites orbiting Mars than to know the number of inactive satellites in Mars’s orbit. Likewise, we would de mand more or less expertise to locate a specific passage in Shakespeare depending on whether the passage is one we could reasonably imagine someone caring about. But an epistemic system that permitted such differing thresholds would strike us as inconsistent. In contrast, by demanding similar thresholds in similar cases, we achieve a consistency that promotes a deep kind of coordination in our basic epistemic practices.6 It allows for the coordination of epistemic rule following across the community, which in turn makes testimony more trustworthy and reliable. In this way, our demand for similar knowledge thresholds is importantly connected to the communal view of knowledge. The practice of having similar thresholds gives rise to a more efficient epi stemic economy because the use of similar knowledge thresholds allows us to each recognize that we are following the same rules and standards. As a result, we don’t have to worry about people applying diverging thresholds in similar cases. When consistent thresholds are shared, I can trust that you will draw the same conclusion from an evidential basis that I would. For these reasons, we typically demand similar levels of justification for similar cases, regardless of whether we can reasonably imagine someone with no practical stake in the proposition. (As mentioned, however, the epistemic standard may go up in cases where an individual’s practical in terest is particularly pressing.) Brown worries that setting a minimum standard for knowledge poten tially gives rise to the problem of arbitrariness (2014: 187). What would fix the minimum standard for knowledge? Without a plausible answer, we land right back in the threshold problem, which is the problem of how to provide an account of what fixes the level of justification required for knowledge in a nonarbitrary way. But the problem of arbitrariness does not arise for the communal impurist. The minimum standard for knowledge derives from our need to identify good sources of information to members of our community. Basic practical concerns thus generate a standard that is fitting to certify
6 See Dogramaci (2012) for a similar approach to rationality.
IMPURE KNOWLEDGE
|
99
actionable information for many people with diverse interests. This sets a general threshold for knowledge in a nonarbitrary way. The level of jus tification needed for knowledge is that which puts the agent in a strong enough epistemic position for her to serve as a reliable source of action able information for her practical reasoning environment, and minimally for the normal standards of her epistemic community. Brown might want to reject communal impurism on the grounds that “appeal to imaginary scenarios does not properly restrain the requirements for knowledge” (2014: 188). For example, Brown worries that we can im agine a scenario in which a lot rests on the proposition that Mars has two moons as well as a scenario in which very little rests on that proposition. This is supposed to illustrate that “appeal to merely imaginary scenarios does not appear to give us a determinate answer to the threshold problem since we can imagine any scenario we like with any stakes whatsoever” (2014: 188). According to the communal approach, however, the threshold for knowing a proposition that is irrelevant to any practical reasoning situation faced by the subject is not set by whatever scenario the subject happens to imagine. Rather, the threshold for knowing an irrelevant proposition is set by the possibilities that are relevant alternatives to members of the epistemic community that might draw on the information. Of course we cannot anticipate every possible scenario in which such information might be put to use, nor must we imagine all of these scenarios when ascribing knowledge in daily fife. But, as I’ve argued, we do have a rough idea of what normally counts as having done enough to establish the propriety of a knowledge claim. As information-dependent and information-sharing creatures, we are accustomed to the fact that others often turn to us for in formation about various topics. But since it is hard to say in advance who will come to rely on our judgments, the threshold for knowledge gravitates toward a level high enough to respect the “typical” or “normal” stakes of others who might appeal to these judgments. This is the communal threshold for knowledge, but this standard can be overridden when the stakes (whether our own or someone else’s) in the relevant practical rea soning environment are especially high.
4.7 Against Subject-Centered Impurism I have argued that, contrary to what many impurists believe (e.g., Hawthorne and Stanley 2008), it is not necessarily the practical reasoning
100
|
What’s the Point of Knowledge?
situation faced by the subject at a time that determines whether she knows. If a knower is someone on whom many people can rely, it makes little sense to think that the epistemic standards would be so closely tied to the interests of the subject. We are interested in whether an informant is epistemically positioned to provide information that is fitting to the community’s interests, not just the subject’s interests. Thus, the threshold for knowl edge will be strongly conditioned by the diverse needs and interests of the community to which the subject belongs. Having such a standard is what makes pooling and sharing information easier. Thus, we have a reason to prefer community-centered impurism to subject-centered impurism.7 I agree with subject-centered impurists that the threshold for knowl edge can be overridden when the stakes for the subject are especially high. However, my view does not collapse into subject-centered impurism for two reasons. First, the practical interests of the subject do not neces sarily set the standard for knowledge. Rather, I have argued that whether someone knows is partly determined by the interests of other members of the community; for instance, I should not recommend to you an informant who barely meets the communal standard if I am aware that your stakes are exceptionally high. Second, I have argued that the subject’s stakes cannot drag the threshold for knowledge lower than the communal level (and I have explained why this asymmetry is not problematic). In contrast, subject-centered impurists claim that a person in a relatively low-stakes situation counts as knowing that p if that person is able to rule out the notp possibilities from a range of alternatives that is narrower than the range demanded by the community. In these ways, communal impurism is im portantly different from subject-centered impurism.
4.8 Concluding Remarks Standard formulations of impurism are vulnerable to serious objections, which has cast doubt on the impurist’s ability to solve the threshold problem. The root of the problem for impurism has been the idea that the level of justification required for knowledge is set by the subject’s practical reasoning situation. However, I have outlined a new way of implementing the impurist strategy: the threshold for knowledge is set by a communal standard, but the required level of justification can increase if the prac tical reasoning situation calls for it. This explains why we often ascribe
7 Greco (2008) and Henderson (2009) make a similar point.
IMPURE KNOWLEDGE
|
101
knowledge as if it were in no way sensitive to context (i.e., when one claims to know, one makes no mention of the context sensitivity of the truth of one’s claim, and one’s claim to know must stand the test of sub sequent challenges as though its content were absolute), and yet we also ascribe knowledge as if it were in some way sensitive to context (i.e., one may claim to know something in one situation, but in a different situation, when the stakes have gone up, one hesitates over the same claim, even if one has the same evidence). By adopting this view, we end up with a plau sible version of impurism that answers the threshold problem and is not vulnerable to the criticisms against the unity and relevance approaches.
102
|
What’s the Point of Knowledge?
chapter 5 | Pluralism about Knowledge
There are many uses of the concept of knowledge, and in sizing up the concept there is no a priori reason to favor its use in picking good informants over its other uses. The task should be to ascertain what the concept must be like to afford all these uses. —FREDERICK SCHMITT (1992: 557)
I have been arguing that the purpose of the concept of knowledge is to identify reliable informants. This practice is necessary, or at least deeply im portant, because we are information-sharing creatures who are profoundly epistemically dependent on others. But why think the concept of knowledge has just one purpose? Perhaps it serves multiple functions. Recently, there have been a variety of proposals about what role our knowledge concept plays in epistemic evaluation. It has been suggested, for instance, that we think and speak of knowing in order to • • • • • • • • •
signal the appropriate end of inquiry track the epistemic norm governing warranted assertion track the epistemic norm governing practical reasoning distinguish between blameworthy and blameless behavior provide assurance to others encourage good testimony give credit for true belief counter doubts indicate that one is certain1
1 These views are suggested, respectively, by Kappel (2010), Williamson (2000), Hawthorne and
This plurality should make us wonder whether the concept of knowledge has just one function. And even supposing it has a single function, how do we decide between these various proposals? To further complicate things, these putative roles are concerned only with adult human knowledge. Presumably, a fully general account of knowledge must confront a broad range of phenomena that includes our tendency to attribute knowledge to children, nonhuman animals, groups, and institutions. All this makes it incredibly plausible to think we speak of knowing for purposes beyond the one I have emphasized. Why, then, should we believe that the func tion of the concept of knowledge is to identify reliable informants? In this chapter, I try to answer this question.
5.1 Pluralism and the Point of Knowledge Within the first few pages of Knowledge and the State of Nature, Craig leaves open the possibility that the concept of knowledge might play mul tiple social roles: Any society that has a well-developed language, sufficiently well devel oped for us to be able to say that it exercises a concept even approximately identifiable with our concept of knowledge, consists of creatures that have reached a considerable degree of mental complexity. Any number of dif ferent sorts of need may, for all we know to the contrary, follow in the wake of this complexity; so there is no a priori reason to think that we are tied by methodological principles to considering only needs of the very basic kind that I have actually tried to restrict myself to. (1990: 4)
By the time we get to page 18, however, we find Craig saying that “the concept that we are constructing” is “one whose sole purpose is to act as a marker for approved sources of information” (emphasis mine). Has he already closed the door on pluralism about the role of knowledge? Not exactly. Craig’s strategy is to “test the explanatory powers of the simple before resorting to the complex” (1990: 4), so he treats his hypoth esis as one that might ultimately need modification—for instance, if it proves too simple, imprecise, or mistaken. Indeed, Craig does modify his initial hypothesis. He starts by imagining an individual inquirer with the
Stanley (2008), Beebe (2012), Austin (1946), Reynolds (2002), Greco (2003), Rysiew (2001), and BonJour (2010).
104
|
What’s the Point of Knowledge?
problem of having to determine the truth as to whether p, and he considers the possibility of some proto-concept of knowledge that might have been used to flag informants who are good enough for our own immediate purposes. Ultimately, however, Craig is led to a more sophisticated hypoth esis by reflecting on the fact that humans often rely on the testimony of others. By considering a community that involves a number of individuals with a shared problem of having to figure out the truth on diverse topics, Craig is led from the primitive notion of a reliable informant (see §2.1) to the idea of an “objectivized reliable informant” (see §2.2 and §2.3). The latter notion is the alleged core of our concept of knowledge. It is at this point, however, that Craig firmly settles on the idea that the purpose of the concept of knowledge is to identify reliable informants (in the objectivized sense). Throughout the rest of his book, no other hypotheses about the role of this concept are seriously considered. This does not imply that every use of the word “know” is concerned with the informant-flagging function. Craig is careful to allow that speaking of knowing is not always concerned with identifying reli able informants (1990: 97). Instead, there will likely be circumstances in which the word “know” is used to fulfill one kind of need but not another, as well as circumstances in which it succeeds in fulfilling multiple needs (I will elaborate on these ideas shortly). Thus, it is no part of the present view that attributions of knowledge play a single communicative role. We may usefully distinguish two ways of understanding talk of the “function” of a concept.2 One common way of thinking about the func tion of a concept is in terms of what speakers may use the concept to do on a particular occasion. For example, it is entirely appropriate to say to someone, “I know you can do it!” even if you have doubts that she will pre vail, since one may say this in order to provide reassurance. In such cases, we use the word “know” in a way that goes beyond its characteristic pur pose.3 In contrast, talk of the function of a concept can also concern what Robin McKenna calls “long-standing and extensive patterns of use within a linguistic community” (2013: 336). Here the focus is on how a com munity uses certain concepts and words to meet needs that are common
21 take this distinction from McKenna (2013), who uses it to clarify the sense in which the function of a word is relevant to its semantics. 3 As mentioned in §1.4, we may distinguish what the function of something is as opposed to what something may function as. For example, the point of a hammer is to drive nails into objects, but hammers may also be used as paperweights, weapons, and so forth.
PLURALISM ABOUT KNOWLEDGE
|
105
to all humans. Instead of focusing on particular uses of the word “know” at particular times, we appeal to extensive and long-standing patterns of use within a whole community. When I speak of the “function” (purpose, point, role) of the concept of knowledge, it is this latter sense of function that I have in mind. We may think of this as the primary function of the concept of knowledge. This distinction can be drawn in another way. We may contrast the point of the concept ofknowledge with the point of specific attributions ofknowledge (or the speech act of ascribing knowledge). The set of considerations for one might be different from the other, since the purpose of attributing knowledge in a cer tain situation might differ from the point of the concept in question. It is not a necessary condition that each proper attribution ofknowledge serves its primary (or characteristic) purpose. Consider the following analogy: it is perfectly ac ceptable to buy a house in order to show off one’s wealth, even though this is not the primary purpose of houses. The point of houses, presumably, is to provide shelter, but houses may be used in ways that go beyond their primary purpose. Why should we think Craig has properly identified the primary purpose of the concept of knowledge? Many scholars have raised this objection. In his review of Craig’s book, Fredrick Schmitt says the concept ofknowledge has many uses and there is no a priori reason to favor the role of identifying reli able informants over other uses (1992: 557). Likewise, Gerken says there are no conclusive arguments for privileging one function over any of the others. In his book On Folk Epistemology, he accuses function-first epistemologists like myself of providing “no principled, non-speculative reason for privileging one function over the rest” (2017: §9.2.b). Rysiew (2012) considers this a prima facie problem for Craig’s proposal.41 take this objection seriously and the point of this chapter is to answer it. I’ll start by explaining why this objection is a bit uncharitable. The point of Craig’s book is to extensively argue for the fruitfulness of his proposal: he attempts to retroactively vindicate his hypothesis by showing that it can account for a range of intuitions about cases and epistemolog ical theories. Indeed, Knowledge and the State of Nature is largely dedi cated to showing that the informant-flagging hypothesis explains several features of our concept of knowledge that have been identified by various philosophical theories, including Nozick’s tracking account, Goldman’s
4 Rysiew focuses on the roles of knowledge attributions rather than the concept ofknowledge. As I’ve just suggested, however, we may distinguish the point of the concept ofknowledge from the various roles that the word “know” might serve in our communicative interactions.
106
|
What’s the Point of Knowledge?
causal theory, reliabilism, and intemalism about justification.5 All this should lead us to think it is no mere coincidence that our central hypothesis captures the primary purpose of the concept of knowledge. Thus, anyone who wishes to argue that we should focus on another hypothesis (or a combination of hypotheses) would need to show that these alternatives do at least as well on this score. Still, we could strengthen our account by explaining why the informant flagging function is the most fundamental by comparing it to other puta tive functions. While I cannot hope to address all the possible functions of our knowledge concept, I will consider a variety of popular alternative proposals.
5.2 Signaling the Appropriate End of Inquiry As inquirers we seek reliable information on which to base our actions. Plausibly, the basic aim of belief is to store correct information, for the rational person wants her actions to be guided by true beliefs. However, the process of inquiry is potentially open-ended. Suppose, for instance, I want to know where my cat is hiding. What does it take to know that my cat is not currently in the kitchen? Must I check under the kitchen table? Must I look behind the kitchen door? Must I open every closed drawer in the room? Must I prove that I am not dreaming? Must I know that I am not a brain in a vat? As suggested in §3.6, inquiry can be thought of as a process of ruling out various possibilities. But it is almost always possible to continue one’s inquiry (logical possibilities are endless). Thus, we need a point at which people can reasonably terminate inquiry.6 According to Kvanvig (2009: 344), this stopping point is cognitively valuable because it satisfies one of the platitudes about the functional roles of knowledge: it signals the point of legitimate inquiry closure. We can make this idea vivid by considering the bank cases outlined in the previous chapter (see §4.4). In the low-stakes case, it is intuitive to say that Keith knows the bank is open tomorrow, and for this reason it seems he needn’t go inside to confirm the bank’s hours of operation. In
5 Cf. Craig’s comment that “the proof is in the pudding” (1990: 4). Kusch and McKenna (2018) provide a more detailed discussion of the idea that Craig partially vindicates these (and other) theories of knowledge. 61 say “reasonably” terminate inquiry because I want to rule out cases in which one simply no longer feels like investigating. Also, false belief would not be a successful end to inquiry, which explains why knowledge is factive.
PLURALISM ABOUT KNOWLEDGE
|
107
the high-stakes case, however, it is intuitive to say he does not know the bank will be open tomorrow; consequently, it seems appropriate for him to check further (e.g., he should go inside and ask an employee). This provides some support for the idea that “knows” signals the point at which further inquiry is unnecessary.7 Cohen’s airport cases (2000: 95) provide another compelling illustra tion that inquiry reasonably ends at knowledge. He asks us to imagine a low stakes situation in which Mary and John are at the L.A. airport and contemplating taking a certain flight to New York. They want to know whether the flight has a layover in Chicago, and they overhear someone ask if anyone knows whether the flight makes any stops. A passenger named Smith replies, “I do. I just looked at my flight itinerary and there is a stop in Chicago.” Now, if you find it intuitive to say that Mary and John know the flight has a layover in Chicago, then it seems they needn’t check any further. But imagine a high-stakes version of this situation in which it is very important that Mary and John arrive in Chicago. Further, imagine that Mary questions the possibility that the itinerary is reHable on the grounds that it could contain a misprint or the schedule might have changed. If you find it intuitive to say they do not know the flight has a stopover in Chicago, then it seems they should check with the airiine agent to find out. Again, inquiry seems to end at knowledge. Some scholars have argued there are exceptional cases in which speakers are wining to ascribe knowledge to somebody and yet it is rea sonable to inquire further.81 will discuss these cases in §5.3 and §5.4. For now I will simply Say it is uncontroversial that “S knows that p” typically conveys that there is no need for further investigation. According to Kappel (2010), Kelp (2011), and Rysiew (2012), Craig does not consider the special role that knowledge ascriptions have for in quiry independently of whatever need we might have for identifying reH able sources of information. They claim that a distinct role for knowledge ascriptions in inquiry is to certify information as being such that it may, or even should, be taken as settled for the purposes of one’s practical and theoretical dehberations. We need a way to command a switch of attention away from these further uneliminated possibiHties—a way to signal when inquiry has gone on long enough.
7 This example has been used to motivate epistemic contextualism, but I am using this example for another purpose. I will discuss contextualism in chapter 7. 8 See Rysiew (2001: 497), Brown (2008: 1444), Bach (2010: 118-119), Reed (2010: 228-229), and Gerken (2011).
108
|
What’s the Point of Knowledge?
Kvanvig, Kappel, Kelp, and Rysiew are surely right about our need for an inquiry-stopper; but this idea is not incompatible with Craig’s hypoth esis about our need to identify reliable informants. Rather, I think these two functions are just different sides of the same coin. This is because the way to reasonably terminate inquiry is by identifying a sufficiently reliable informant. A reliable informant as to whether p is someone from whom we can take it that p, which is to say that we treat her word on whether p as set tling the question of whether p. For example, I take it from the bank teller that the bank will be open on Sunday, but I do not take it from a street psy chic that I should invest my savings in a certain stock. In saying “S knows that p” we are saying that S is sufficiently reliable for one to take it that p. This connection explains why the functional role of flagging reliable informants also serves to mark the point at which further inquiry is unnec essary. Spending more time and resources to continue one’s inquiry would be impractical: continuing to inquire beyond this point would commit us to paying higher “informational costs” that are not worth the lessened risk of being wrong. Everyday life does not demand that our chances of being wrong are absolutely zero. If our informant knows, then there is no need to investigate further. Attributing knowledge to someone is a way of expressing the attitude that someone’s epistemic position (with respect to a given proposition) is good enough to stop further inquiry. That’s precisely what makes such an informant reliable enough. The dual roles of identifying reliable informants and terminating inquiry can therefore be given a unified treatment. Instead of replacing Craig’s hy pothesis with the view that knowledge ascriptions are used to terminate inquiry, both hypotheses appear to be mutually supporting. In fact, the role of identifying reliable informants appears to be more fundamental than the inquiry-stopping function because the former is explanatorily prior. That we have found a reliable informant explains why we may terminate inquiry at a certain point. Without such an explanation, it is unclear when we should reasonably end inquiry.
5.3 Tracking Epistemic Norms Knowledge has also been associated with the role of tracking the epi stemic norms of assertion and practical reasoning. This is because knowl edge ascriptions (and denials) align with natural assessments of assertion and practical reasoning in ordinary language. Here’s an example: it seems appropriate to challenge assertions by asking the asserter, “How do you
PLURALISM ABOUT KNOWLEDGE
|
109
know that?” Similarly, we can rightfully criticize a person’s actions or reasoning when that person acted without knowledge; for instance, “You shouldn’t have gone down this street, since you didn’t know that the res taurant was there.” In addition to questions and challenges, there are data from lottery assertions: one is not warranted in asserting, “Your ticket is a loser” (or using this as a premise in practical reasoning) on the grounds that one does not know that the ticket is a loser.9 These considerations illustrate that knowledge-talk is tightly connected to warranted assertion and practical reasoning. In contrast, competing accounts in terms of evidence, justified belief, warrant, or reliability do not straightforwardly align with ordinary language usage.10 Admittedly, theorists who investigate the epistemic norms governing assertion and practical reasoning typically do not make any claims about functional role, for they are not working within the function-first framework I have adopted. However, these theorists often support their preferred view about epistemic norms by appealing to knowledge ascriptions and denials, which assumes that knowledge-talk is tracking the relevant epistemic norms. For the sake of convenience, we can formulate the relevant epistemic norms as follows: KA: One is in a good enough epistemic position to assert that p if (and only if) one knows that p. KPR: One is in a good enough epistemic position to rely on p in prac tical reasoning if (and only if) one knows that p.
I will refer to these as “the knowledge norms.” Occasionally I will consider these norms together because in many cases our intuitions about assertion, practical reasoning, and knowledge go together. For example, DeRose’s bank scenarios have been used to show that the epistemic propriety of both assertion and practical reasoning depend on whether one is thought to have knowledge. For this reason, many arguments for (and against) the knowl edge norm of assertion apply mutatis mutandis to the knowledge norm of practical reasoning, and vice versa. Three points of clarification are needed. First, I put the necessity claim of KA and KPR in parentheses to indicate that each norm can be under stood as a sufficiency claim, a necessity claim, or a biconditional. Second,
9 See Williamson (2000: 246, 252), Unger (1975: 250-265), Hawthorne and Stanley (2008: 571), and Hawthorne (2004: 21). 10 Cf. Hawthorne and Stanley (2008: 573) and Gerken (2015).
110
|
What’s the Point of Knowledge?
a contextualist would need to rephrase these norms to accommodate the context sensitivity of “knows” (see DeRose 2002). Third, I have followed Hawthorne’s (2004) formulation of the practical reasoning norm, but my discussion applies to other formulations as well, such as Fantl and McGrath’s (2009) “acting on p” and Hawthorne and Stanley’s (2008) “treating p as a reason for action.” For present purposes, I treat the epi stemic norms of assertion and practical reasoning as relevantly similar, al though this is a simplification (see Gerken 2011). Montminy (2013) argues that assertion and practical reasoning must be governed by the same epi stemic norm, but Brown (2012) and Gerken (2014) raise doubts about the plausibility of this view. I do not claim the norms of assertion and practical reasoning coincide. Rather, I merely assume that some of the arguments used to support (and criticize) both KA and KPR are sufficiently similar to allow us to discuss these norms together. The knowledge norms have attracted several defenders, although these norms have also been contested for many reasons. I will mention three. First, there are circumstances in which asserting that p and relying on p in one’s reasoning seems proper or warranted even though we would not say that the agent knows p. In particular, cases of warranted false belief and warranted yet lucky true beliefs (i.e., Gettier cases) go against the idea that knowledge is necessary either for warranted assertion or for practical reasoning (Brown 2008a; Gerken 2011; Smithies 2012). Although people are typically unwilling to ascribe knowledge in these cases, this does not change our view about the epistemic appropriateness of one’s reasoning or assertion. This indicates that knowledge ascriptions might not track the necessity direction of the knowledge norms. Second, our tendency to cite knowledge when defending or criticizing actions does not necessarily indicate that knowledge ascriptions track the relevant epistemic norm. As Brown (2008a) and Reed (2010) point out, we also cite epistemic states both stronger and weaker than knowledge when defending and criticizing actions. For example, I might say, “You shouldn’t have left so late because you weren’t certain that the trains were operating after midnight.” I might also say, “When I arranged your blind date, I had every reason to believe that her divorce was final” (see Brown 2008a). That we cite factors both stronger and weaker than knowledge when defending and criticizing action counts against the idea that knowl edge ascriptions are tracking the governing epistemic norm. Third, there are plausible cases in which more than knowledge is re quired to rely on p in practical reasoning. Brown (2008b: 176) discusses the case of a surgeon who is about to operate on a diseased kidney. Before
PLURALISM ABOUT KNOWLEDGE
|
111
operating, the surgeon double-checks the patient’s notes, despite knowing which kidney is diseased. The nurse explains what’s happening to a nearby student as follows: “Of course, she knows which kidney it is. But, imagine what it would be like if she removed the wrong kidney. She shouldn’t op erate before checking the patient’s records.”11 This illustrates that knowl edge ascriptions might not track the sufficiency direction of the knowledge norm of practical reasoning.11 12 Defenders of the knowledge norms have responded to these criticisms in a variety of ways. For example, it has been argued that apparent counterexamples to the necessity direction of both KA and KPR are merely excusable violations of the relevant norms. However, this strategy is beset with problems.13 Against the idea that sometimes more than knowledge is required for practical reasoning (as in Brown’s surgeon case), Jonathan Ichikawa (2012) has argued that cases in which knowledge is intuitively present, but action is intuitively epistemically unwarranted, provide no traction against the knowledge norm. It is unlikely that this debate will be settled anytime soon. For the pre sent purpose, I will simply assume there are genuine counterexamples to the knowledge norms. Thus, I will assume knowledge ascriptions are not used to track the epistemic norms governing assertion and practical reasoning. Anyone who denies the knowledge norms faces an explanatory chal lenge: why is the word “know” often appropriately used in the assessment of assertion and practical reasoning? It is undeniable that we use “know” far more frequently than alternative epistemic vocabulary, such as “jus tified” or “reliable” or “warranted,” which gives knowledge a presump tion of importance that these other terms lack. This places a burden on anyone who denies the knowledge norms to explain why knowledge-talk is so prominent in the epistemic assessment of assertion and practical rea soning. I will now consider an answer to this challenge—one that is com patible with the idea that the primary role of the concept of knowledge is to flag reliable informants.
11 Rysiew (2007), Gerken (2011), and Sosa (2015) also think the epistemic requirements on action can be higher than those on knowledge. 12 See Lackey (2011) for an argument against the sufficiency direction of the knowledge norm of assertion. For a challenge to Lackey’s cases, see Benton (2016); for her reply, see Lackey (2016). 13 Williamson (2000: 256), Hawthorne and Stanley (2008: 586), and DeRose (2009: 93-95) appeal to this “excuse maneuver,” while Douven (2006), Lackey (2007b), Brown (2008), and Gerken (2011) criticize it.
112
|
What’s the Point of Knowledge?
5.4 The Threshold-Marking Function What could explain the prominence of knowledge-talk if not the knowl edge norms? Here’s a hypothesis: perhaps “knowledge” normally picks out the epistemic standard for warranted assertion and practical reasoning, but not always. This would explain why competent and rational speakers frequently use “know” when evaluating assertions and practical reasoning (because knowledge is normally required) even though knowledge is not the relevant epistemic norm (because sometimes more, or less, than knowledge is needed). Gerken (2015) suggests this view. He puts forward a thesis called “normal coincidence” according to which the degree of warrant required for knowledge normally coincides with the degree of warrant necessary and sufficient for assertion and practical reasoning (2015: 146; see also Douven 2006: 469).14 On this view, we typically know what we assert and we typically rely on known propositions for practical reasoning, yet knowledge is not the norm governing these actions. As the previous sec tion illustrates, there are some circumstances in which more (or less) than knowledge is required; for instance, more is required when the stakes are very high and less is required in Gettier cases. Assuming these are genuine counterexamples, the knowledge norms are incorrect. Nevertheless, we often cite “knowledge” in epistemic assessment because it is normally a reliably accurate way of indicating that the standard required for epistemically warranted assertion and practical reasoning is met. Why would the level of warrant required for knowledge normally coincide with the level of warrant that is both necessary and sufficient for assertion and practical reasoning? According to Gerken, one of the communicative functions of knowledge ascriptions is to serve as a threshold-marker. In normal cases of epistemic assessment, ascriptions of knowledge mark the threshold of warrant that a person must possess with regard to p in order to properly assert p and to be epistemically rational in acting on (the belief that) p. With respect to practical rea soning, we need such a threshold because there must be some point at which one’s epistemic position provides a strong enough basis for action. “Know” usually indicates that this threshold has been met and so the
14 Actually, Gerken focuses specifically on action, not assertion. He does say that assertion is a speech act and, hence, an act, but he nevertheless sets aside the special case of assertion because he considers it a distinctive kind of act with distinctive features. For simplicity, however, I will ignore this complication and continue to focus on both assertion and practical reasoning.
PLURALISM ABOUT KNOWLEDGE
|
113
agent is warranted in acting. (This is closely related to our need for an inquiry-stopper, since action typically begins where inquiry ends.) With respect to assertion, we have such a threshold because there must be some point at which one’s epistemic position is strong enough with respect to p to serve as a reliable source of information from which others can take it from one that p. Information transmitted via assertions is commonly used as reasons for action, so we generally expect people to assert only what they know. On this view, knowledge ascriptions mark the threshold of warrant for proper assertion and practical reasoning in normal cases of epistemic as sessment. What makes a case of epistemic assessment “normal”? I will not pursue a full characterization of this notion here (see Gerken 2011 and 2015 for a discussion), but Gerken assumes that environmental factors, such as objective frequencies, partly determine whether a context is normal. “Epistemic normality” is perhaps best illuminated by reflec tion on paradigm cases, so I’ll provide two examples. A person in an area where, unbeknownst to him, all of the barn-looking objects are merely bam facades would be in abnormal circumstances because such areas are rare (or only imaginary). In such an environment “knowledge” would not mark the threshold for epistemically warranted assertion or practical reasoning. Similarly, cases in which a subject’s practical stakes are very high are abnormal because usually much less is at stake (most practical reasoning circumstances are not of life-or-death importance). These facts about epistemic normality help explain why assertions and actions can sometimes be reasonable even in the absence of knowledge, as well as why possessing knowledge isn’t always enough.15 Is the threshold-marking function any less important or fundamental than the function of identifying reliable informants? I think these hypotheses are so closely related that they can be seen as mutually supporting. I will suggest, however, that the function of identifying reliable informants is more explanatorily fundamental.
151 am simplifying Gerken’s view here. More precisely, he says that person S meets the epistemic conditions required for her to properly assert p or rationally use (her belief that) p as a premise in practical reasoning if and only if S is warranted in believing that p to a degree that is adequate relative to her deliberative context (see Brown 2008b for a similar view). The notion of a deliberative context can be thought of as the agent’s reasonably believed or presupposed practical context, which is determined by parameters such as alternative courses of action, urgency, availability of further evidence, and the stakes (Gerken 2015: 144). This warrant demand frequently coincides with the degree required for knowledge, which helps explain why knowledge-talk is so central to our practice of epistemic evaluation. However, some deliberative contexts are abnormal and thus require more (or less) warrant than knowledge.
114
|
What’s the Point of Knowledge?
Why would the level of warrant required for knowledge normally co incide with the level of warrant that is both necessary and sufficient for assertion and practical reasoning? Gerken provides compelling answers. In the case of assertion, it is because we need to mark a point at which one’s epistemic position is strong enough with respect to p to serve as a reliable source of information from which others can take it from one that p. In the case of practical reasoning, it is because we need to mark a point at which one’s epistemic position provides a strong enough basis for ac tion. These functions are importantly related: information transmitted via assertion is commonly used as a reason for action, so we generally expect people to assert only what they know. On these points, I completely agree with Gerken. Notice, however, that none of this explains precisely where the level of warrant needed for knowledge is set, nor does it explain why it is set there (wherever that is) rather than elsewhere. We therefore run into a version of the threshold problem outlined in chapter 3. In short, Gerken provides a plausible explanation for why the level of warrant required for knowledge normally coincides with the level of war rant that is both necessary and sufficient for assertion and practical rea soning, but we are left wondering at what point we have enough warrant for knowledge. Two issues are thus left unclear. First, at what point do we have enough warrant to justify action? Second, at what point is our epi stemic position strong enough for us to serve as a reliable source of infor mation for others? It pushes the problem back to say, “the point at which one knows.” What we want to know is how strong an epistemic position that is and why the standard is set at this point. We can answer these questions by appealing to the account I have been developing throughout this book. The point of the concept of knowledge, I claim, is to certify reliable informants to members of our epistemic community. To say that an agent knows that p is to say that she is epistemically positioned with respect to p so as to be a good source of actionable information, which means that we may take it from her that p. Our practical concerns generate a standard that is fitting to certify in formation that is good enough for the practical reasoning situations of many people with diverse interests. As argued in chapters 2-4, we may think of this as a general (or communal) threshold. The level of warrant needed for knowledge is that which puts the agent in a strong enough epistemic position for her to serve as a reliable source of actionable in formation for members of her epistemic community. Nevertheless, there will sometimes be cases in which more, or less, warrant is needed than is required for knowledge. In the case of Brown’s surgeon, for example,
PLURALISM ABOUT KNOWLEDGE
|
115
the cost of error is so high, and the cost that one must pay to check further is so low, that the standard normally required for action (i.e., knowledge) doesn’t suffice. And this is precisely what we should ex pect, since someone with a particularly important reason for wanting to get the right belief on a question is unlikely to rely on the general public standard that an individual knows. Thus, there are two ways in which contextual factors like human interests, purposes, and the like get into the picture: first, they deter mine the general (or typical) threshold for epistemically warranted as sertion and practical reasoning; second, they shape one’s deliberative context (see note 15 for further discussion of a deliberative context). An interesting feature of this view is that “knowledge” is doubly relative. It is relative to the deliberative context and to broader features of the human situation (e.g., our psychology, societal makeup, circumstances, and so on). This idea allows us to make epistemological progress in three ways. First, we may unify the threshold-marking function with the other functions we have considered so far. For example, if one is in a good enough epistemic position to assert p or rely on p in practical reasoning, then further inquiry into whetherp will typically be unnecessary. Further, a common way to reasonably terminate inquiry is by identifying a suf ficiently reliable informant. Thus, we may unify the seemingly distinct goals of identifying reliable informants, signaling the appropriate end of inquiry, and marking thresholds. Second, we may unify these roles in a way that preserves the basic insight that there are strong associations between knowledge, assertion, and practical reasoning, but without succumbing to the existing criticisms against the knowledge norms. Third, we may appeal to my account of flagging reliable informants to explain where the level of warrant needed for knowledge is set, and why it is set at this point rather than elsewhere. This allows us to explain why knowledge terminates inquiry and why knowledge functions as a threshold-marker. As argued in chapter 3, we may solve the threshold problem for fallible knowledge in a clear and nonarbitrary way by ap pealing to our need to identify reliable informants to members of our community. In this way, our central hypothesis provides the backbone for the other functions we have thus far considered. It is for this reason that we may reasonably claim that flagging reliable informants is the primary purpose of the concept of knowledge.
116
|
What’s the Point of Knowledge?
5.5 Distinguishing Blameworthy and Blameless Behavior In “Social Functions of Knowledge Attributions,” James Beebe suggests a different purpose of the concept of knowledge. Drawing upon work in ex perimental philosophy and evolutionary game theory, he argues that the con cept of knowledge is sometimes used “to distinguish which agents are most worthy of blame for inflicting harms, violating social norms, or cheating in situations of social exchange” (2012:222). Let’s call this the blame-flagging function. Beebe intends this hypothesis to supplement rather than replace the informant-flagging function. In doing so, he escapes a prima facie problem for anyone who presumes the concept of knowledge has a single purpose. Unlike Kappel and Kelp, Beebe does not seek to replace one monistic view about the function of the concept of knowledge with another. Rather, he is a pluralist about the purpose of the concept of knowledge. Although Beebe’s hypothesis is intended to supplement rather than re place the informant-flagging function, his proposal still poses a challenge to my view. I have been arguing that the primary point of the concept of knowledge is to identify reliable informants. But why should we think this function is more fundamental than the blame-flagging function? In this section, I try to explain why. Let’s first look at the considerations that mo tivate Beebe’s hypothesis. Beebe draws on two sources of evidence for his view. First, he ap peals to some basic lessons that have emerged from evolutionary game theory about when individuals should cooperate and the conditions under which cooperation can arise and achieve stability within various populations (Maynard Smith 1982; Axelrod 1984). Among other things, Beebe highlights the importance of information exchange to human sur vival and flourishing, our interest in developing strategies for cooperation, and our need to distinguish people who are blameworthy for uncoopera tive behaviors from those who are not (2012: 227). After drawing on many of the lessons we have learned from evolutionary game theory, Beebe concludes that part of the point of having a concept of knowledge is to ful fill our need to distinguish agents who are most worthy of blame. This hypothesis is further supported by work in experimental philos ophy, which provides a complementary source of data for the blame flagging function. Recent investigations in experimental philosophy have found that
PLURALISM ABOUT KNOWLEDGE
|
117
people are strongly inclined to count as knowledge the beliefs of agents who are blameworthy for actions undertaken in light of those beliefs—even when the epistemic positions of those agents do not seem to warrant such attributions and even when the attributors would not otherwise be inclined to attribute knowledge to them. (Beebe 2012: 232)
Specifically, Beebe draws on a previous study he conducted with Wesley Buckwaiter (2010) in which they discovered that people are more likely to say that an agent knew her action would bring about a certain side effect if the side effect were bad than if it were good. This finding builds on Joshua Knobe’s (2003) pioneering work on folk attributions of intentionality, which revealed that the perceived goodness or badness of an action’s side effects will influence people’s ascriptions of intentionality about those side effects. This finding has been dubbed the “side-effect effect.” Beebe and Buckwaiter therefore call their result the “epistemic side-effect effect.” In Beebe and Buckwaiter’s study, participants were either given the “help” or the “harm” version of the following vignette: The vice-president of a company went to the chairman of the board and said, “We are thinking of starting a new program. We are sure that it will help us increase profits, and it will also help/harm the environment.” The chairman of the board answered, “I don’t care at all about helping/harming the environment. I just want to make as much profit as I can. Let’s start the new program.” They started the new program. Sure enough, the environ ment was helped/harmed. Did the chairman know that the new program would help/harm the environment?
On a seven-point Likert scale, ranging from -3 (“the chairman didn’t know’) to 3 (“the chairman did know”), they found that almost twice as many people chose the strongest affirmation of the chairman’s knowl edge in the harm case (67.5 percent) as in the help case (35.5 percent). Since the agent in the help case has epistemically equivalent justification for believing the side effect will occur as the agent in the harm scenario, it seems wrong to maintain that one informant is more reliable than the other. This suggests that the function of the concept of knowledge in such cases is not entirely to flag reliable informants. Rather, Beebe claims the concept of knowledge is being used to distinguish those agents who are most blameworthy. I am open to the possibility that this hypothesis might supplement the idea that the purpose of the concept of knowledge is to identify reliable
118
|
What’s the Point of Knowledge?
informants. Nothing I have said conflicts with the idea that knowledge is sometimes attributed to mark blameworthy behavior. However, I need to explain why I regard the informant-flagging function of knowledge as more fundamental than the blame-flagging function.16 After all, Beebe also appeals to our interest in cooperating with others in ways to foster in formation exchange, the goal of which is to facilitate human survival and prosperity. Thus, I cannot simply appeal to these considerations in order to display the comparative fundamentality of my hypothesis. There are two ways to cast doubt on the idea that the blame-flagging function is as fundamental to our concept of knowledge as the informant flagging function. First, there is ample experimental evidence that the asymmetry characteristic of the epistemic side-effect effect is found in many folk psychological concepts. Beyond knowledge and intention ality, a similar effect has been found regarding folk concepts of desire, deciding, being in favor of, being opposed to, advocating, and believing.17 These similarities strongly suggest there is nothing special about the con cept of knowledge in identifying blameworthy behavior. Rather, as Beebe and Buckwaiter observe, “the various side-effect effects stem from gen eral facts about folk psychological attributions, rather than from features unique to the concepts in question” (2010: 480). Paraphrasing Pettit and Knobe (2009: 590), we may conclude that not only does the impact of moral judgment extend beyond the concept of knowledge, moral judg ment appears to impact just about every concept that involves holding or displaying a positive attitude toward an outcome. For this reason, it is highly doubtful that our concept of knowledge functions primarily to draw the relevant distinction between blameworthy and blameless behavior. Second, we may plausibly explain the asymmetry in these cases as a pragmatic phenomenon. Consider the harm case. People might be inclined to attribute knowledge to the chairman in this scenario because denying or withholding ascriptions of knowledge would generate the unwanted implicature that he deserves no blame for harming the environment. This explanation presupposes that the chairman did not actually know the new program would harm the environment but people said he knew to get across that he was blameworthy. Now consider the help case. People might be disinclined to attribute knowledge to the chairman in this sce nario because it might generate the unwanted implicature that he deserves
16 Beebe does not argue that either of these hypotheses is more primary than the other. Thus, I am not challenging his view by claiming the blame-flagging function is less fundamental. 17 See Pettit and Knobe (2009) and Beebe (2013).
PLURALISM ABOUT KNOWLEDGE
|
119
praise for helping the environment.18 This explanation presupposes that the chairman did know the new program would help the environment but people said he didn’t know to get across that he wasn’t praiseworthy. We thus have two pragmatic stories to tell about this phenomenon. But which of these two accounts is more plausible? I find the latter far more convincing. After all, the vice president of the company told the chairman in no uncertain terms that he (the vice pres ident) was sure the new program would increase profits as well as help/ harm the environment. The vignette also makes clear that the chairman heard exactly what the vice president said and we are provided with no reason to think the chairman should doubt what the vice president told him. Finally, the vignette stipulates that the new program did in fact help/ harm the environment, just as the vice president said it would. Thus, it seems like a clear case of knowledge via testimony. Others have likewise suggested the chairman has knowledge in both cases of Knobe’s original study. Hugh McCann (2005: 739), for instance, says that “in both vignettes, he [the chairman] knows perfectly well what he is doing.” Bertram Malle (2010), Fiery Cushman and Alfred Mele (2008: 172), and Jennifer Wright and John Bengson (2009) all make sim ilar comments. Fred Adams even claims that most people probably think the chairman intended to harm the environment “because they were told that he knew the harm would occur if they went ahead with the program in question” (2006: 261, emphasis mine). Adams’s claim is mistaken because the word “know” does not feature in either case, but it seems clear that participants were all but told the chairman knew the new program would harm (or help) the environment. Thus, the most plausible pragmatic story is that participants said something strictly speaking untrue in the help case (that the chairman didn’t know) because it implied something important and true (that he did not merit praise). It is far less plausible to maintain the chairman actually didn’t know what would happen but that participants said something untrue in the harm case (that he did know) because it im plied something important and true (that he was blameworthy).19 It is therefore puzzling why Beebe takes the experimental data to reveal a robust tendency to overattribute knowledge to blameworthy agents. He
18 Similarly, Adams and Steadman (2004a, 2004b) have argued that the asymmetry found between the harm case and the help case in Knobe’s study of intentionality is a pragmatic phenomenon. 19 This contrasts with the most plausible pragmatic explanation for ascriptions of intentionality. In those cases, it seems the chairman did not intentionally to help/harm the environment because he “doesn’t care at all” about helping/harming the environment. Thus, participants tend to attribute intentionality to convey the chairman is blameworthy.
120
|
What’s the Point of Knowledge?
claims “people are strongly inclined to count as knowledge the beliefs of agents who are blameworthy . . . even when the epistemic positions of those agents do not seem to warrant such attributions and even when the attributors would not otherwise be inclined to attribute knowledge to them” (2012: 232). This assumes the chairman did not gain knowledge via the vice president’s testimony, which (as I’ve argued) is implausible. A better interpretation is that the data points to our tendency to underattribute knowledge to agents who knowingly do something for which they merit no praise—and this can be given a pragmatic explanation. Thus, it is unlikely that these findings reveal much about the folk’s core concept of knowledge. I do not deny that ascriptions of knowl edge may sometimes play the social role of marking out individuals who are the most worthy of blame. On this point, Beebe is likely cor rect. Further, I do not deny there are important conceptual connections between knowledge and blame. As Beebe points out, “In making distinctions of blame, it is important to discriminate between those who have informational access to certain facts and those who do not— and to the quality or kind of informational access they have and the de liberative use they make of that information” (2012: 229). My claim is that we have little reason to think these findings reveal much about the primary purpose of the concept of knowledge. Beebe and Buckwaiter’s survey is likely tapping into pragmatic aspects of knowledge attribu tion and its role in praise and blame, not the semantic content of the concept of knowledge.
5.6 Assurance In her book Assurance, Krista Lawlor defends Austin’s view that knowledge claims are typically used to offer assurance to others. In claiming to know that p one usually performs the speech act of assuring one’s audience that p. For example, if I say, “I know that Luke is hiding in the garden,” I assure my listeners that Luke is hiding in the garden. Let’s call this the assurance-giving function. What does it mean to give assurance to others? On Lawlor’s account, this involves a commitment to having “conclusive reasons” for believing that p (roughly, reasons that would satisfy any reasonable person). This is stronger than merely asserting that p. When I assert that p to my audi ence, I commit myself to having a reason to think p is true (or perhaps as having more reason to think p than not-p), but I do not commit myself to
PLURALISM ABOUT KNOWLEDGE
|
121
having conclusive reasons.20 By offering you assurance, in contrast, I rep resent myself as in a position to provide you with reasons that will settle the matter of p’s truth, thereby permitting you to “stop weighing reasons and get on about [your] business” (Lawlor 2013: 20). For example, the speech act of assuring you that Luke is in the garden involves a guarantee that I have conclusive reasons for thinking that Luke is in the garden.21 If you accept my assurance, you may then stop wondering where he is. Assurance thus plays an important role in the process of inquiry. Inquirers seek out someone who knows—someone who can give sound assurances. It is by giving our assurance to others that we grant them the right to rely on us for epistemic security. Lawlor’s starting point is the function of assurance-giving rather than the function of the concept of knowledge. Her concern “is not the concept of knowledge per se,” but rather the speech act of making a knowledge claim (2015:194 n. 1). Further, Lawlor outright denies that she thinks pro viding assurance is the “constitutive function” of knowledge ascriptions. For these reasons, her account does not conflict with the hypothesis that the primary purpose of the concept of knowledge is to flag reliable informants. It is perfectly conceivable that knowledge claims typically perform the speech act of assuring others even though this is not the primary purpose of the concept of knowledge. Still, we might wonder why knowledge claims would usually perform the assurance-giving function if, as I argue, the primary purpose of our knowledge concept is to flag reliable informants. In what follows, I briefly suggest that the informant-flagging function is more fundamental than the assurance-giving function because the former nicely explains the latter but not vice versa. Suppose knowledge claims typically perform the speech act of assuring others, on the one hand, and yet the purpose of the concept of knowledge is to flag reliable informants, on the other hand. Why would this be? A nat ural answer is that someone who finds a good informant has a reason to feel assured. Inquirers want more than just to acquire a true belief; they want to be relieved of their concerns. An assurer represents herself as in a position to vouch for p with conclusive reasons, which means that no reasonable alternatives survive the assurer’s evidence. This explains why
20 See Lawlor (2013: 12-15) for more on the distinction between assertion and assurance. 21 In chains of assurance giving, an assurer may not be privy to the first-order reasons to believe p. In these cases, the assurer “does not represent herself as having conclusive reasons for p, but only represents that there are conclusive reasons for p” (Lawlor 2013: 18).
122
|
What’s the Point of Knowledge?
knowledge claims provide inquirers with assurance. By finding someone who knows that p, we may reasonably terminate inquiry into whether p and confidently proceed with Our plans, insofar as they depend on the truth of p. Someone who tells me she knows p thereby performs the speech act of assuring me that p, since she represents herself as having reasons or ev idence that settle all reasonable doubts. Assuming I accept my informant’s assurance, I may stop inquiring into whether p and dismiss whatever reasons I had to think that not-p. In short, I may rest assured that p. In con trast, it would be mysterious why knowledge claims perform the speech act of providing assurance if they didn’t pick out reliable informants and allow us to settle all reasonable doubts. We may therefore think of assurance as a byproduct of the primary function of the concept of knowledge. The reason we have a concept of knowledge at all, as I’ve argued, is to mark out people on whom we can rely for actionable information. We use this concept to indicate that someone’s epistemic position is sufficiently strong to rule out all the al ternative possibilities that the reasonable person would want eliminated before claiming to know p. However, saying, “I know” does more than just signpost our epistemic position—it allows us to take “a new plunge” and give others a guarantee of something (see Austin 1979: 100). Assurance can thus be regarded as an extra feature that “tags along” with the concept of knowledge. To make this point clearer, consider the following analogy. We likely have the concept of “fun” to pick out a cer tain kind of value; however, the phrase “This is fun” is often used not just to pick out a value but also to encourage people to participate in certain activities. We may therefore say the point of the concept of “fun” is to identify a certain kind of value, but uttering “This is fun” often performs the speech act of encouraging others to participate.22 Likewise, we may say that knowledge claims typically perform the speech act of assuring others even though this is not the primary point of the concept. I doubt Lawlor will find any of this objectionable. Her reasonable alternatives account of knowledge is a species of the same genus as the relevant alternatives view I developed in chapter 3. On her view, an agent knows that p only if she is able to rule out all the reasonable alternatives to p, which are those alternatives a “reasonable person” would want ruled out.23 We may express Lawlor’s point another way by importing my own terminology developed earlier in this book: an agent knows that p only if
22 Thanks to Krista Lawlor for suggesting this example. 23 Chapter 5 of Assurance provides a detailed discussion of the “reasonable person.”
PLURALISM ABOUT KNOWLEDGE
|
123
her epistemic position is strong enough to rule out all the not-p possibilities that are relevant alternatives to members of the epistemic community that might draw on the agent’s information. In short, we both think the relevant alternatives are those that are fitting or reasonable to members of the epi stemic community. Moreover, Lawlor and I agree on the following points: knowledge claims are used to represent oneself as ready to terminate inquiry; one who claims to know that p represents oneself as having reasons that other reasonable people will find empirically adequate; knowledge plays a cru cial role in practical reasoning; and there must exist a default or communal standard of knowledge to effectively communicate with others. We both also think practical interests can affect what alternatives count as reason able in a situation, and we both put limits on how an agent’s practical interests affect the set of reasonable alternatives. Thus, I find very little to disagree with in her view. Rather than being forced to choose between her account and my own, I see them as mutually supporting.
5.7 The Descriptive Use What else might one do with the statement “I know”? The simplest an swer is one claims to know something. This is true in both first- and thirdperson ascriptions of knowledge, as well as knowledge ascriptions to past, present, and future individuals. For example, we might say, “In 1616, Galileo knew that the earth orbits around the sun.” We need not say this to provide anyone with assurance or to distinguish blameworthy behavior from blameless behavior—it is just a point about Galileo having sufficient evidence to know (see Gerken 2015: 226). Thus, one might wonder: is the purpose of the concept of knowledge simply to refer to knowledge, or to report our epistemic position to others? Craig considers a version of this idea. He asks, “Couldn’t it just be that knowledge, like water, is common and important stuff, and that the purpose of the concept is simply to enable us to think and talk about it?” (1990: 3). The idea that we have the concept of X in order to refer to X is a plausible and straightforward explanation for many of our concepts. In the case of water, for instance, it is immensely plausible that we have this concept because it makes it possible for us to think and talk about water, which is something every community has an obvious need to do. Likewise, we presumably have a concept of gold because it allows us to think and talk about gold, which is a common and valuable substance.
124
|
What’s the Point of Knowledge?
In the case of knowledge, however, this type of explanation is not very plausible. To claim the purpose of the concept of knowledge is to refer to knowledge presupposes that the concept of knowledge is determined by the nature of knowledge itself and our experience of it. Put differently, it seems to suppose that knowledge is a natural phenomenon that can be understood and studied independently of our concept of knowledge, just as water is a natural phenomenon that can be understood and studied inde pendently of our concept of water. As argued in §1.5, however, knowledge is not something we find in the world, like gold or water, but rather is something we impose on it, like marriage and SUVs. As a social or arti factual kind, the identity conditions for knowledge are deeply intertwined with human interests and intentions. On this picture, it makes little sense to say the purpose of the concept of knowledge is simply to refer to knowl edge. Indeed, one of the central ideas in this book is that we cannot make sense of what knowledge is without reflecting on the role of the concept of knowledge in our cognitive ecology. For this reason, it is highly unlikely that the point of the concept of knowledge is simply to refer to knowl edge. At the very least, that proposal requires us to have a very different view about the metaphysics of knowledge than the pragmatist view I am defending here.
5.8 Group Knowledge Epistemology has largely centered on the conditions for an individual’s belief to qualify as knowledge, but knowledge is also widely and sys tematically attributed to groups made up of individuals.24 One might say, “The FBI knows where the terrorists have planted the bomb” or “Barnes & Nobel knows it’s losing the battle over the book market to Amazon.” Now, if we assume that the purpose of knowledge attributions is to iden tify reliable informants, then it follows that group knowledge attributions also identify reliable informants. However, Lackey (2012) doubts this is true. She considers several examples of group knowledge attribution that systematically fail to identify reliable informants. She concludes that the informant-flagging function is not the primary purpose of knowledge attributions. I will now defend my proposal from this objection. Central to Lackey’s argument is the idea that a “reliable informant” must satisfy the following necessary condition, which she calls BELIEF:
24 See Goldman (2004), Schmitt (1994), and Hakli (2007).
PLURALISM ABOUT KNOWLEDGE
|
125
S is a reliable informant with respect to the question whether p only if either p and S believes that p, or not-/? and S believes that not-/? (Lackey 2012: 274). In other words, reHable informants must be reliable believers. Assuming this is correct, it follows that the purpose of knowledge ascriptions is to identify informants who at least either believe that p when p, or beHeve that not-/? when not-/?. However, Lackey claims that group knowledge attributions often fail to identify reliable informants in this sense; in particular, we often ascribe knowledge to groups of individuals in which there is no reHable believer. Lackey considers two accounts of group beHef: the summative (or re ductionist) view and the nonsummative (or inflationary) view. According to the summative view, a group’s believing thatp can be understood in the minimal sense that all or some of the members of the group beHeve that /?. As Anthony Quinton puts it, “To say that the industrial working class is determined to resist anti-trade union laws is to say that all or most in dustrial workers are so minded” (1975: 17). Precisely how to aggregate the individual beliefs of group members is a tricky issue. Does the belief of the group reduce to the beHef of all its members, or the majority of its members, or to a single dictatorial member (see List 2005)? I leave this matter open. Assuming the belief of a group is reducible to all or some of its members, the weakest version of this view will hold that there must be at least one reliable behever in a group to properly ascribe knowledge to that group. A problem for this view, according to Lackey, is that we often attribute knowledge to groups where not a single member of the group has the belief in question. Since these knowledge attributions are not flagging any behevers, they cannot be flagging any reHable behevers. Lackey provides two cases where it seems acceptable to attribute group knowledge and yet not a single member of the group holds the relevant be Hef. The first arises when group knowledge is determined by aggregating the judgments of the individual members. Lackey (2012: 248-249) cites the following case by PhilHp Pettit:
Judgment Aggregation'. The employees of a company are deciding whether to forgo a pay-raise in order to spend the saved money on implementing a set of workplace safety measures. The employees are supposed to make their decision on the basis of considering three separable issues: “first, how serious the danger is; second, how effective the safety measures that a pay sacrifice would buy is likely to be; and third, whether the pay-sacrifice is bearable for members individually. If an employee thinks that the danger is sufficiently serious, the safety measure sufficiently effective, and the
126
|
What’s the Point of Knowledge?
pay-sacrifice sufficiently bearable, he or she will vote for the sacrifice; oth erwise he or she will vote against” (Pettit 2003: 171). The company’s three employees vote in the following way:
Serious danger? A. Yes
Effective measure? No
Bearable loss? Yes
Pay sacrifice? No
B.No C. Yes
Yes Yes
Yes No
No No
If the group’s [judgment] is determined by how the members vote on the premises, then the group’s conclusion is to accept the pay sacrifice since there are more ‘Yes’s than ‘No’s in each of the premise columns. In such a
case, “the group will form a judgment on the question of the pay-sacrifice that is directly in conflict with the unanimous vote of its members. It will form a judgment that is in the starkest possible discontinuity with the corre sponding judgments of its members” (Pettit 2003: 183).
Pettit uses this example to illustrate why a summative account of group belief is untenable, but Lackey uses it to show the falsity of the informant flagging function when group belief is understood summatively. In Judgment Aggregation we can assume it is true that the pay sacrifice is in the employees’ best interest and that all of the yes votes are wellgrounded, reliable produced, etc. In such a case, the group may be said to have knowledge of the relevant proposition even though not a single member believes that p. Thus, if group belief is understood summatively, the group knowledge ascription cannot be identifying a reliable believer (Lackey 2012: 249). The second type of case is when group knowledge is determined by the decision-making power of a group that establishes its “official position.” Consider the following example: Official Position-. The Board of Directors of the American Academy of Pediatrics convenes and decides that their official position is that there is not a causal connection between the MMR (Measles, Mumps, Rubella) vaccine and autism. There is substantial, though not conclusive, evidence supporting this verdict, and the Board agrees that the benefits of vaccines strongly outweigh the possibility that a very small percentage of chil dren will become autistic as a result. Despite this, all of the doctors who are members of the American Academy of Pediatrics recognize that the
PLURALISM ABOUT KNOWLEDGE
|
127
evidence is inconclusive and so has some doubts that prevent them from believing that there is not a causal connection between the MMR vaccine and autism. (Lackey 2012: 249-250)
Assume there in fact is no causal connection between the MMR vaccine and autism, and that the substantial evidence upon which the doctors’ conclusion is based is sufficient for epistemic justification. Lackey says there is no reason to deny knowledge to the American Academy of Pediatrics (2012: 250). Yet not a single member of the group reliably believes that p, so the attribution of knowledge cannot be identifying a reliable believer in the summative sense. To escape this conclusion, one might appeal to a nonsummative (or in flationary) account of group belief. According to this view, group belief is irreducible to the beliefs of all or some of its members. As Lackey puts it, “in some very important sense the group itself is the subject of the belief, where this is understood as over and above, or otherwise distinct from, the beliefs of any of its individual members” (2012: 251). This account of group belief would enable us to ascribe knowledge to groups even when not a single member of the group has the belief in question. Lackey raises a number of objections to the joint acceptance account offered by Gilbert (1989), which is the most popular inflationary view of group belief. On this view, a group believes that p if and only if the members of G jointly accept p.25 Joint acceptance does not require a single member of the group to believe that p, but it does require one who participates in the joint acceptance of p to “accept an obligation to do what he can to bring it about that any joint endeavors ... among the members of G be conducted on the assumption thatp is true” (Gilbert 1989: 306-307). This does not require one to accept an obligation to believe or to try to be lieve that p, but it does require one who believes that not-p to not express that belief baldly. Thus, a group that jointly accepts p may be said to be lieve p even if not a single member holds this belief. Inflationary accounts of group belief have several problems (see Lackey 2012: 251-254). First, Gilbert’s position cannot clearly handle cases like Official Position because it is not clear that each member of the AAP en gaged in joint acceptance; moreover, we can easily imagine many indi vidual members baldly expressing their dissenting beliefs to their own patients.26 Second, the case of Judgment Aggregation involves nothing like
25 Schmitt (1994) and Tuomela (2004) also hold joint acceptance views. 26 Similarly, the Supreme Court may issue a group decision while individual judges nevertheless issue a dissenting opinion.
128
|
What’s the Point of Knowledge?
joint acceptance. Third, even if we use the premise-based aggregation pro cedure to motivate an inflationary view of Judgment Aggregation, as Pettit does, it is not clear which of the many judgment aggregation procedures we should use (see Cariani 2011). Finally, there are many concerns for in flationary accounts in general. For example, group belief seems far more directly voluntary than individual belief (Wray 2003; Hakli 2007) and group belief is far less governed by evidence than individual doxastic states (Wray 2003; McMahon 2003). Lackey concludes that the informant-flagging function of knowledge is unsupported by either a summative or a nonsummative account of group belief. She does not claim that knowledge ascriptions never flag reHable informants. Rather, her claim is that we do not attribute knowledge prima rily to identify reliable informants. Is there any hope left for the informant-flagging hypothesis? I will sug gest three ways to defend it from Lackey’s objections. First, one might deny that reHable informants must satisfy BELIEF. Second, one might argue that BELIEF should be restricted only to individuals. Third, one might argue that group knowledge attributions are acceptable but nonliteral and therefore strictly speaking false. I will briefly defend each of these ideas. 5.8.1 Option 1: Deny BELIEF Lackey’s argument against the informant-flagging function crucially relies on the claim that a “reHable informant” must satisfy BELIEF. If we accept this condition, then it follows that the purpose of knowledge attributions is to identify informants who at least either believe that p when p is the case, or beheve that not-p when not-p is the case. But should we accept this requirement? According to Craig, to whom Lackey attributes this requirement, “the matter is not so simple” (1990: 13). On the one hand, Craig claims that reHable informants will generally tell the truth because they beheve it, and the fact that someone beheves p is often a reliable and relatively accessible “indicator property” that the behever possesses good evidence (see §2.3). Thus, inquirers will tend to seek informants who satisfy BELIEF. On the other hand, there may be circumstance in which other factors about our informant will determine whether we accept what he tells us and these may override our tendency to believe even a very confidently made claim (Craig 1990: 13). In other words, befief is not an indicator property we are always willing to use, even though it correlates well with telling the truth. As a result, Craig says we may find some cases of knowledge in which the
PLURALISM ABOUT KNOWLEDGE
|
129
subject does not believe the relevant proposition.27 We may therefore con cede some grounds to people who want to play down the belief require ment (because it may not be necessary), yet belief need not drop out of the picture altogether (as it would on traditional analyses). A major advantage of function-first epistemology, as mentioned in §1.2, is that we can explain why belief is a central component of knowledge even if it isn’t absolutely necessary. I think Lackey should be amenable to this idea. Indeed, she argues there are independent reasons for doubting that belief in p is a necessary condition for being a reliable informant (2012: 255). Although Lackey takes this to show that knowledge attributions do not primarily identify reliable informants, we might draw a different conclusion: knowledge attributions do flag reHable informants, but reliable informants needn’t al ways satisfy BELIEF. Why might Lackey accept this alternative conclu sion? She should be open to denying BELIEF because her own account of group knowledge does not go in for necessary and/or sufficient conditions (2012: 267); moreover, she claims we properly attribute knowledge to groups even when there are no befievers. 5.8.2 Option 2: Restrict BELIEF to Only Individuals
You might not want to eliminate the befief condition for knowledge. Consider Lackey’s example of the creationist schoolteacher (introduced in §2.3), which indicates that knowledge may require befief. In this example, we are asked to imagine a creationist schoolteacher who learns the theory of evolution in order to teach it to her students. As a result of her highly competent teaching, her students come to know various truths about ev olution; however, the schoolteacher does not herself seem to know these various things because she does not believe them. Let’s assume that individuals must believe thatp in order to know that p. Does it follow that groups must also believe what they know? Lackey says it would be arbitrary to restrict BELIEF to only individuals (2012: 254).28 However, she notes there are several important differences between indi vidual beliefs and group beliefs; for example, group belief seems far more
27 See Radford (1966) and Myers-Schulz and Schwitzgebel (2013). 28 There is an interesting tension in Lackey’s work. On the one hand, her example of the creationist schoolteacher indicates that belief is required for knowledge. On the other hand, she argues there are cases of group knowledge in which there is no believer. The only way to resolve this tension (that I can think of) is to say belief is a necessary condition for individual belief but not group belief.
130
|
What’s the Point of Knowledge?
directly voluntary than individual belief. As Official Position illustrates, groups have the ability to simply decide at will to endorse a specific view or proposition, whereas individual belief is far less voluntary. Lackey also highlights research which suggests that group belief is far less governed by evidence than individual doxastic states. These facts might lead us to conclude that groups as such do not actually believe things, but rather they participate in what L. J. Cohen (1995) calls “acceptance.” On Cohen’s view, “belief that p is a disposition . . . normally to feel it true that p and false that not-p” (1995: 4). He goes on to say that belief “is classifiable as a disposition to have a certain kind mental feeling, not as a disposition to perform a certain kind of action” (1995: 11). In contrast, “To accept that p, is to have or adopt a policy of deeming, positing, or postulating that p—i.e., of including that proposition or rule among one’s premises for de ciding what to do or think in a particular context, whether or not one feels it to be true that p” (Cohen 1995: 4). We cannot deny there are several differences between individual be lief and what we might sometimes call group “belief.” Indeed, many philosophers of mind have claimed that groups lack the conscious agency that is necessary for genuine mentality (e.g., Block 1978). Further, Gilbert and Pilchman (2014) note that individual beliefs tend to be involuntary, aimed at truth, shaped by evidence, subject to the ideal of integration, context-independent, and come in degrees; in contrast, group “beliefs” are voluntary, aimed at utility, shaped by pragmatic goals, not subject to the ideal of integration, context-dependent, and do not come in degrees. On the basis of these differences, we may reasonably deny that groups liter ally have beliefs and instead maintain they can only accept propositions.29 As Cohen says, “When we look closely enough, and get behind the met aphor or the accidents of vocabulary, we find that organizations are typi cally engaged in accepting premises or pursuing goals, not in experiencing beliefs or desires” (1995: 55).
5.8.3 Option 3: Deny Group Knowledge
It might also be argued that group knowledge attributions are merely met aphorical, “loose talk,” or are otherwise not fully legitimate, and hence the very idea that group knowledge attributions identify reHable informants should be rejected from the start. Lackey considers this response
29 Another option, explored by Gilbert and Pilchman (2014), is to argue that Cohen provides a perfectly good account of individual belief but not group belief.
PLURALISM ABOUT KNOWLEDGE
|
131
(2012: 254), but she says it is utterly arbitrary to hold this position in the absence of any substantial argumentation. Further, she points to the fact that we do attribute knowledge to groups in “a widespread and systematic way” as evidence for the existence of group knowledge. But the fact that people are willing to attribute knowledge to groups tells us little about whether these groups actually have knowledge. There is abundant evidence from cognitive psychology that people tend to overattribute mental states. In one of the most familiar experiments, participants tended to describe the systematic movements of various ge ometric shapes in intentional terms (Heider and Simmel 1944). It is un likely, however, that the participants in this study really believed the circles and triangles were agents with genuine beliefs and desires. Rather, the data suggest that people find it easy to adopt the perspective from which behav ior could be described in intentional terms. In fact, people tend to ascribe mental states wherever they find purposive behavior (Huebner 2016). We should therefore not be surprised to find people attributing knowledge to groups of individuals that are committed to a common purpose. Our ten dency to attribute knowledge to groups may simply reflect our tendency to describe purposive behavior by using mental state language.30 None of this denies the propriety of group knowledge attributions.31 It may be perfectly acceptable to say, for instance, “The FBI knows the bomb was homemade.” But the fact that sentences in which knowledge is attributed to groups sound acceptable does not discriminate between interpretations on which these sentences are regarded as literally true and interpretations of these sentences on which they are regarded as merely metaphorical. To explore this issue, we might ask individuals to categorize a fist of sentences (including sentences that ascribe knowledge to groups and individuals) as making figurative or literal claims. In a relevant study, Adam Arico and his colleagues (2011) found that people tended to inter pret sentences about group mentality literally. Thus, we might be tempted to conclude that people will also interpret sentences about group knowl edge literally. But, as Bryce Huebner (2016) points out, this will not settle the issue because participants may interpret these sentences literally but treat them as claims about the individual members of the relevant groups.
30 Also, there are reasons to think most appeals to collective mentality are explanatorily superfluous (Rupert 2005; Huebner 2013). 31 At one point Lackey seems to assume that a deflationist about group knowledge ascriptions must deny the propriety of group knowledge ascriptions (2012: 266).
132
|
What’s the Point of Knowledge?
For example, the sentence “The FBI knows the bomb was homemade” might be interpreted as a literal claim about the mental states of indi vidual FBI agents. People might use group-terms as instrumentally useful shorthands to refer to group members. This wouldn’t explain Lackey’s examples of group knowledge in which there is no believer: ascriptions of group knowledge cannot be lit eral claims about the mental states of individual members if no individual believes the relevant proposition, let alone knows it. However, Lackey’s cases of group knowledge without belief only threaten my account if two conditions hold: first, people in fact are willing to ascribe knowledge in these cases; second, people interpret these knowledge claims as literally true. We may question both these assumptions. Although Lackey claims we often attribute knowledge to groups where not a single member of the group has the belief in question, she gives no evidence for this claim. Rather, she provides us with two cases that she asserts are paradigmatic examples of group knowledge. But are they? For what it’s worth, I do not find it intuitive to attribute knowledge to these groups, and I wonder whether others will agree. This is of course an em pirical question that requires some experimental work. In the absence of any evidence, however, we have little reason to think people will attribute knowledge to groups in which no member even believes the relevant prop osition. And even if some people do find it appropriate to attribute knowl edge in these cases, we have no evidence to suggest they are interpreting these claims literally. Finally, I should point out that my argument does not require us to deny the existence of group knowledge tout court. We may simply deny group knowledge when no member of the group believes the relevant propo sition. Thus, my view is perfectly compatible with the idea that group knowledge is a genuine phenomenon. What would be required, however, is that some or all members of the group believe p in order for the group to count as knowing that p.
5.9 Epistemic Injustice My story about the concept of knowledge is deeply social, but some might find it insufficiently political. I have not said much, if anything, about the effects of social categorization upon one’s status as a knower. Drawing on Miranda Flicker’s (2007) work on “epistemic injustice,” one might worry that stereotypes and biases lead individuals to systematically underestimate
PLURALISM ABOUT KNOWLEDGE
|
133
how much people in certain social groups know, and overestimate how much people in other social groups know. When a speaker receives less credit than she deserves, she suffers a credibility deficit. When a speaker receives more credit than she deserves, she receives a credibility excess. I will focus specifically on cases in which a speaker suffers a credibility deficit due to prejudice, since these are among the clearest cases of epi stemic injustice.32 In these cases, the concept ofknowledge seems to serve the function of systematically downgrading the epistemic status of some people, thereby oppressing them. Race and gender are clear cases in which the testimony of others might be dismissed as dubious. Consider Kant’s infamous remark about a “Negro carpenter”: And it might be, that there was something in this which perhaps deserved to be considered; but in short, this fellow was quite black from head to foot, a clear proof that what he said was stupid. (1960: 113)
These sorts of examples abound. As Cocker (1998: 317) notes, during slavery black people were generally denied the right to testify against white people because they were not regarded as credible witnesses. We can also easily call to mind cases in which sexist stereotypes cause a hearer to assign a deflated level of credibility to an informant’s word. On these grounds, one might conclude that our concept of knowledge often plays a social role that is morally problematic. We need to distinguish two ways in which the concept of knowledge might serve to perpetuate injustice. First, it could be that we fail to re gard a sufficiently reliable informant as a knower due to inappropriate and harmful prejudices (whether explicit or implicit). In these cases, we employ the concept of knowledge in an unjust way: the subject has met the conditions required for knowledge, but they are not recognized as a knower.331 will call this an epistemically unjust withholding ofknowledge. Second, it could be that the concept of knowledge itself embodies and perpetuates injustice. In this case, the concept embedded in our practice of epistemic evaluation would itself be sexist, racist, or otherwise morally problematic. For example, it might be that the methods we tend to count as knowledge-producing are more often associated with men than women.34 32 Jose Medina (2011) investigates credibility excesses. 33 There are also cases in which an individual is recognized as a sufficiently reliable informant, but is intentionally denied the status of a knower as a result of an identity prejudice. 34 See Haslanger (1999) for a discussion of this idea.
134
|
What’s the Point of Knowledge?
I will call this an epistemically unjust concept of knowledge. How do these two forms of epistemic injustice relate to my account of knowledge? Assuming the informant-flagging function is basically right, we would be employing the concept of knowledge incorrectly if an identity preju dice led us to mistakenly deny knowledge to someone who in fact meets the relevant epistemic standard. In this case, withholding an attribution of knowledge would run against the purpose of the concept of knowledge be cause we would fail to count as knowers people who in fact are sufficiently reHable for the community at large. Our practice of epistemic evaluation works correctly when an informant meets the level of justification required for knowledge and the inquirer recognizes that the informant meets this standard. But when an inquirer does not befieve what she should, or when she mistakenly regards her informant as insufficiently refiable when she shouldn’t, then our practice of epistemic evaluation fails to fulfill its pur pose. A potential informant receives less credit than she deserves, and the result is an epistemic injustice. In contrast, if the concept of knowledge itself embodies and perpetuates injustice, then it seems we could use this concept correctly even when it facifitates the oppression of certain social groups. In this case, the morally and politically problematic conditions of a knowledge denial would be enshrined in the application conditions of the concept itself. Assuming our concepts and our social practices are as deeply intertwined as I have been treating them, it would not be totally surprising to discover that our epi stemic concepts, norms, and practices are shaped by and reflect the racist and patriarchal circumstances from which they emerge. Following Helen Longino (1994:4—7) and Sally Haslanger (1999: 469), we might distinguish between “constitutive” and “contextual” values of our epistemic practices. The constitutive values of a practice are those that constitute the goal or end of the practice (or are a necessary means towards those ends), while the contextual values are “those present in the context where knowledge is sought, discovered, attributed, denied, forgotten, etc., but that do not define what the practice is, or what makes it an epistemic practice” (Haslanger 1999: 469). Another way to draw this distinction is between the content of the concept of knowledge on the one hand, and the problematic conditions of a knowledge denial (or attribution) on the other hand. If the concept of knowledge were itself epistemically unjust, then pre sumably our goal should be to consider what our concept of knowledge ought to be fike in order to achieve epistemic justice (among our other goals). It is not a part of my project here to make the case that our concept
PLURALISM ABOUT KNOWLEDGE
|
135
of knowledge actually does embody this form of injustice. Rather, I have treated the conditions for knowledge as neutral on matters of gender, race, and other aspects of our social identity. This is because I am focusing on human needs and interests so basic and universal that they can be modeled in a manner that abstracts away from considerations of social identity. But let’s suppose the concept of knowledge does embody some form of injustice. If this were correct, then the method of function-first epis temology would be the perfect way to help us determine what sorts of concepts, norms, and practices we should want to organize our epistemic lives. I encourage others to take up this project. In lieu of completing this task myself, I’ll simply point out that my methodology is amenable to this aim. Indeed, my book is predicated on the idea that epistemologists should be interested not just in what knowledge is but also what it ought to be like—and the functions it ought to serve. Finally, my account is also com patible with the highly plausible idea that we often attribute (and deny) knowledge in ways that perpetuate injustice. As I have argued, this would frustrate the purpose for which we have a practice of epistemic evaluation.
5.10 Concluding Remarks This chapter has explored a variety of proposals about why humans think and speak of knowledge. In particular, I have investigated the claim that knowledge attributions are used to signal the appropriate end of inquiry, to track the epistemic norm governing assertion and practical reasoning, to distinguish between blameworthy and blameless behavior, to provide as surance, to describe one’s epistemic position, and to pick out group knowl edge. I have not denied that knowledge claims may be used for each of these purposes. Rather, I have argued that this plurality does not threaten my claim that the primary function of the concept of knowledge is to iden tify reliable informants. This does not imply that every application of our knowledge concept is concerned with the informant-flagging function. By asserting “S knows that p,” I might aim to provide you with assurance or encouragement without flagging S as a reliable informant as to whether p. I may also deny that S knows p even though S is a reliable informant because, for in stance, attributing knowledge to S would have the unwanted implicature that she deserves praise for her actions. But none of this suggests that our concept of knowledge has no primary function or that different functions are in conflict. To return to my earlier analogy, a hammer has a primary
136
|
What’s the Point of Knowledge?
function but it can also serve multiple nonconflicting purposes. Indeed, the fact that hammers can function as paperweights and weapons is something we can explain by appealing to the “point” of hammers—viz., they have a certain shape and weight in order to achieve their primary purpose. In the case of knowledge, I have appealed to our common need to identify reliable informants in order to explain why attributions of knowledge may also function as inquiry-stoppers, assurance-givers, and threshold-markers that indicate when the epistemic standard that is usually necessary and sufficient for assertion and practical reasoning has been met. Put more generally, I have argued that the concept of knowledge is valuable because it helps us to collaborate with others to achieve certain goals. The fact that knowledge attributions play a plurality of roles only further indicates the importance of knowledge in our folk epistemology. For the sake of brevity, I’ve had to limit my discussion in two ways. First, I have focused on knowledge attributions to adult humans (and their groups), not young children or nonhuman animals. It seems intuitive, how ever, that young children and animals know a great many things—yet nei ther of these creatures are obviously reliable informants in the sense under consideration. So you might ask: Do I think these are genuine cases of knowledge? I do not wish to deny the propriety of attributing knowledge to either young children or animals, since we can explain and understand much of their behavior in terms of knowledge. In the case of animals, we might even classify many of them as reliable informants. For example, vervet monkeys have different warning cries for aerial predators (e.g., hawks), terrestrial predators (e.g., lions), and climbing predators (e.g., snakes). When one monkey sees a predator and gives the appropriate cry, other members of the troop come to its aid.35 So they, too, have something of a need for “reliable informants.” Still, I think adult human knowledge is often more extensive and sophisticated than animal knowledge and in fant knowledge in various ways.36 Adult human knowledge is the kind of knowledge that has been the main object of epistemological interest, but a complete theory of knowledge must provide an account of animal and infant knowledge. I leave this for another occasion. Second, it is plausible that the plurality of roles played by knowledge ascriptions is far greater than those I have considered here. For example,
35 Thanks to Catherine Elgin for this example. 36 See Sosa (2007) and Williams (2015) for a defense of this idea. For a contrary view, see Kornblith (2002).
PLURALISM ABOUT KNOWLEDGE
|
137
you might think that knowledge ascriptions are used to give credit for true belief (Greco 2003), to counter doubts (Rysiew 2001), or to indicate that one is certain (BonJour 2010). I have done what I can to address the major alternatives to my view, but a fully general account of knowledge would have to confront this vast range of complex phenomena. In the next chapter, we will examine another type of epistemic pluralism, namely the view that the cognitive norms that determine what counts as knowledge vary with and are dependent on local cultural frameworks.
138
|
What’s the Point of Knowledge?
chapter 6 | Epistemic Diversity
There seems to be no known language in which sentences using ‘know’ do not find a comfortable and colloquial equivalent. The implication is that it answers to some very general needs of human life and thought. . . —EDWARD CRAIG (1990: 2)
a deep presupposition guiding this book is that we can derive important theoretical insights by investigating the English word “know” and the con cept it expresses. But English is just one of over 6,000 spoken languages, and native English speakers make up less than 6 percent of the world’s population. Further, some evidence suggests there are substantial and sys tematic differences in the epistemic concepts, judgments, and norms that different people use to evaluate human cognition. This raises a consid erable worry. Am I simply describing the epistemic terms, concepts, or norms of some specific cultural group? In short, am I engaging in ethnog raphy—or what some scholars call “ethno-epistemology”?1 There is nothing wrong with ethnography per se; however, epistemologists customarily take questions about the nature of knowledge, justification, and so forth to be questions about “universal principles and concepts unmarked by any particular demographic or geographic limita tions” (Kukla 2015: 202). For example, skepticism is supposed to be wor rying and interesting because it purports to show that there is no human knowledge, or almost none. Like most epistemologists, I take myself to be investigating general epistemic facts, not culturally local ones. This
1 Weinberg, Nichols, and Stich (2001) use this term.
raises an important methodological question about the nature and aims of function-first epistemology. Why should we be interested in studying the concept expressed by the English word “know”? What is the philosoph ical significance of a concept that we acquired as a result of the seemingly arbitrary fact that we happened to learn one language and acquire certain concepts rather others? In this chapter, I will argue that the concept of knowledge is universal in the sense that it is expressed by some word in every (or virtually every) nat ural language. There is some empirical data suggesting the concept of knowl edge is not universal, but I will argue this data is not robust. I then construct a theoretical argument for why language users living in social communities would develop (or adopt) a concept of knowledge that is shared crossculturally.2 My argument, in a nutshell, is that all humans have certain basic practical needs that the concept of knowledge satisfies, and we could not meet these needs with an epistemic lexicon that does not contain any words with a content that is (near-) synonymous with “know.” Thus, we should ex pect every culture to have a central epistemic concept that is similar to that expressed by the English word “know.” I conclude this chapter by exploring the impheations of this thesis for philosophical inquiry.
6.1 Evidence for Epistemic Diversity Are there substantial and systematic cross-cultural differences in the epi stemic concepts, judgments, and norms that people employ when evaluating human cognition? According to a group of experimental philosophers, the cognitive norms that determine what counts as knowledge vary with and are dependent on local cultural frameworks. Before discussing the evi dence for this sort of epistemic diversity, we must distinguish two senses in which the concept of knowledge might be culturally relative. First, there might be systematic variation in the extension of the word “know” as used by English speakers from different cultural backgrounds. For example, people of Western descent might refrain from attributing knowledge to Gettier victims, whereas people of East Asian descent might be willing to ascribe knowledge in such cases. This may suggest these groups have different concepts.3 Second, the concept of knowledge might be culturally
21 will sometimes speak of “adopting” concepts, but my argument does not depend on anything like Platonism. We may also speak of “constructing” concepts. 3 One might prefer to speak of different conceptions of knowledge, but I am concerned with the standards for what it takes to know, not just conceptions of what it takes to know. Following
140
|
What’s the Point of Knowledge?
relative because the word best translated as “know” in another language expresses a different concept from that expressed by the English word “know.” How plausible are these possibilities of conceptual variation? In a widely cited study, Jonathan Weinberg, Shaun Nichols, and Stephen Stich (2001) report systematic variation in the extension of the word “know” as used by English speakers from different cultural backgrounds.4 They test a number of important cases in the history of contemporary epis temology, including the following Gettier case: Bob has a friend Jill, who has driven a Buick for many years. Bob there fore thinks that Jill drives an American car. He is not aware, however, that her Buick has recently been stolen, and he is also not aware that Jill has replaced it with a Pontiac, which is a different kind of American car. Does Bob really know that Jill drives an American car, or does he only believe it?
Weinberg, Nichols, and Stich (WNS) report that the majority of East Asian (56 percent) and South Asian (61 percent) participants responded with the judgment that Bob has knowledge (“really knows”), whereas the majority of Western participants (74 percent) maintained that Bob lacks knowl edge (“only believes”). Thus, Westerners seem more likely to deny knowl edge to the victim of this Gettier scenario than their Eastern counterparts (Weinberg, Nichols, and Stich 2001: 444). WNS also report that participants of East Asian heritage were less likely than Western participants to ascribe knowledge to the protagonist of Keith Lehrer’s “Truetemp” case (Lehrer 1990: 163-164). They test the following example: One day Charles is suddenly knocked out by a falling rock, and his brain becomes re-wired so that he is always absolutely right whenever he estimates the temperature where he is. Charles is completely unaware that his brain has been altered in this way. A few weeks later, this brain re-wiring leads him to believe that it is 71 degrees in his room. Apart from his estimation, he has no other reasons to think that it is 71 degrees. In fact, it is at that time
authors like Jackson (2011), Nagel, Juan, and Mar (2013), and Machery et al. (2015), I will speak in terms of concepts rather than conceptions. 4 They also report that epistemic judgments differ based on socioeconomic status and educational background. I focus specifically on cultural differences, but most of what I say applies equally to these other factors. Some people have claimed there are gender differences between epistemic judgments (Buckwaiter and Stich 2011), but a large number of follow-up studies report no gender differences.
EPISTEMIC DIVERSITY
|
141
71 degrees in his room. Does Charles really know that it was 71 degrees in the room, or does he only believe it?
While both East Asian and Western subjects were more likely to classify Charles’ belief as not a case of knowledge, WNS found that East Asian subjects were significantly more likely to deny knowledge (88 percent) than Western subjects (68 percent). WNS also tested a version of Fred Dretske’s (1970) Zebra case: Mike is a young man visiting the zoo with his son, and when they come to the zebra cage, Mike points to the animal and says, “that’s a zebra.” Mike is right—it is a zebra. However, as the older people in his community know, there are lots of ways that people can be tricked into believing things that aren’t true. Indeed, the older people in the community know that it’s pos sible that zoo authorities could cleverly disguise mules to look just like zebras, and people viewing the animals would not be able to tell the differ ence. If the animal that Mike called a zebra had really been such a cleverly painted mule, Mike still would have thought that it was a zebra. Does Mike really know that the animal is a zebra, or does he only believe that it is?
WNS found that only 32 percent of Westerners maintained that Mike re ally knows that the animal is a zebra, whereas 50 percent of participants from the Indian subcontinent attributed knowledge to Mike. Thus, it seems that epistemic judgments may vary as a function of differences in cultural background. If we assume these epistemic judgments (or “intuitions”) are a basic source of evidence about our concepts, then this data indicates that the concept of knowledge is culturally relative. The idea behind this thought, put roughly, is that words express concepts and that concepts are the meanings of words. It is my concept of knowledge that enables me to say, for instance, that a Gettier victim does not know, and this reveals something about the shape of that concept.5 The cultural variation reported by WNS challenges the idea that the concept of knowledge is universal. The upshot is that epistemologists who study the concept expressed by the English word “know” are only investigating a concept that is possessed by a particular cultural group (e.g., Caucasian English speakers), not a concept that we all have de spite our different heritages. This poses a prima facie challenge to a lot
5 See Jackson (2011) for a defense of this idea.
142
|
What’s the Point of Knowledge?
of epistemological theorizing. Why should we care about the features of a concept that only belongs to Westerners? Does this implicitly assume that Western epistemology is the benchmark by which all other cultures’ epis temological activities are to be understood and measured? Isn’t it arbitrary to privilege the epistemic judgments of one cultural group over another? While the results reported by WNS have been heavily cited, it is doubtful they are robust. In their extensive review of the literature, Kenneth Boyd and Jennifer Nagel (2014) report “there is no robust evidence that the epi stemic intuitions of different demographic groups are deeply at odds with each other.” For example, Nagel, San Juan, and Mar (2013) were unable to replicate WNS’s Gettier case results, finding no statistically significant difference in knowledge ascriptions across ethnicity, gender, age, or phil osophical training. Minsun Kim and Yuan Yuan (n.d.) also failed to detect a cross-cultural difference in Gettier intuitions by using the same methods and case taken verbatim from WNS. Hamid Seyedsayamdost (2015) tried to thoroughly replicate all the results presented by WNS, but he found no ethnically or socioeconomically correlated differences. John Turn (2013) also reports a lack of difference in evaluations of Gettier cases in tests involving Western participants and participants from South Asia. Nagel (2013: 182) makes three additional observations. First, only two of the eleven reported comparisons by WNS found that the majority of one group went one way (i.e., “only believes”) and a majority of the other group went another way (i.e., “really knows”). In other words, actual group disa greement was reported only for two of the eleven comparisons discussed. Second, WNS only found a statistically significant difference in a single Gettier case on a single multicultural pool of American undergraduates, whereas they found no such difference between Westerners and East Asians in two additional experiments designed to elicit Gettier intuitions. Third, the cases they investigated were not tested repeatedly and they did not test other Gettier cases with different content. As a result of these critiques, Stich himself now questions the veracity of the data presented by WNS. During his talk at the Cambridge Moral Sciences Club on 14 May 2013, he called the 2001 study “crude” and admitted he’s not sure he trusts it. In a more recent and comprehensive study, Edouard Machery and his team (including Stich) report the first cross-linguistic evidence that Gettier intuitions are found across languages. They surveyed four different cul tural groups (Brazil, India, Japan, and the United States) that speak quite different languages and found that cases of justified true beliefs were not judged to be cases of knowledge. On these grounds, Machery and his team hypothesize there might be “a core folk epistemology.” They write,
EPISTEMIC DIVERSITY
|
143
If this hypothesis is correct, then people in all cultures will possess epi stemic concepts that require more than justification, truth and belief, and in most cultures that concept will be expressed by the epistemic term com monly translated into English as “know.” (Machery et al. 2015: 452)
As Machery et al. point out, this does not entail that people in all cultures have the same knowledge concept. Even if every culture has a central epi stemic concept that requires more than justification, truth, and belief, there could still be considerable variation in other ways. For example, it might be that Westerners are less willing than East Asians to attribute knowledge in conversational contexts where unrealized possibilities of error have been mentioned. This is just one of many possible ways in which folk episte mology may vary as a function of demographic variables. Thus, the fact that we do not detect cross-cultural or cross-linguistic differences for some cases, such as Gettier cases, does not imply that individuals from different backgrounds will not have diverging judgments about some other situations. Still, there might be robust agreement across a sufficiently wide range of cases to postulate the existence of a universal knowledge concept. How one comes down on this issue will largely depend on one’s account of concept individuation. Whether people who assign slightly different extensions to the word “know” have different concepts is a vexing topic because “there is no consensus, either among philosophers or among linguists, about what constitutes a difference in meaning, or about how concepts are to be individuated” (Stich 2013: 5). According to Jackson (2011), if two people disagree about whether the victim of a Gettier scenario qualifies as knowing, then they simply have different concepts of knowledge and mean different things by “know.” The assumption here is that someone presented with a particular case will judge one way rather than another way because they she possesses one concept rather than another. But you might also think that concepts are individuated in a less fine-grained way. Perhaps two people have the same concept if the conditions that govern its application are the same for both people across a sufficiently wide range of cases. On this view, the universality of the concept expressed by the word “know” is compatible with considerable cultural variation. I prefer this way of individuating concepts because it is the only way to make sense of the idea that two people can ever share the same concept. The hypothesis I want to explore in this chapter is that there is a cultur ally invariant core to the concept of knowledge. According to this view, there is a sufficiently large core set of epistemic judgments that do not vary cross-culturally. If this view is correct, then we have a good reason to
144
|
What’s the Point of Knowledge?
regard the concept of knowledge as universal—at least at its core. What’s more, this conclusion would have significant implications for philosoph ical methodology. If cross-cultural judgments about whether an agent knows were to overlap sufficiently, then epistemologists could still derive significant theoretical payoffs by investigating the concept expressed by the word “know.” Differences in the extension of the word “know” that exist at the periphery (rather than the core) of our concept should not bar epistemologists from studying the deeper structural features of the concept of knowledge.61 will elaborate on this idea in the next section.
6.2 The Universal Core of Knowledge There is already evidence to support the hypothesis that there is a universal core to the concept of knowledge. For example, the study conducted by WNS found that for one crucial probe there was no statistically significant difference among any of the groups they examined. They tested the fol lowing case, Coin Flip: Dave likes to play a game with flipping a coin. He sometimes gets a “special feeling” that the next flip will come out heads. When he gets this “special feeling,” he is right about half the time, and wrong about half the time. Just before the next flip, Dave gets that “special feeling,” and the feeling leads him to believe that the coin will land heads. He flips the coin, and it does land heads. Did Dave really know that the coin was going to land heads, or did he only believe it?
Among all participants there was widespread consensus that Dave only believed (did not know) that the coin was going to land heads. Since almost everybody agrees that we should not treat cases of subjective certainty as knowledge, then at least some of our epistemic judgments are universal. Indeed, experimentalists will frequently use control condition cases in which there is near universal agreement that the protagonist does not know (i.e., cases involving subjects who make lucky guesses or who hold obvi ously false beliefs).7 This provides further evidence for a universal core
6 Nagel (2013) has a similar view, but she defends this idea in a very different way. 7 Boyd and Nagel (2014) make a similar point. They also claim the empirical literature on crosscultural variation in mind reading—which is the capacity to interpret, predict, and explain the behavior of others in terms of their underlying mental states—does not support the claim that the basic structural features of knowledge differ between cultures.
EPISTEMIC DIVERSITY
|
145
to the concept of knowledge. Weinberg, Nichols, and Stich are open to the possibility of such a universal core. They write: “the fact that subjects from all the groups we studied agreed in not classifying beliefs based on ‘special feelings’ as knowledge suggests that there may be a universal core to ‘folk epistemology’ ” (2001: 450). Admittedly, the empirical data in favor of a universal knowledge concept is far from conclusive. To confirm this hypothesis, we would have to show that the epistemic judgments of the vast majority of people converge over a wide range of cases. This would require testing thousands of cases across dozens of languages and cultures. In lieu of completing this daunting task, I will mention some supplementary em pirical data from linguistics that indicates the existence of a core folk epistemology. Specifically, the program of research known as natural semantic metalanguage suggests there is a small inventory of universal concepts that are expressed by some word in every human language; further, it seems the concept expressed by the word “know” is universal in this sense (Goddard 2010). More precisely, the sense of “know” that embeds a propositional or wh- complement is arguably universal.8 Martin Haspelmath and Uri Tadmor (2009) also report that a word for “know”—in the sense that embeds a propositional complement— features in all the languages in the World Loanword Database, which includes a broad sampling of languages from every inhabited conti nent. Thus, linguists have isolated “know” as one of a small number of words that are universal. What, then, should we say about the empirical data for epistemic diver sity? As I’ve argued, the existing data is not very robust. But let us suppose for the sake of argument that there is a degree of cultural variability. Even if this data were robust, it is important to realize just how narrow this evi dence really would be. Experimental epistemologists have largely focused on Gettier cases, Truetemp cases, and the effects of skeptical scenarios on knowledge judgments.9 They have investigated these cases because sev eral major movements in epistemology over the past 50 years have relied on the judgments elicited by these thought experiments; however, these
8 Evidence suggests that the situation is more complicated vis-a-vis translations of “know how” (e.g., Jim knows how to ride a bike). While it might be true that propositional and question embedding verbs are not different across languages, this is clearly not true for all uses of “know.” My focus is on propositional knowledge and I remain neutral on whether knowledge-wh reduces to propositional knowledge. 9 For a sample of views, see Nichols, Stich, and Weinberg (2003), Swain, Alexander, and Weinberg (2008), Nagel, Juan, and Mar (2013), and Buckwaiter and Stich (2014).
146
|
What’s the Point of Knowledge?
studies do nothing to show that cross-cultural divergence in epistemic judgments is a widespread phenomenon. This is because Gettier cases, Truetemp cases, and the like compose only a tiny portion of the full range of cases involving knowledge attributions. These findings are thus per fectly compatible with convergence across a vast range of cases in which people epistemically evaluate others. Furthermore, the cases investigated by experimentalists are a fairly un representative lot. People do not commonly need to determine whether the unlucky victim of a Gettier scenario knows (see §3.8), or whether someone who has miraculously had his brain reliably rewired is someone on whom we should rely, or whether we are really brains in vats (as I’ll discuss in chapter 8). Gettier scenarios, Truetemp cases, and skeptical possibilities are not the kinds of situations we routinely face in our day-today affairs. Further, our ordinary capacity to think about epistemological issues has likely been shaped by circumstances that are far less esoteric and far-fetched. It is for this reason that I would not be very surprised if our judgments about knowledge diverged in these cases, since the hypothetical circumstances that bring them out are unusual. Our practice of epistemic evaluation almost certainly did not develop with an aim to regulate these sorts of situations or scenarios. Joshua Alexander and Jonathan Weinberg seem to agree: they conjecture that differences in folk epistemologies will be “more likely in the unusual and marginal sorts of cases that are pop ular with epistemologists than in ones that are evaluated and discussed in some frequency in civilian life” (2014: 136). Even Hugo Mercier (2011), a staunch critic of those who believe there are deep group differences in cog nition, thinks we should expect to find fine-grained cultural differences in knowledge ascriptions. But this takes place against a widely shared back ground of epistemic convergence. Joshua Alexander, Jonathan Weinberg, and Stacey Swain have else where resisted this line of reasoning. They claim that as increasing amounts of empirical data display culturally based epistemic diversity, it becomes “less reasonable to suppose that the intuitions experimental philosophers attack are an unrepresentative lot” (Swain, Alexander, and Weinberg 2008: 148). As I’ve argued, however, the empirical case for cultural variation has concentrated on a narrow and unrepresentative set of examples. Admittedly, these cases have been central to much epistemological theory, so if epistemologists have been wrong about the intuitiveness of these cases, then this finding would be disastrous for many contemporary theories of knowl edge (I’ll return to this point in §6.5). But this would not at all undermine
EPISTEMIC DIVERSITY
|
147
the claim that there is a universal core to our folk epistemology—and per haps a very large one. Rather, it is likely that some cases will strike us all as clear examples of knowledge (e.g., beliefs gained through direct percep tual observation under ideal conditions), other cases will strike us as clear examples of ignorance (e.g., a false beHef), and for still other cases there may be no widely accepted answer. Assuming there is cross-cultural agree ment about paradigm cases of knowing and paradigm cases of ignorance, we may rightly say there is a universal core to our central epistemic con cept. In other cultures this concept will likely be expressed by the epistemic term commonly translated into English as “know.”
6.3 An Explanation for a Universal Knowledge Concept If “know” were universal in this way, it would be a remarkable fact that cries out for an explanation. Why is it that no human language seems to get by without a word that accurately translates “know”? It is by reflecting on the purpose of epistemic evaluation that we come to understand why language users like us, Hving in social communities, would have a cross-cultural concept of knowledge. The basic idea is that all humans have certain fundamental practical needs that the concept of knowledge satisfies, and we could not meet these needs with an epistemic lexicon that does not contain any words with a content that is (near-) syn onymous with “know.” We should therefore expect every sufficiently developed culture to have a central epistemic concept that is similar to that expressed by the Engfish word “know.” It is of course possible that a near-synonymous word, such as “schnow,” could do the job of “know” by playing the same functional roles.1011 It is unfikely, however, that we could meet our present communicative needs with an epistemic lexicon that does not contain any words with a content that is at least nearly synonymous with “know.” What basic human needs does the concept of knowledge fulfill? As argued earlier, humans in every culture throughout history have required true beliefs in order to successfully guide their actions and to satisfy their natural curiosity. It is by having enough truths (and not too many falsehoods) that we are able to reliably satisfy our other fundamental needs, such as our need for food, water, shelter, and countless other things.11
10 Thanks to Mikkel Gerken for pointing this out. 11 False beliefs sometimes have survival value, as mentioned in chapter 2.
148
|
What’s the Point of Knowledge?
Consequently, it is in our interest to find sources of information that will lead to true beliefs. In particular, we tend to seek out informants because they are often convenient and helpful (see §2.5). With respect to almost any issue, however, people vary in terms of their reliability. Thus, we may expect any community of individuals to have an interest in evaluating the reliability of potential informants. Further, we come to assess the ade quacy of informants for people and purposes beyond our own immediate concerns because we seek to utilize the practical advantages of pooling and sharing information. The concept of knowledge is hypothesized to arise in connection with these goals. This illustrates why the concept of knowledge is deeply important to humans. All societies will need to solve the problem of how to identify, pool, and share reliable information with a community of inquirers, so we should expect every sufficiently developed society to possess a concept that helps us achieve this goal.12 This idea contrasts with the usual emphasis that cross-cultural linguis tics places on the contingency of our concepts. Anna Wierzbicka (1999) and Cliff Goddard (2010) argue that the language of mental state concepts displays a great deal of cultural variation, which reminds us of the contin gency of our concept possession and how easily things might have been different. The concept of knowledge, however, is not highly contingent in this way. Given some basic facts about humans, our conditions of life, and some of our interests and aims, we have needs that the concept of knowledge satisfies. My claim is not just that the concept of knowledge is beneficial in certain ways, but that this concept must be this way (or sufficiently similar) in order to satisfy these needs. We use this concept to meet certain fundamental human interests, although our practice of epistemic evaluation was not necessarily designed with these benefits ex plicitly in view. This does not imply that the concept ofknowledge arises out of anything like metaphysical necessity.13 We need to distinguish two possibilities. First, it might be that a concept which arose to meet certain needs must, necessarily, have a certain shape in order to meet those needs. Second, it might be necessary for us to have a specific concept that meets these needs. I am defending the former claim but not the latter. The shape of the concept ofknowledge is necessitated by its arising to meet certain needs;
12 At some point in human history societies would not have been far enough along in their development to address these needs effectively. 131 take this point from Kappel’s (2010) brilliant analysis of Craig’s work.
EPISTEMIC DIVERSITY
|
149
moreover, the basic nature of these needs makes it incredibly likely (al though not necessary) that we would have and use this concept. On this hypothesis, a species with a radically different psychology from ours might not have anything like the concept of knowledge. As Neta (2006: 277 n. 25) observes, a community of omniscient beings would have no need for the concept of knowledge because there would be no need for them to distinguish reHable informants from unreliable informants. Evaluative concepts arise in connection with certain needs and practices that they are meant to regulate, so concepts that do not serve any function are unlikely to be adopted. (This function could be as simple as identifying a natural kind.) This reasoning will presumably apply to any word or con cept in our linguistic and conceptual repertoires.14 Humans are social an imals and thus societies will adopt words and concepts that are useful to them. The more basic and universal the human needs are that the con cept we are explicating satisfies, the more essential this concept will be to our life. A consequence of this idea is that the concept of knowledge is species relative. But this does not entail that knowledge is something only humans possess. Once we have this concept, we may apply it to creatures that are both less and more intelligent than us, and who have radically different needs and interests from our own. For example, we may accurately say that an omniscient being is “all knowing,” even though the concept of knowledge would likely not serve any purpose in a community of omnis cient beings. If the concept of knowledge has a universal core, what does this core look like? More empirical research is needed to fully answer this ques tion, but I believe the pragmatic theory outlined in this book provides an initial answer: the core of knowledge is sufficiently reHable true befief, where “sufficiently reliable” is unpacked roughly as “reliable enough to serve the purposes of members in the epistemic community.” As argued in chapter 3, an agent with a sufficiently reHable true befief is in a strong enough epistemic position with respect to some proposition p to eliminate all of the not-/? possibilities that are relevant to members of the commu nity that might draw on her true befief. The relevant alternatives are those that are widely (but tacitly) agreed to be the ones that an informant must rule out.
14 Similarly, recall Hume’s insight that a community of perfectly altruistic beings in a land of abundant resources would have no need for a concept of justice (Treatise 3.3.2).
150
|
What’s the Point of Knowledge?
6.4 Conceptual Pluralism and Ameliorative Epistemology I have provided a theoretical argument for the existence of a universal folk epistemology. However, this argument is open to empirical refuta tion. Perhaps experimental philosophy or cross-linguistic semantics will eventually reveal that the concept of knowledge is not a cultural universal in the way I have envisaged, just as many other concepts are language and culture specific. Imagine we locate some culture where the notions of epistemic evaluation embedded in their language are radically different from those in the West. For example, imagine a society in which nothing remotely similar to the Western concept of knowledge is expressed by a word in their language. It is perfectly conceivable that in some society the word best translated as “know” is one that means “true belief’ or “highly justified belief.” Consequently, you might think it is arbitrary to study the Western concept of knowledge because it is possible that some society got along without it. As Nichols, Stich, and Weinberg write, Without some reason to think that what White, Western, high SES [soci oeconomic status] philosophers call ‘knowledge’ is any more valuable, desirable, or useful than any of the other commodities that other groups calls ‘knowledge’ it is hard to see why we should care if we can’t have it. (2003: 245)
But not all epistemic concepts are on par. The concept of knowledge is used to satisfy basic communicative and epistemic needs, which explains why we should expect this concept (or a sufficiently similar one) to arise in any human society. Insofar as this concept has arisen to satisfy these basic pragmatic needs, it is plausible that any society lacking this concept would be conceptually and epistemically impoverished. The concept of knowledge is deeply important to human life, so any society that lacked this concept might find it difficult, perhaps impossible, to effectively sat isfy the same basic needs that the concept of knowledge satisfies. Indeed, the very possibility of a well-developed human society getting by without this concept (or something sufficiently similar) might be dubious if my account is correct. Epistemologists can therefore go beyond the descriptive project, which reveals the present shape of our epistemic concepts, and engage in a nor mative project, which evaluates how well or poorly people’s epistemic practices satisfy their needs and goals. Following Sally Haslanger (2012), we can think of this as ameliorative epistemology. An ameliorative project
EPISTEMIC DIVERSITY
|
151
has two broad phases. We start by asking, “What is the point of having the concept in question?” For example, why do we have a concept of knowl edge or a concept of understanding? This requires us to thoroughly and critically reflect on the purpose (or purposes) of using a specific concept. Subsequently, we ask: “What ought our concept be like in order to best satisfy our needs and goals?” At this stage we may reformulate a concept in view of the purposes we have established. We introduce a new theo retical concept by stipulating the new meaning of the term, and its con tent is determined by the practical or theoretical role it plays (Haslanger 2012: 367). The aim is to enhance our conceptual resources (or the concep tual resources of others) in a way that better serves our critically examined purposes. In this way, putative facts about the roles of the concept of knowledge can be put to work for normative theorizing; in particular, we legitimately critique the epistemic standards or concepts of a society that fail to satisfy basic human needs and goals. This would not be arbitrary, xenophobic, or unjustified. There are clear pragmatic grounds on which we can con fidently judge our practices of epistemic evaluation as superior to a so ciety that lacked a (near-) synonym for “know.” Any such society would be unable to effectively meet its fundamental communicative and epistemic goals and would therefore be disadvantaged as a result. Moreover, this is an area of philosophical inquiry where experimental philosophy has no bearing. The results of recent experimental studies only reveal how we in fact categorize cases of knowledge, not how it would be best to categorize them—i.e., how we ought to categorize them for our purposes or benefit. None of this assumes that our actual concept of knowledge is the best one to have, all things considered. There may be room for considerable improvement. My claim is simply that our knowledge concept is good to have because it does important work and solves coordination problems centering on the need to identify, pool, and share information.
6.5 Problems and Prospects for a Universal Epistemology Suppose our epistemic judgments about Gettier victims, Truetemp cases, and skepticism are in fact culturally variable. This would indicate that our core folk epistemology does not include a number of epistemic judgments that have played a central role in philosophy. Would this be a problem for epistemologists who want to rely on these judgments when theorizing? It certainly would be. To illustrate this point, consider the Gettier problem.
152
|
What’s the Point of Knowledge?
If there were genuine cultural variability about whether the victim of a Gettier scenario knows, then epistemologists who seek to solve the Gettier problem would only be attempting to solve a problem that exists for some people (i.e., Westerners) and not others (i.e., East Asians). The Gettier problem would simply not arise for those cultures or dialect groups that do not have a knowledge concept that requires more than justification, truth, and belief. But the fact that some limited range of epistemic judgments are shown to be culturally variable would not threaten my epistemological project. I aim to show that folk epistemology has a universal core, but I do not claim that we can expect near universal agreement about the extension of “knows” in all cases. It is implausible that any theory of knowledge will accommodate every imaginable case as nicely as one might wish, so we should be attentive to the possibility epistemic pluralism. Thus, I am open to the possibility that our knowledge concept might be culturally elaborated in particular ways; for instance, it might be that Chinese participants are more likely than American participants to ascribe knowledge to Charles in the Truetemp case (I have no data on this). Precisely why people from different cultural backgrounds might assign different extensions to the concept of knowledge is not something I will attempt to answer here. My proposal leaves room for gaps that might be filled in different ways, in different cultures, and at different times. In this way, we may supplement Craig’s imaginary genealogy with real genealogy, by engaging with the reality of historical and cultural contingencies. Nevertheless, we may still speak of a universal core folk epistemology that might receive different forms of cultural elaboration. This raises an important question about philosophical methodology, namely, which are the philosophically interesting cases? You might think, as Jennifer Cole Wright puts it, that “philosophy is often most interesting (and of most value) when it is working ‘at the margins,’ wrestling with unclear and borderline cases” (2010: 500). This sort of project certainly has value. But another goal of epistemology is, or should be, to elucidate the deeper core of our epistemic practices and our epistemically evaluative notions. What characterizes this core? Presumably it is those features of the concept that enable it to play an essential role in satisfying the basic human needs that epistemic evaluation serves to regulate. In contrast, the more peripheral aspects of our folk epistemology will include features that are not necessary to satisfy these needs. On my account, it is likely that these peripheral features will permit more cultural variation because they are not responsive to fundamental and universal human interests. In this
EPISTEMIC DIVERSITY
|
153
way, my approach allows us to separate those aspects of folk epistemology that emerge from the purpose of its origin, on the one hand, from aspects that are a contingent and peripheral matter of its history, on the other hand. An important question remains unanswered. Do we have a reliable method by which to track epistemic convergence? Put differently, can we predict when our epistemic judgments are not merely the expression of a local cultural framework? Assuming our epistemic judgments are some times culturally variable, we cannot simply assume that our own judgments always reflect or express a universal folk epistemology.15 It is clear that experimental epistemology must play a crucial role in this project. As we continue to empirically study epistemic evaluation, the contours of our folk epistemology will become clearer. However, we might also wonder whether there is an a priori (or armchair) way to tell which epistemic judgments are expressions of a core folk epistemology. I’d like to conclude by cautiously suggesting that there might be a method available; in particular, people may have introspective access to certain at titudinal states that can serve as a reliable indicator of our core epistemic judgments. Let me explain. Swain, Alexander, and Weinberg (2008) found that attributions of knowledge are sometimes vulnerable to an “order effect.” In particular, they found that participants’ judgments about the Truetemp case were significantly influenced by what case the participant had previously considered. For example, if the participant had previously read Coin Flip, in which Dave forms the true belief on the basis of a “special feeling,” then the participant was more likely to say that Charles knew (as opposed to merely believed) that the temperature was 71 degrees. If, however, the par ticipant had previously read Testimony, a case in which a woman named Karen forms a true belief about chemistry on the basis of reading an article in a top scientific journal, then the participant was less likely to say that Charles knew.16 This reveals an order effect on our epistemic judgments. But not all epistemic judgments were vulnerable to an order effect. In particular, there was no order effect on participants’ judgments about Coin Flip or Testimony. Regardless of which case they read previously, participants were equally likely to deny knowledge in Coin Flip. Likewise, participants were equally likely to ascribe knowledge in Testimony ir respective of which case immediately preceded it. Thus, it seems that
15 Versions of this objection have been raised by Weinberg (2007), Swain, Alexander, and Weinberg (2008), and Alexander and Weinberg (2014). 16 The same pattern was true for the fake-barns case, but I set that aside for simplicity.
154
|
What’s the Point of Knowledge?
only certain epistemic judgments are vulnerable to this irrational bias. Is there a way to tell which of our epistemic judgments are stable and which are not? Recent work by Wright (2010, 2013) indicates that people tend to be much more confident in, and believe more strongly, epistemic judgments that turn out to be stable both within an individual and shared across populations. In line with Swain, Alexander, and Weinberg, Wright found that participants’ knowledge attributions were unstable in Truetemp as a result of which case they previously considered. Also in line with Swain, Alexander, and Weinberg, she found that participants’ knowledge attributions were stable in Coin Flip and Testimony. What was novel about her study is that she also asked participants to rate how confident they were about their answer (0 = not very confident to 5 = very confident). She found that participants were significantly more confident in their judgments about Coin Flip and Testimony than they were in their judgments about Truetemp; moreover, this was true regardless of the order in which the cases were presented. Her results therefore indicate that one’s confidence is a significant predictor of whether the case in question was stable or un stable. In other words, participants were able to introspectively track the stability of their intuitive judgments. How does this relate to epistemic diversity? Recent work by Asher Koriat (2012) indicates that a general feature of intuitive judgments is that the more confident one is about one’s intuitive judgments, the more likely it is that these judgments will be stable—both within an individual and shared across populations. Thus, it might be possible for us to sub jectively track judgments that vary as a function of differences in cultural background. This is an empirical hypothesis that must be tested, of course. Further, there are some important differences between intuitive judgments that are affected by biases and intuitive judgments that vary as a func tion of differences in cultural background. Specifically, the former involve judgments that are “unduly influenced by information present/salient at the moment our judgments are formed,” while the latter involve “a much more complex story about the ways in which socio-cultural belief systems/norms become internalized, shaping our understanding of the use of certain concept—and perhaps even the concepts themselves” (Wright 2016: 569). Still, it might be that high degrees of confidence track not just stability from order effects but also cross-cultural stability. If this were true, then high degrees of confidence would track cross-cultural agreement about whether a subject knows, while low degrees of confidence would track cases where cross-cultural disagreement is more likely. This ability
EPISTEMIC DIVERSITY
|
155
would be unsurprising on my account, since the paradigm cases of knowl edge would be those at the core of our folk epistemology.
6.6 Concluding Remarks By reflecting on why humans think and talk about knowledge, we are provided with independent evidence contrary to the view that the concept of knowledge is culturally variable. This shields much of epistemology from the recently popular objection that we cannot derive important the oretical insights by investigating the English word “know” or the con cept it expresses. If the concept expressed by the English word “know” is universal, then epistemologists can investigate a concept that we all share despite our diverse cultural and linguistic backgrounds. In this way, my proposal vindicates a common methodology in epistemology. Moreover, we need not deny that there is (or could be) cross-cultural variation in epistemic judgments. Variation in some areas is compatible with uniformity in others. This leaves space to explore systematic vari ation in the extension of the word “know” as used by English speakers from different cultural backgrounds, as well as the ways in which the word best translated as “know” in other languages has a different exten sion from the English word.
156
|
What’s the Point of Knowledge?
chapter 7 | Epistemic Pragmatism
Pragmatists think that if something makes no difference to practice, it should make no difference to philosophy. —RICHARD R0RTY (1995: 281)
is the correct semantic theory of “knows”? One of the most prev alent debates in contemporary epistemology is about whether the seman tics of knowledge ascriptions is invariant or context sensitive.1 A variety of moves have been made to try to make headway on this issue, but no clear victor has emerged. For example, philosophers have appealed to our intuitive judgments about cases, the plausibility of psychological biases that people routinely exhibit, “semantic blindness,” and certain tests of the genuine context-sensitivity of a term.1 2 All these factors (and more) have been carefully investigated in an effort to determine whether contextualism, invariantism, or some other view is true of our knowledge ascriptions. The jury is hung. In a further attempt to adjudicate this dispute, a handful of philosophers have recently appealed to putative facts about the purpose of knowl edge ascriptions. John Greco (2008, 2009), David Henderson (2009, 2011), Robin McKenna (2013), and a former version of myself (Hannon 2013, 2015b) argue that considerations about the function of knowledge what
1 For a sample of views, see Brown (2006), Cohen (1999), DeRose (2009), Fantl and McGrath (2009), Hawthorne (2004), Lewis (1996), Rysiew (2001), and Stanley (2005). 2 See DeRose (1992), Brown (2005), Stanley (2005), Schaffer and Knobe (2012), and Pynn (2017) for a discussion of the intuitive basis of contextualism; see Nagel (2010a) and Gerken (2013) for a discussion of psychological biases; see Schiffer (1996), Hawthorne (2004), DeRose (2009), and Montminy (2009) for discussions of semantic blindness; and see Cappelen and Lepore (2003), Partee (2004), and Stanley (2004) for tests of context sensitivity.
ascriptions support epistemic contextualism. In contrast, Stephen Grimm (2015b) claims the social role of knowledge ascriptions supports sen sitive invariantism, while Christoph Kelp (2011) thinks these same considerations favor insensitive invariantism. Will this strategy work? Can we derive the correct semantic theory of “knows” from putative facts about the function of knowledge ascriptions? This chapter raises doubts about the viability of this strategy. In the next two sections, I examine how this functionalist account has been used to support both contextualism (§7.1) and sensitive invariantism (§7.2). I argue that subject-sensitive invariantism is implausible if the informant-flagging function of knowledge ascriptions is correct. Further, I claim that Grimm’s non-subject-centered version of sensitive invariantism collapses into a ver sion of contextualism. Subsequently, I argue that insensitive invariantists can also motivate their view by appealing to considerations about the func tion of knowledge ascriptions (§7.3). What does all this show? Contrary to what many epistemologists claim, it seems that putative facts about the function of knowledge ascriptions will not instruct us as to whether con textualism, sensitive invariantism, or insensitive invariantism is true. This might seem like an unhappy result, but I try to draw a positive lesson from it: our practice of epistemic evaluation functions perfectly well regardless of whether contextualism or some type of invariantism is true (§7.4). In the second half of this chapter (§§7.5-7.7), I make a more radical proposal: I suggest this entire debate about the semantics of “knows” mistakenly presupposes that we should account for the meaning of epistemic claims by determining their truth conditions. A more natural approach, I argue, is to ask what practical function epistemic evaluation serves us in communicating with each other. I call this view epistemic pragmatism. According to this view, natural language leaves indeterminate whether knowledge claims have contextualist or invariantist semantics.
7.1 Contextualism and the Purpose of Knowledge Following Craig, I have been arguing that humans think and speak of knowing primarily to identify good informants to members of their commu nity. Although each of us needs reliable information to successfully guide our own actions, we also have a shared need to pool and transfer informa tion to make it easily accessible. We therefore must identify individuals who are reHable enough not just for our own immediate purposes, but also for a variety of individuals with a wide range of interests, projects,
158
|
What’s the Point of Knowledge?
and purposes. In saying that someone knows, we are certifying an agent as epistemically positioned such that others can freely draw on his or her information. Assuming this hypothesis is correct, what—if anything—can it tell us about the semantics of knowledge ascriptions? In a series of papers, Greco (2007, 2008, 2009) appeals to this hy pothesis to motivate contextualism about “knows.” One of the central components of Greco’s view is that knowledge attributions are “used to flag good information and good sources of information for use in practical reasoning” (2008: 433). He arrives at this view by combining Craig’s idea that knowledge attributions function to identify good informants with the increasingly popular idea that knowledge is central to practical reasoning (as discussed in §5.3). In epistemology, “attributor contextualism” refers to the following view: The truth-value of sentences of the form “S knows that p” (and the like) varies with the context of the speaker of the sentence. That is, for the very same S and p, at the very same time, a sentence of the form “S knows that p” can be true relative to one speaker context and false relative to a different speaker context. (Greco 2008: 417)
Greco agrees that knowledge attributions are context sensitive, but he says the attributor’s context is sensitive to a variety of practical interests. On Greco’s view, the truth-value of a knowledge ascription is determined by context, but that might be the practical reasoning environment of the attributor, the subject, or someone else (Greco 2008: 433). Why does Greco allow for this type of sensitivity? Because if the function of knowl edge ascriptions is to serve us in our practical reasoning, the standards for knowledge should be tied to whatever interests and purposes are relevant to the practical reasoner at issue. Thus, the interests and purposes that are operative in the relevant practical reasoning context will set the epistemic standard to know.3 Although the interests of the attributor, the subject, or some third party can affect the standards for knowledge, Greco says there are still restrictions on how a person’s interests may affect the operative standard. On the one hand, the standards for knowledge cannot drop too low because the demands of practical reasoning require that knowledge be imported
3 Following Greco, I will not put quotation marks around “know,” “knowledge,” and “knowing” to indicate semantic ascent. My discussion will often be stated in the object language for ease of expression.
EPISTEMIC PRAGMATISM
|
159
across various practical environments. For S to be a useful informant, the standards by which she counts as a knower must be at least as high as practical reasoning typically requires. On the other hand, the standards required for knowledge cannot be too high because that would make it less than widely available, which would prevent knowledge from playing its role in practical reasoning. Practical reasoning is widely necessary, so knowledge must be widely available.4 According to Greco, “this creates considerable pressure towards stability regarding the truth-conditions of our knowledge claims” (2008: 429). Even if the interest-dependent truth conditions for knowledge ascriptions vary across practical environments, they will not vary wildly. This must be true if knowledge ascriptions are to satisfy the information-sharing function in our practical and social activities.5 Although the truth conditions of knowledge ascriptions are fairly stable on Greco’s view, his account is still contextualist because the truth-value of knowledge attributions is variable over attributor contexts. He therefore defends a version of interest-dependent contextualism (Greco 2008: 433). Henderson (2009, 2011) also motivates contextualism by appealing to the idea that knowledge attributions identify good informants. According to Henderson, to say “S knows that p” is to indicate that S is epistemically positioned with respect to p so as to be a good source of “actionable in formation,” which means that one can take it from S that p (Henderson 2009: 120). This motivates contextualism when combined with a second idea: whether an agent qualifies as a sufficiently good source of action able information depends on who is expected to draw on that information. Since knowledge attributions are used to certify agents as reliable sources of information to some audience or group, whether it is true to say that an agent knows will depend on the audience for which that agent is evaluated as a source of information. In particular, Henderson distinguishes “applied source communities,” which are focused on some particular project, from “general source communities,” which are devoted to developing a body of results on which people in various other communities might confidently draw, whatever their projects or purposes might be (2009: 126). When one
4 A skeptical invariantist would disagree. She would argue that knowledge ascriptions can be appropriate even though the epistemic standard isn’t met as long as the subject is close enough to meeting the relevant standard (see Davis 2007). I will discuss this view in the next chapter. 5 Hawthorne (2004), Williamson (2005), and Rysiew (2012) argue that contextualism is incompatible with the role knowledge reports play in storing, retrieving, and transmitting useful information. I defend contextualism from this objection in my article “Stabilizing Knowledge” (2015b).
160
|
What’s the Point of Knowledge?
is certifying an informant for an applied source community, one should be concerned with sources of information in view of the practical interests of the applied community; but when one is certifying an informant for a general source community, one should be concerned with sources of information that will be reliable for an indeterminate range of applied communities and their various practical interests. Greco and Henderson endorse contextualism because of what they take to be a plausible hypothesis about the point of the concept ofknowledge: it is for certifying people as good sources of information to a community of inquirers. Further, Greco and Henderson posit versions of contextu alism with fairly stable truth conditions across a wide range of contexts, but they also allow the epistemic standards to shift in order for knowl edge ascriptions to continue serving their function of identifying good informants. Let’s call this view functional role contextualism. Functional role contextualists argue that whether an informant qualifies as a reliable source of actionable information is a context-dependent matter; thus, whether knowledge is truly ascribed to an informant will depend on the context. Functional role contextualism is prima facie plausible. Indeed, I defended this view for years. However, contextualists will score a victory only if the functionalist story to which they appeal favors their view over rival semantic accounts. I now believe that contextualists fail to achieve this goal, and I will explain why in the sections to follow.
7.2 Sensitive Invariantism and the Purpose of Knowledge John Hawthorne (2004) and Jason Stanley (2005) believe that facts about practical rationality must be incorporated into the truth conditions of knowledge ascriptions. They defend interest-dependent versions of invariantism according to which the truth-value of a knowledge claim depends on the interests and purposes of the subject—the person to whom knowledge is attributed. Stanley therefore dubs the view “subject-sensitive invariantism.”6 The main thesis of subject-sensitive invariantism is that whether a subject knows a proposition at a given time is partly determined by his practical interests, i.e., by how much is at stake for that person at that
6 Unlike contextualism, which is exclusively a semantic view, sensitive invariantism combines semantic and metaphysical theses. Here it is the semantic issues that are under discussion; however, to the extent that the relevant metaphysical claims are motivated by linguistic data, linguistic considerations will indirectly bear on the metaphysical issue.
EPISTEMIC PRAGMATISM
|
161
time. This is still a version of invariantism, however, because a knowledge attribution’s truth-value is invariant across different contexts of utterance (unlike contextualism). Thus, while both functional role contextualists and subject-sensitive invariantists highlight a connection between knowl edge and practical reason, subject-sensitive invariantists emphasize that a knower must meet the epistemic standard commensurate with the subject’s own need for actionable information. If knowledge ascriptions are used to certify good informants to an ep istemic community, then subject-sensitive invariantism faces a significant problem. The subject-sensitive invariantist claims the epistemic standards for knowledge are tied to the practical reasoning situation of the subject. As argued in §4.7, however, it makes little sense to think that the epistemic standards would be so closely tied to the interests of the subject rather than a more general standard. If knowledge attributions are for certifying good informants to a community of people, then the epistemic standards would be strongly conditioned by the general needs and interests of the community to which the subject belongs. Thus, our hypothesis about the purpose of knowl edge ascriptions gives us a reason to reject subject-sensitive invariantism.7 One advantage of contextualism is it allows the attributor’s context to be sensitive to the needs and interests operative in the subject’s context as well as those of the attributor and her epistemic community. For ex ample, the attributor might be trying to help the subject decide how to proceed on an issue that is of the utmost importance to the subject but not to the attributor. In such a context, the attributor will be affected by the subject’s interests, which call for more demanding epistemic standards. Contextualism can thus provide a type of sensitivity that is attuned to the need to identify good informants to a community of inquirers. What determines the relevant epistemic standard in a given context is a complex function of the attributor’s context, the practical interests of the subject whose belief is epistemically assessed, and the epistemic community to which the attributor and the subject in question belong. The social role of the concept of knowledge thus puts considerable pres sure on subject-sensitive invariantism. However, Grimm (2015b) purports to defend a non-subject-centered version of sensitive invariantism that I will call community-sensitive invariantism. On this view, the epistemic standards for knowing are sensitive to the concerns not just of the subject
7 Subject-sensitive invariantism also has difficulty explaining why speakers in high-stakes contexts will use third-person knowledge attributions to deny knowledge to subjects in faraway, low-stakes contexts (DeRose 2009).
162
|
What’s the Point of Knowledge?
but to certain third parties as well. More specifically, our judgments about knowledge are sensitive to the highest practical stakes in the relevant practical reasoning situation. As argued in chapter 3, this is because the standards must be sensitive to (high enough for) other people who might appeal to the subject’s belief, where “might” tracks the notion of a “real possibility” that “might actually” happen. For example, although it is log ically possible that someone with life-or-death stakes might need to know whether the Red Sox won last night, we would not regard this as a pos sibility that could reasonably be expected to occur. Thus, my belief that the Red Sox won needn’t be based on justification high enough to meet life-or-death stakes (unless perhaps I actually encounter such an inquirer). I would, however, need enough justification to satisfy people who are in terested in the game’s outcome for more ordinary reasons. Grimm motivates his view by appealing to the familiar hypothesis that knowledge attributions certify good informants to a community of potential inquirers. As information-dependent and information-sharing creatures, we are accustomed to the fact that others often turn to us for information about various topics; however, it is hard to say in ad vance who will come to rely on our judgments. Grimm thus argues that the threshold relevant for knowledge will gravitate toward a level high enough to respect the “typical” or “normal” stakes of people who might appeal to these judgments. This is the “default” threshold for knowl edge, but this standard can be overridden when the stakes are especially high (as I argued in chapter 4). In this way, knowledge is tied to prac tical interests and is therefore stakes-sensitive, yet what it takes to know does not change radically from case to case. Knowledge is fairly stable across contexts because this allows knowledge ascriptions to serve their information-sharing function. Are there important differences between Grimm’s view and functional role contextualism? Grimm prefers to describe his view “not as a version of contextualism but rather as a version of stakes-sensitive invariantism” (2015b: n. 22). But the motivation for this claim is unclear. What makes sensitive invariantism an invariant view is that the truth-value of a knowl edge claim is invariant across different contexts of utterance (unlike con textualism). What makes it sensitive is that the epistemic standards of knowledge are partly determined by (i.e., sensitive to) the subject’s prac tical reasoning environment. According to Grimm, however, the standards of knowledge are sensitive to the concerns not just of the subject but to certain third parties as well. This seems to permit contextual variability. So in what sense is Grimm’s view a version of invariantism!
EPISTEMIC PRAGMATISM
|
163
We can interpret his view in two ways. Perhaps there is one community that sets the threshold for knowledge, which would make for an invariant standard. To make this idea more plausible, we could appeal to my earlier claim (in §4.6) that all humans are, in an important sense, members of the same epistemic community. However, an invariantist could not permit the communal standard for knowledge to be overridden when the stakes of some individual are especially high (the standard must be, fixed by the community). Yet Grimm acknowledges that, in some concrete settings, the standards for knowing may rise if the situationally relevant stakes are higher than normal. So this interpretation of his view seems incorrect. Further, it would give up the sensitive aspect of sensitive invariantism. We may interpret Grimm’s “community-sensitive invariantism” another way. Perhaps there are multiple communities and the epistemic standards are invariant with respect to the community at issue (i.e., in the context). On this view, the community plays a similar role to that of the subject on classical versions of subject-sensitive invariantism: just as the practical reasoning situation of one person may differ from that of another person, so too may the standards differ across communities. But then there doesn’t seem to be much difference between this view and the versions of contex tualism defended by Greco and Henderson. If one is keeping epistemic gate for various contextual communities, and if the standards are fitting to the higher ranges plausibly encountered in the relevant community, and if this varies across communities, then why is this not just contextualism? Thus, Grimm’s view likely collapses into a version of contextualism that is not substantially different from the versions of functional role con textualism we have thus far considered. The informant-flagging function therefore provides little reason to prefer one view to the other. Let me draw out this comparison a little more clearly. According to both Grimm and the contextualists, the truth of a knowledge ascription is sensitive not just to some individual’s context but also to the demands of other people in their community, which includes a variety of individuals with diverse prac tical interests. This type of sensitivity is needed if knowledge ascriptions are to serve their hypothesized function of identifying good sources of informa tion to a community of inquirers. Further, this sort of sensitivity allows both Grimm and the contextualist to deal with potential counterexamples to their view. For example, if the attributor’s context can be sensitive to the practical reasoning demands of the subject (or other inquirers), then contextualists can accommodate cases in which it is intuitive to say that a subject in a highstakes context knows even though the attributor has little stake in the matter. Similarly, if the epistemic standards for a subject to know can be sensitive
164
|
What’s the Point of Knowledge?
to the practical interests of inquirers outside the subject’s own practical rea soning environment, then Grimm can explain why speakers in high-stakes contexts will use third-person knowledge attributions to deny knowledge to subjects in faraway, low-stakes contexts. In addition, both Grimm and the contextualists claim the truth conditions for knowledge will be fairly stable across a wide range of contexts. Thus, we cannot appeal to considerations of stability to evaluate the plausibility of these views, which is a strategy used by Hawthorne (2004), Williamson (2005), and Rysiew (2012). What about insensitive invariantisml If the purpose of knowledge ascriptions were incompatible with this view, it would be a significant strike against what is widely regarded as the orthodox view in the his tory of epistemology. Indeed, several philosophers (myself included) have argued that insensitive invariantism is incompatible with the role knowl edge attributions purportedly play in identifying reliable informants. However, I will now argue that functional considerations do not actually tell against insensitive invariantism.
7.3 Insensitive Invariantism and the Purpose of Knowledge Insensitive invariantism is the main source of resistance to contextualism. According to this view, what counts as being in a sufficiently good epi stemic position to know does not vary with practical facts about the con text. Whatever the subject’s or the attributor’s (or someone else’s) practical interests might be, there is some good epistemic position in which an agent must stand with respect to a proposition in order for that agent to know it. If the standard for knowledge does not vary with practical factors in a context, then how do insensitive invariantists account for our willingness to ascribe knowledge in some cases (i.e., low-stakes) but not in others (i.e., high-stakes)? Following Grice (1989), many insensitive invariantists argue that what varies with context are not the truth conditions of knowledge (or “knowledge”) but rather the conditions of warranted assertability (see Rysiew 2001 and Brown 2006).8 They usually spell out the idea of warranted assertability in terms of Gricean rules of conversation, which allow speakers to be pragmatically warranted in asserting something false
8 Another type of insensitive invariantism appeals to various psychological factors that lead us to mistakenly attribute/deny knowledge (e.g., Bach 2005 and Nagel 2010b). I bracket this form of invariantism because I am concerned with recent attempts to draw conclusions about the semantics of knowledge attributions from facts about their purpose, yet nobody has defended this version of invariantism by making such a move.
EPISTEMIC PRAGMATISM
|
165
and pragmatically forbidden from saying something true. More specifi cally, insensitive invariantists often claim that asserting a true knowledge ascription in a context where much is at stake would have a false implicature and thereby convey something false to the audience; for instance, it would convey that the subject’s evidence is strong enough for current practical purposes. In contrast, falsely asserting “S doesn’t know that p” would have a true implicature and convey that S’s evidence is not strong enough for current practical purposes, given what is at stake. Here I focus on nonskeptical versions of invariantism according to which our knowledge attributions in low-stakes cases are true and ap propriate, whereas our knowledge denials in high-stakes cases are false but appropriate.91 will argue that insensitive invariantism is equally well supported by our hypothesis about the function of knowledge ascriptions. First, I will provide a functionalist explanation for why the semantics of knowledge ascriptions might settle at an invariant level; second, I will an swer a recent objection to insensitive invariantism that has been made by functional role contextualists. As interdependent information-sharing creatures, we have a signifi cant need for identifying reliable sources of information. If knowledge ascriptions are for certifying an informant’s befief to members of our ep istemic community, then we must identify informants who are reliable enough for a variety of individuals with a wide range of interests. In order to ensure that informants are reliable enough for many people, a prac tice would develop of setting the standard fairly high. According to Craig, this will edge us towards the idea of someone who is a good informant as to whether p whatever the particular circumstances of the inquirer, whatever rewards and penalties hang over him and whatever his attitude to them. That means someone with a very high degree of reliability, someone who is very likely to be right—for he must be acceptable even to a very demanding inquirer. (1990: 91)
This passage arguably supports a version of insensitive invariantism. On this view, the standard for knowledge would settle at a level high enough to satisfy the function of identifying good sources of information to a com munity of inquirers.
9 See Davis (2007) for a defense of skeptical pragmatic invariantism, and Dinges (2016) for a reply. I will discuss skepticism at length in the next chapter.
166
|
What’s the Point of Knowledge?
Why might the semantics of knowledge ascriptions become settled and thus invariant? If knowledge ascriptions are used to share information on which a variety of individuals may act, then the relevant epistemic standard must not be tied too closely to the concerns of particular individuals—that would allow too much variability for knowledge ascriptions to serve their information-sharing function. To share information on which others may rely, there must be a sufficiently high standard (as Craig says above). In other words, there must be a common point of view such that a variety of potential inquirers could rely on the epistemic assessments of others for their own practical and theoretical deliberations. A knowledge ascrip tion with the insensitive content posited by the invariantist seems well designed to communicate this sort of situation-insensitive message to the epistemic community. An invariant standard would easily allow knowl edge attributions to “serve as a piece of common coin, usable by and useful to any number of people or groups not part of the immediate sit uation” (Rysiew 2012: 286). It is therefore plausible that the semantics of knowledge ascriptions might settle at a level that is most useful to the inquiring public at large. I previously rejected this argument. In “The Practical Origins of Epistemic Contextualism” (2013), I argued that (/insensitive invariantism were correct, then the truth conditions of knowledge ascriptions would sometimes be at odds with their function. More specifically, I claimed that insensitive invariantists are committed to the view that an informant may truly count as a knower even though her epistemic position is not strong enough for the inquirer’s purposes. If this were true, it would give us a reason to prefer functional role contextualism to insensitive invariantism. Before explaining why I no longer hold this view, let me clarify this objection by using the “bank case” discussed previously. Imagine that Keith and Rachel stop at the bank to deposit their paychecks. They have an impending bill that must be paid and little money currently in their account, so it is very important that they de posit their pay checks by Saturday. Keith remarks that two weeks ago he was at the bank on Saturday morning and it was open, but Rachel points out that banks do change their hours. In such a case, it seems wrong to attribute knowledge to Keith. According to the (nonskeptical) insensi tive invariantist, Keith does meet the standard for knowledge, but it is inappropriate to attribute knowledge to him because it would have the false implicature that his evidence is strong enough for current practical purposes. Thus, insensitive invariantism allows for cases in which an
EPISTEMIC PRAGMATISM
|
167
informant in a high-stakes situation could truthfully, albeit improperly, be called a knower. So far so good. But if the point of knowledge attributions is to identify reliable informants, one might wonder why it would be true to say that Keith knows. It seems his evidence isn’t strong enough for current practical purposes, so why would he truly qualify as a knower? After all, knowledge ascriptions are used to identify reliable informants, yet it seems that Keith is not reli able enough. For this reason, I used to think insensitive invariantism was committed to the view that the truth conditions of knowledge ascriptions could be inconsistent with their putative function. (In contrast, contextu alism faces no such inconsistency: when the stakes are high, it would be false to say that Keith knows precisely because he is not reHable enough for current practical purposes.) I am no longer persuaded by this argument. The invariantist has a plau sible story for why it would be true to say that Keith knows even though he is not reliable enough for his audience in the context. The story is this: if knowledge ascriptions are used to certify informants to an epistemic community, then we should expect the semantics of “know” to settle at a level high enough to satisfy the community’s standard, even though an informant might not be in a strong enough epistemic position for people with uniquely pressing concerns. Any high-stakes context that calls for stricter standards would then be dealt with at the level of pragmatics. In this way, uttering a sentence with the insensitive content posited by the invariantist could still communicate a situation-sensitive message over and above what the ascription semantically encodes. In the high-stakes context, for example, Keith is reliable enough for the community, so he knows; however, it is inappropriate to say that he knows because it would falsely convey to Rachel that his information is good enough to meet the particularly pressing concerns in their context. Pragmatics accounts for the expression of situation-geared information, while traditional seman tics provides a stable, insensitive content that the social role ofknowledge ascriptions seems to demand.10
101 take this point from Rysiew (2012: 285), who led me to see the force of this view. Rysiew also advises caution about arguing directly from facts about the function of knowledge ascriptions to facts about their semantics, but our views differ in at least two ways. First, Rysiew focuses on the hypothesis that knowledge ascriptions play a “special role in signaling an appropriate end to particular lines of inquiry” (2012: 270). Second, he argues that the plurality of roles played by knowledge ascriptions makes it difficult to base claims about their semantics on some select function (2012: 277).
168
|
What’s the Point of Knowledge?
Thus, insensitive invariantism is also compatible with our hypothesized function about the role of knowledge ascriptions. Indeed, it seems just as attuned to our need to identify good informants as functional role contex tualism. Where does this leave us?
7.4 Diagnosing the Problem We have considered several attempts to use putative facts about the role of knowledge ascriptions as evidence for one’s favored semantic view. As we’ve seen, this strategy does not help us adjudicate between func tional role contextualism and invariantism. This is because our hypothesis about the role of knowledge ascriptions cannot tell us whether the se mantic content of “knows” has firmly settled at a level high enough to satisfy the community’s standard (in which case our need to communicate situation-sensitive information would be dealt with at the level of prag matics) or whether the content of “knows” is stable across a wide range of circumstances and yet changes when the stakes are high.11 While this might seem like an unhappy result, I think we can draw a positive lesson from it. Notice that whether you are a contextualist or some type of invariantist, competent speakers will typically use “knows” to identify good informants, thereby satisfying the purpose of the concept of knowledge. Suppose, for example, that Stewart (a contextualist) and Jessica (an insen sitive invariantist) are at the L.A. airport contemplating taking a certain flight to New York. They want to know whether the flight has a layover in Chicago. Both Stewart and Jessica overhear someone ask a passenger, Krista, if she knows whether the flight stops in Chicago. Krista looks at her flight itinerary and responds, “Yes, I know—it does stop in Chicago.” But Stewart and Jessica have a very important business contact they have to meet at the Chicago airport. Stewart says, “How reliable is that itinerary? It might contain a misprint. They could have changed the schedule at the last minute.”11 12 In this case, Stewart (the contextualist) and Jessica (the insensitive invariantist) will likely disagree about whether the sentence uttered by
11 It is possible that some other hypothesis about the function of knowledge ascriptions will shed more light on their semantics. In chapter 5, however, I argued that many of the possible functions of knowledge ascriptions can be given a fairly unified treatment. This provides us with some reason to doubt that any of these individual functions will help us resolve this issue. 121 adapt this case from Cohen (1999: 58).
EPISTEMIC PRAGMATISM
|
169
Krista is strictly speaking true. What matters for their purposes, however, is not whether Krista’s utterance is true (vs. merely warrantedly asserted); what matters, presumably, is whether it would be reasonable for them to act on her assertion. They want to know whether her information is reli able enough for action: should they board the plane now or check further with the airline agent? Similar considerations will apply to other examples in the literature on contextualism. Consider the high-stakes bank case. Contextualists and nonskeptical invariantists disagree about whether it is strictly speaking true to say that Keith knows the bank is open (the former say it is false; the latter say it is true). But a fact that is accepted by all parties in this debate is that competent speakers do regularly affirm instances of “S knows that p” in some contexts (i.e., those with low stakes) and that these same speakers will deny such attributions of knowledge in other contexts (i.e., those with high stakes).13 As DeRose observes, the issue dividing invariantists and contextualists is not whether in different conversational contexts, quite different standards govern whether ordinary speakers will say that someone knows something. Of course, what we’re happy to call knowledge in some (“low-standards”) contexts we’ll deny is knowledge in other (“high-standards”) contexts. The issue is whether these varying standards for when ordinary speakers will attribute knowledge, and/ or for when they’re in some sense warranted in attributing knowledge, re flect varying standards for when it is or would be true for them to attribute knowledge. (1999: 195-196)
In other words, these utterances strike people as natural and appropriate irrespective of whether the semantics of these attributions takes a contextualist or an invariantist form. Thus, we may reasonably suppose that if the invariantist and the contextualist found themselves together facing a situation of the sort described in the bank scenario (or the airport sce nario, etc.), they would be able to get along just fine with each other, as far as understanding what each other meant (and intended) by their words.14 What matters to Keith and Rachel is not whether Keith’s knowledge claim is strictly speaking true or false. Rather, what matters is whether it would 13 The phenomenon of shifty epistemic judgments has been doubted by May et al. (2010), Feltz and Zarpentine (2010), and Buckwaiter (2010); however, these criticisms have been largely addressed by Schaffer and Knobe (2012), Sripada and Stanley (2012), Pinillos (2012), and DeRose (2011). I presuppose this phenomenon is a real one. 14 Baz (2012: 325-326) makes this point.
170
|
What’s the Point of Knowledge?
be reasonable for them to act on the assertion that Keith knows the bank is open tomorrow. They want to determine whether they should get in line now or return the next morning. This illustrates that our practice of epistemic evaluation achieves its purpose whether this practice is accounted for in terms of what is pragmat ically conveyed or what is semantically expressed. This fact should give us pause. If knowledge ascriptions play their role whether or not a specific semantic view is true of them, then what practical difference does it make whether the semantics of knowledge ascriptions is contextual or invariant? Uttering “S knows thatp” is a way to communicate that S’s information is reliable enough according to some contextually salient epistemic standard irrespective of whether that standard reflects warranted assertability or lit eral truth. Either way, knowledge attributions are made in ways that make one’s conversational contribution relevantly informative. Thus, it seems that knowledge ascriptions can do the work that we need or might reasonably expect them to do, at least in the sorts of situations that contextualists and invariantists focus on, without our ever needing to at tend to the question of whether their semantics should be understood along contextualist or invariantist fines (Baz 2012). So one lesson we might draw from this is to shift our focus away from the truth conditions of knowl edge ascriptions and instead focus on the function of human discourse— by asking what difference it makes to us to have this practice of epistemic evaluation. Perhaps a better way to understand our knowledge claims is to better understand their role in the games we currently play. I take this to be an important lesson from J. L. Austin’s epistemological writings. A striking feature of Austin’s work is that he consistently avoids the words “true” and “false” when investigating the use of knowledge attributions in various contexts. He instead prefers to use expressions like “it will not do to say,” it would be “right” to say (1979: 98), we would be “justified” in saying (1979: 101), or it would be “silly” or “within reason” to say certain things (1979: 84). He rarely, if ever, asks whether it would be true to say “He knows thatp” of some putative knower. Moreover, in “A Plea for Excuses” Austin follows his famous claim that ordinary language should be the “first word” in philosophy with a footnote in which he makes the following plea to the reader: “And forget, for once and for a while, that other curious question Ts it true?’ May we?” (Austin 1979: 185). Many philosophers will find Austin’s remark puzzling.15 Whether sentences of the form “S knows that p” express truths has been the focus of 15 See DeRose (2002: 196), for example.
EPISTEMIC PRAGMATISM
|
171
many recent debates about the semantics of knowledge ascriptions. Austin, however, urges us not to ask, when investigating our knowledge claims, precisely the question at the center of the debate between contextualists and invariantists, namely: “Is what S said true?” He prefers to remain si lent on this issue. Austin’s remark might seem unmotivated or strange, and he says little to elaborate on this feature of his writings. We are left to wonder why we should temporarily “forget” about pressing the question of troth and falsity. Perhaps we can gain some clarity by reminding ourselves that Austin is primarily interested in the appropriate use of “knows that” in various contexts. He wants to explore what we do with knowledge claims. I have been emphasizing a similar point. When we examine the types of situations in which knowledge claims are made, both contextualists and noncontextualists tend to agree on the appropriate use of “knows that.” In the high-stakes bank scenario, for instance, the contextualist and the invariantist agree that Keith ought not say that he knows. Further, the con textualist and the invariantist agree that if a person has a true belief and is in a sufficiently strong epistemic position for the situation, then (ce teris paribus) it is “within reason” to say that she knows. Each semantic view proposes a more or less equivalent way to achieve what we set out to achieve, namely, the identification of good informants. This might in dicate why Austin is not very interested in determining the literal truth or falsity of such utterances. Insofar as one’s knowledge claim is “justified” or “within reason” or “right,” the further question about its troth has little bearing on what we do with these claims. Whatever Austin’s reasoning might be, it seems clear that our hypoth esis about the purpose of knowledge ascriptions will not instruct us as to whether these knowledge claims are strictly speaking true or merely warrantedly assertable.16 Thus, philosophers should probably stop trying to resolve the contextualism-invariantism dispute by appealing to considerations about function.17 I have been repeatedly urged to conclude this chapter here. But I cannot resist the temptation to explore a deeper reason why considerations about function will not guide us on this matter. In the remainder of this chapter,
16 Craig is also skeptical about drawing conclusions about semantics from function. Although his method leads to an account of our linguistic practices surrounding “knows that,” he recognizes that “it does not determine how to apportion the underlying mechanics of the practice between invariantist semantics and contextually motivated pragmatics” (1990: 167). 17 As mentioned earlier, this is not to suggest that reflecting on the function of knowledge claims will shed no light on their semantics.
172
|
What’s the Point of Knowledge?
I will propose the following unorthodox view: perhaps there is not always a determinate answer as to whether a knowledge claim is strictly speaking true (on the one hand) or false but appropriately asserted (on the other hand). In such cases, it would be a mistake to insist on the utterance’s truth or falsity. I call this view epistemic pragmatism, and I’ll explain why in §7.6. While I often quote Austin favorably, I do not claim he endorses this view. The fact that Austin asks us to temporarily “forget” about the truth of our knowledge claims is perfectly compatible with the com monly held view that knowledge claims have either a contextualist or invariantist semantics. However, rejecting this common idea will pro vide us with an alternative explanation for Austin’s enigmatic remark. Perhaps he thinks we should give up on the assumption that all serious and literal uses of “know” are in principle assessable in terms of truth and falsity. At times Austin speaks as though knowledge claims are nei ther true nor false, and indeed this is how David Lewis (1983), Avner Baz (2012), and others read Austin. For instance, Austin writes, “To suppose that T know’ is a descriptive phrase, is only one example of the descriptive fallacy” (1979: 103). However, I am not (that) inter ested in Austin exegesis. Whether this is Austin’s considered view or not, I claim it is a position worth taking seriously and that it can be recovered from his writings. Before outlining this idea, I want to manage your expectations. Epistemic pragmatism attacks a widely held picture of language, and a full defense of this view would require a detailed argument. My goal is less ambitious: I want to provide a sketch of this position in order to make it more visible. I believe that if we take the purpose of “knows that” as pri mary in our theorizing, we might realize there is something confused about the standard attempts to determine the semantics of knowledge ascriptions by identifying their truth conditions. While this is a fairly radical view, it is really only radical at the level of theory. It does not call for a revision of our ordinary practice of knowledge ascriptions; at most, we are required to rethink philosophical practice.
7.5 Austin on Truth and Falsity In How to Do Things with Words, Austin considers the question whether it is true or false to say that France is hexagonal. He refuses to give a straightforward answer:
EPISTEMIC PRAGMATISM
|
173
Suppose that we confront “France is hexagonal” with the facts, in this case, I suppose, with France, is it true or false? Well, if you like, up to a point; of course I can see what you mean by saying that it is true for certain intents and purposes. It is good enough for a top-ranking general, perhaps, but not for a geographer. . . . How can one answer this question, whether it is true or false that France is hexagonal? (1975: 143)
The most natural thing to say, according to Austin, is that this description is “pretty rough,” without ascribing to it a determinate truth-value. In such cases, insisting that the utterance must be either true or false is a mis take: “It is just rough, and that is the right and final answer to the question of the relation of ‘France is hexagonal’ to France” (Austin 1975: 143). We may interpret Austin’s remarks in multiple ways. On one interpreta tion, Austin is endorsing the contextualist idea that the truth of a sentence can vary depending on contextual features that go beyond those needed to fix the value of indexical elements in the sentence. To support this reading, we might remind ourselves that Austin says that “in the case of stating true or falsely, just as much as in the case of advising well or badly, the intents and purposes of the utterance and its context are important” (Austin 1975: 143). On this contextualist reading, it would be a mistake to insist that the utterance per se must be either true or false. To see why, we must distinguish two questions. First, is the sentence “France is hexagonal” true? Second, is what is stated in using that sentence on a particular occa sion true? To answer the first question affirmatively, every use of that sen tence would have to be true. But a contextualist will deny that the sentence can be classified as true or false without considering the circumstances in which it is used.18 In some contexts (e.g., when speaking to a top-ranking general) the sentence can be used to make a true statement, while in other contexts (e.g., when speaking to a geographer) it might be used to make a false statement. Thus, a contextualist will say the second question can be answered affirmatively in certain contexts, while the answer to the first question is indeterminate. An insensitive invariantist will think there is only one true answer to the question “Is France hexagonal?” This answer will not be shaped by features of the context but rather will be determined by the relation of
18 Austin says: “If you take a bunch of sentences .. . there can be no question of sorting them out into those that are true and those that are false; for... the question of truth and falsehood does not turn only on what a sentence is, nor yet on what it means, but on, speaking very broadly, the circumstances in which it is uttered” (1975: 110-11).
174
|
What’s the Point of Knowledge?
“France is hexagonal” to France. A strict insensitive invariantist will say that France is not hexagonal, so it is always false to say “France is hex agonal.” This sort of invariantist might claim it is perfectly appropriate to say “France is hexagonal” in certain situations (e.g., when speaking to a top general), but this is just loose talk. Given her strict standards for what counts as “hexagonal”—viz., all internal angles must be 120 degrees—it would never be true to say that France is hexagonal. A moderate insensi tive invariantist, by contrast, will have more lax standards for what truly counts as hexagonal.19 Assuming France meets this standard, the moderate invariantist will say that “France is hexagonal” is true in every conversa tional context, including those in which it would be inappropriate to utter this sentence (e.g., when speaking to a geographer). So far this parallels our discussion of contextualism and invariantism in epistemology. According to the epistemic contextualist, the sentence “Keith knows the bank is open” does not have the same truth-value on every occasion, even once we fix the values of the relevant indexicals. We must consider the circumstances in which it is used to determine truth and falsity. The insensitive invariantist denies this: what varies is not the statement’s truth-value but rather the standard for appropriate assertion. More specifically, the strict invariantist will deny that Keith meets the (high) standard to know, while the moderate invariantist will maintain that he meets the (moderate) standard to know. I now want to consider an alternative to both contextualism and invariantism. The main idea, put roughly, is that natural language leaves in determinate whether knowledge claims have a contextualist or invariantist semantics.20 This proposal casts doubt on a view of language that structures many philosophical discussions. Scholars who debate the meaning of knowledge ascriptions tend to share the belief that there is a definite an swer to the problem disputed (i.e., contextualism or invariantism?); they disagree only on the more specific question as to what the objective an swer is. When investigating our knowledge claims, these theorists jointly presuppose that the question “Did she say something true or something
19 As far as I know, nobody defends moderate invariantism with respect to geometric terms: the most plausible theories are contextualism and strict invariantism. I mention this view, however, because it provides a better comparison with moderate (i.e., nonskeptical) invariantism about “knows.” While moderate invariantism is not a very plausible theory about the semantics of geometric terms, it is a prima facie plausible theory about the semantics of knowledge claims. 20 This is not to say that any semantic theory is just as good as any other. Some are nonstarters (see Unger 1984: 33-35).
EPISTEMIC PRAGMATISM
|
175
false?” is always in order, in the sense that the agent must have spoken truthfully or falsely. I want to question this presupposition. Some of Austin’s remarks may be taken to support this idea. After all, he never explicitly claims that the truth-value of the thought expressed by the sentence “France is hexagonal” changes in different contexts. He instead says, “It is a rough description; it is not a true or a false one” (1975: 143). Likewise, when he considers the sentence “Lord Raglan won the battle of Alma,” Austin says it is “exaggerated and suitable to some contexts and not to others,” and he tells us “it would be pointless to insist on its truth or falsity” (1975:143-144). 21 Indeed, he shies away from using the words “true” and “false” when discussing this example: Did Lord Raglan then win the battle of Alma or did he not? Of course in some contexts, perhaps in a school book, it is perfectly justifiable to say so—it is something of an exaggeration, maybe, and there would be no ques tion of giving Raglan a medal for it. (1975: 143, emphasis mine)
The closest Austin comes to evaluating this sentence in terms of truth and falsity is when he says, “what is judged true in a school book may not be so judged in a work of historical research” (1975: 143). Notice, however, that he frames his discussion in terms of what we judge to be true, not what is true. Similarly, he says, “I can see what you mean by saying that [‘France is hexagonal’] is true” (1975: 143, emphasis mine). Again, his language is cagey and he stops short of claiming the utterance is true. We might reasonably conclude, then, that Austin thinks it would be a mistake to insist that these utterances must be either true or false even once we’ve considered the circumstances in which each is used. Perhaps the description is just rough, and that is the right and final an swer to the question. Before I argue (in the next section) that this type of semantic indetermi nacy is a plausible feature of our epistemic vocabulary, let me garner some support from outside epistemology. Consider the following case: Wanting milk for his morning cereal, Henry asks Sarah “Is there some milk in the fridge?” “No,” Sarah answers, “There’s none.” Hearing Sarah’s an swer, Henry puts the cereal away and fixes toast instead. All the while, Sarah
21 A bit of context: Lord Raglan’s orders were not transmitted to some of his subordinates.
176
|
What’s the Point of Knowledge?
has been studying the state of the fridge, including some dried splashes of milk on the shelves.22
Is what Sarah said true or false? If you incline toward contextualism, you’ll think what Sarah said is true because her utterance’s truth conditions are partially determined by her context—in this case, “There’s none” is true because there is no quantity of milk in the fridge sufficient for the usual purposes milk serves at breakfast. If you incline toward insensifive invariantism, you’ll think what Sarah said is strictly speaking false. There is some milk in the fridge, namely, the dried splashes of milk on the shelves; thus, Sarah spoke falsely. The strict insensitive invariantist will, or at least should, think that what Sarah said is appropriate because she implied something true: there is no usable milk for Henry’s cereal. It would have been “silly” (to use Austin’s term) for her to tell Henry there is milk. Were she to answer this way, she would miss the point of Henry’s question. Put differently, what Sarah says is false even though what she means is true. What Sarah means is there is no quantity of milk in the fridge sufficientfor the purpose of eating cereal; however, this is not what the sentence she uses means, according to the strict invariantist. What does this example illustrate? One lesson, which I stressed earlier, is that whichever semantic theory you prefer, we can all agree that Sarah said the right thing. Her utterance was maximally informative for the intents and purposes present in her context. Of course this leaves unan swered that other curious question, the one Austin asks us to temporarily forget, namely: Is what Sarah said true! While all sides of this dispute have spilled much ink over it, I find it plausible to say there is no fact of the matter as to whether contextualism or some type of invariantism is the cor rect account of Sarah’s utterance. They are equally good semantic models for understanding her claim, but neither has priority. Rather, they simply offer different ways to carve up semantics. I am not the first to make this suggestion. Peter Unger defends a similar view in his book Philosophical Relativity. In that work, Unger endorses the “relativist” conclusion that contextualism and invariantism are equally good theories and that there is no fact of the matter as to which view is cor rect. On this point, I fully agree with Unger; however, our views part com pany in two important ways. First, Unger says that if “know” exhibits this type of semantic indeterminacy, then “there is no objectively right answer
221 take this example from Lawlor (2005: 219), who borrows it from Travis (1989: 18-19).
EPISTEMIC PRAGMATISM
|
177
as to what ‘know’ means” (1984: 5). I will argue that we should reject this further claim.23 Second, my focus is much narrower than Unger’s. I will not argue that for some fairly wide class of expressions (e.g., “know,” “flat,” “cause,” “free,” etc.) there are plausible semantics of both the invariantist and contextualist type such that no matter of fact could decide between them. My attention is concentrated mainly on knowledge ascriptions. That said, Unger might be right to amplify the view. Global pragmatists like Huw Price (2011) would want to extend this account to a much wider class of expressions than “knows that,” and I see no reason to think this kind of indeterminacy would apply only to knowledge claims. To further motivate this idea, let’s consider another example from out side epistemology: Elizabeth lives in New York City and enjoys long walks. One day she walks from Battery Park to the Metropolitan Museum of Art, where she meets her friend Juliana. Juliana asks Elizabeth how long it took to walk that far. “Two hours,” Elizabeth answers. However, if Elizabeth were timing herself care fully she would realize that she actually walked for two hours, one minute, and 38 seconds.
According to the strict insensitive invariantist, what Elizabeth said is false. It is false because, strictly speaking, it took her more than two hours to walk from Battery Park to the Met. On this account, what Elizabeth meant is that it took her about two hours, but what she said is false. Suppose this is right. What would it have taken for Elizabeth to say something true? Imagine she carefully timed herself and told Juliana that her walk took “two hours, one minute, and 38 seconds.” Would that utter ance be true? It depends how strict you want to be. Elizabeth could have been more precise, and we can easily imagine contexts in which greater precision is required. Must she also know the number of milliseconds to speak truthfully? Why stop here? What about the number of microseconds or nanoseconds? How much is enough? And how do we draw the line in a principled way? Are all sentences involving measurements strictly speaking false? If so, does no one ever know what time it is? If the insensitive invariantist were stuck with this absurd conclusion, this would clearly be a point in favor of contextualism. However, Davis (2007:417)
23 By rejecting this idea, we also avoid a further undesirable conclusion drawn by Unger, namely, “there may be no objectively right answer to the problem of whether or not we know such and such to be so” (1984: 5).
178
|
What’s the Point of Knowledge?
convincingly argues that this conclusion does not follow. We can know that Elizabeth’s walk took about two hours, and that suffices to know how long it took. Thus, one can truthfully say, for instance, “The walk took about two hours.” What one cannot know, or speak truly about, is precisely how long the walk took, with no margin of error. But we need not know that in order to how long her walk took. Even though “two hours” would be strictly speaking false, one can still use this utterance to imply that one knows how long it took. At this point, the contextualist might appeal to something like the fol lowing principle: it is a strike against a theory of a common term of a natural language that it involves the speakers of that language in systematic and widespread false hood in their use of that term. (DeRose 1995: 46)
DeRose calls this the “methodology of the straightforward” (2005: 192). However, it is not clear why it is an advantage on linguistic grounds to make such claims count as true. As Davis (2007: 429) argues, “nothing in linguistics forces us to hold that the uses of ‘Two hours’ and other para digm examples of loose use are strictly and literally true.” A separate argu ment is needed to claim that the troths communicated by these utterances were communicated as part of lexical meaning. Thus, the contextualist is not clearly at an advantage here.24 Indeed, I doubt that anything about the meaning of the English words “two hours” will instruct us as to which of these ways is really the right way of accounting for things, for the very good reason that neither of these is “the right way.”
7.6 Epistemic Pragmatism
•
Now let’s return to epistemic evaluation. Many philosophers presuppose that the meaning of epistemic claims is to be settled by determining the truth conditions of those claims. This presumption is at the center of debates between contextualists and their rivals. However, a natural way to approach the meaning of epistemic claims is “to ask not about their truth-conditions but about what practical function they serve in our commerce with one
24 Jason Bridges (2008) also provides compelling criticisms of this methodology. His argument, in a nutshell, is that speakers can easily be brought to doubt their natural and appropriate straightforward uses of words, and DeRose provides no good reason to limit his focus to speakers’ initial judgments about the use of words. Thus, DeRose engages in simplistic and distorted accounting of the relevant phenomena.
EPISTEMIC PRAGMATISM
|
179
another” (Chrisman 2010: 116). If we assume in advance that debates about the meaning of epistemic claims must be settled by determining their truth conditions, we obscure from view accounts that explain the meaning of the target claims in terms other than truth conditions. This sort of nonfactualism about the semantics of epistemic vocabulary is essentially a pragmatist view, so I call it “epistemic pragmatism.” The epistemic pragmatist encourages us to rethink a common presupposition in epistemology about how best to account for the meanings of epistemic claims. We are to look for the socialpractical roles of words and concepts rather than referents in order to under stand their meaning. William Alston (2000), Matthew Chrisman (2010), and Avner Baz (2012) suggest something similar. “Pragmatism” is of course a vague, ambiguous, and overworked word, as Richard Rorty reminds us (1982: 60). In other philosophical contexts it might be important to emphasize the differences between myself and other so-called pragmatists, but for the purposes of this chapter these differences may be ignored. I use this word because it loosely reflects an attitude or approach that I share with pragmatists such as Wittgenstein, Rorty, and Price. These authors seek to dismiss or demote philosoph ical puzzles in favor of more practical questions about the roles and functions of the words and concepts in terms of which we talk and think about the things that lead to philosophical puzzles.25 This type of prag matism has a second-order, or linguistic, focus. An epistemic pragma tist will investigate the role of the term or concept “knowledge.”26 As I have been emphasizing throughout this book, I am likewise interested in why humans think and talk about knowledge. The guiding thought is that if we can explain how creatures in our circumstances naturally come to speak in the ways that we do, there will be no further puzzle about the truth conditions of knowledge ascriptions. Indeed, pragmatists tend to deny that questions about the truth conditions of statements, sentences, or beliefs, as well as the referents of terms, have determinate answers that philosophy is tasked to provide. A better approach to knowledge is to in vestigate the function of human discourse, to ask what difference having such a concept makes to us.
25 What “things” are these, precisely? Some common targets include causation, goodness, and truth, but knowledge is a plausible candidate for this list. Pragmatists have asked about the social role of epistemic concepts such as knowledge (see Rorty 1979 and Brandom 2000). 26 The epistemic pragmatist might also focus on our concepts of rationality, justification, understanding, wisdom, etc. Here I am concerned solely with knowledge, but I will investigate the concept of understanding in chapter 9.
180
|
What’s the Point of Knowledge?
“Pragmatism” is an apt label for another reason. On my view, what it takes for a true belief to qualify as knowledge is partly determined by the relevant practical reasoning environment—whether it is that of the subject, the attributor, or the broader epistemic community. Knowledge is therefore closely tied to practical matters, not just factors traditionally recognized as epistemic. Elsewhere this idea has been labeled “pragmatic encroachment.” Thus, pragmatic considerations shape my view in at least two ways. There are doubts about pragmatism in general, of course, but I set these aside in the hope of getting clearer about a distinctively epistemic type of pragmatism. This view could be motivated by worries about the possi bility of there being absolute facts about whether someone knows some thing. There is a sense in which the invariantism/contextualism debate is about whether there are such facts, but it is usually conducted separately from traditional realism/antirealism debates, and it is almost entirely fo cused on various data concerning linguistic usage. Paraphrasing Matthew Chrisman, we may put the point as follows: Epistemic pragmatists encourage us to ask not about the nature of epistemic facts or values but instead about the nature of epistemic evaluations. They do so because they think we can satisfactorily answer the latter question without ever countenancing some sui generis kind of thing—epistemic facts or values—to answer the first question. Their idea is that epistemic evaluations are not representations of how the world is; rather, they are expressions of our attitudes towards the way we represent the world to be. (Chrisman 2010: 115)
Beyond the usual motivations for pragmatism, this is a natural view to adopt if we remind ourselves of three things: first, it is plausible that eval uative concepts are shaped by the purposes for which they are developed; second, our epistemic evaluations achieve their characteristic purpose whether we understand their appropriate use in terms of literal content or conversational pragmatics; third, we should not assume in advance that the distinction between traditional semantics and pragmatics will always be useful to us when we wish to elucidate some word or concept. There may be no way to apportion the complexity inherent in usage between the se mantics of knowledge attributions and pragmatic considerations. In other words, we should not presume in advance that one of these opposing se mantic theories of “knows” is inherent in our concept of knowledge. Thus, I am not just making the epistemic claim that knowledge ascriptions have determinate truth conditions but we cannot know them. My hypothesis
EPISTEMIC PRAGMATISM
[
181
is that there simply is nothing to discover as to whether contextualism or some type of invariantism is ultimately correct. These observations are not likely to impress the contextualist or the invariantist, given the methodological paradigm in which they seek to understand our concept of propositional knowledge. Even if we assume that knowledge ascriptions can fulfill their role irrespective of whether their semantics is contextualist or invariantist, it does not follow that nei ther of these views is correct, or that the answer is indeterminate. One might argue that it is far more plausible that the semantics of knowledge ascriptions could have been one way or the other, at some point, but then the semantics became settled. (An analogy: the foundation of my house could have built from either wood or stone. Both materials would have got the job done, but that doesn’t mean that my house’s foundation is now made from both, or that it is indeterminate.) Our ability to answer the tra ditional semantic question may be no part of what our everyday, compe tent employment of “knows that” requires, but this does not mean there is something confused in principle about standard approaches to semantics. The problem with this reply, however, is that it simply assumes the truth of the orthodox view about how language works, namely, that the se mantics of “knows” is determinate between contextualism and some type of invariantism. What is under investigation is the way we insist, in the course of doing philosophy, on the truth (or falsity) of a sentence that is uttered. Why expect our knowledge ascriptions to come equipped with a semantics that is determinate between contextualism and invariantism? The fact that our knowledge ascriptions achieve their purpose whether or not a specific semantic theory is true of them should lead us to wonder why we must assume that one of these semantic accounts is somehow fun damental to our concept of knowledge. None of this questions the factivity of knowledge. If someone knows that p, then her belief that p must be true. Further, I am not claiming that the truth of a knowledge attribution is indeterminate because the truth of the putative knower’s belief as to whether p is indeterminate (although that may happen). What I am arguing is that there are cases in which the truth of an attribution of knowledge to a subject is indeterminate. We may find this indeterminacy even in cases where the putative knower’s belief is determinately true; for example, Keith’s belief that the bank is open to morrow might clearly be true even though it is indeterminate whether it is true to say that he knows this (or even whether he knows this). This is all still highly schematic and impressionistic—the details need to be filled out and amplified by more considerations than can be offered
182
|
What’s the Point of Knowledge?
here. There are many important questions to which I can give only very in complete answers. These include the following: Are epistemic pragmatists incapable of providing a full semantic account of knowledge ascriptions? Do knowledge ascriptions fail to express complete propositions on this view? Is the epistemic pragmatist unable to account for the appearance of knowing that some epistemic statements are true, or the possibility of fun damental error? Am I claiming there is no sensible distinction between the semantics and pragmatics ofknowledge ascriptions? I think the answer to all these questions is “No,” and I will briefly gesture as to why.
7.7 Unanswered Questions Does epistemic pragmatism entail that all serious and literal uses of “knows that” cannot be assessable in terms of truth and falsity? If it does, how do we account for our practice of trying to communicate facts when we evaluate people epistemically? Surely we attempt to state what is true when making epistemic claims, not just what we think is appropriate. So the objection goes. But the epistemic pragmatist need not deny that in certain situations, for particular intents and purposes, one might use a sentence to state a truth. Our knowledge claims certainly have intuitive truth-values; for in stance, someone who claims to know that p on the basis of an unedu cated guess seems to speak falsely, and someone who claims to know that p on the basis of evidence that guarantees the truth of her belief seems to speak truly. Acknowledging this fact does not sell us out to contextu alism or invariantism. The point is that we can legitimately speak of truth and falsity—we have the right to treat epistemic commitments as truthapt—when this question naturally arises within suitable contexts as part of ordinary and normal practice. Following Austin, it might be useful to characterize truth and falsity as labels for positive and negative poles in a variety of more specific forms of appraisal.27 He writes, We become obsessed with “truth” when discussing statements, just as we become obsessed with “freedom” when discussing conduct. So long as we think that what has always and alone to be decided is whether a cer tain action was done freely or was not, we get nowhere: but so soon as we turn instead to the numerous other adverbs used in the same connexion
27 Austin doubts that every statement aims to be true (1979: 131).
EPISTEMIC PRAGMATISM
|
183
(“accidentally,” “unwillingly,” “inadvertently,” &c.), things become easier, and we come to see that no concluding inference of the form “Ergo, it was done freely (or not freely)” is required. Like freedom, truth is a bare min imum or an illusory ideal (the truth, the whole truth, and nothing but the truth about, say, the battle of Waterloo or the Primavera). (1979: 130)
The idea is that we use numerous specific forms of appraisal to evaluate what people say: they might be justified, reasonable, fair, adequate, and so forth. Each would be occasion-bound. But we should not fall into the trap of what Austin calls “the true/false fetish” (1975: 151).28 A statement might be “exaggerated or vague or bald, a description rough or misleading or not very good, an account rather general or too concise” (1979: 129). In cases like these, Austin says it is “pointless” to insist on deciding whether the statement is true or false. Is it true or false that Belfast is north of London? That the Galaxy is the shape of a fried egg? That Beethoven was a drunkard? That Wellington won the battle of Waterloo? There are various degrees and dimensions of suc cess in making statements: the statements fit the facts always more or less loosely, in different ways on different occasions for different intents and purposes. (1979: 130)
We should not expect to find a straightforward answer by reference to or dinary language. Instead, what we might say is this: there is a reason to say that the Galaxy is the shape of a fried egg (or that there’s milk in the fridge, or that one knows the bank is open, etc.) and there is a reason to deny this. We are not, however, ignorant of some truth that still awaits dis covery. If epistemic pragmatism is correct, there is no truth of the matter beyond the facts already described—no further evidence that will lead us to the “real” answer. Am I just fudging over an important distinction? If we accept epi stemic pragmatism, do we risk running together semantic and pragmatic considerations that ought to be kept apart?
28 With Austin, we might endorse a form of deflationism about truth—a view on which truth is a thin or nonexplanatory notion. According to this form of deflationism, saying that a statement is true is just a way of saying that the statement has one or another of a range of more specific positive qualities—for example, that it is satisfactory, correct, fair, etc. (1979: 130; 250-251; 180). The predicate “is true” does not always have a descriptive function: it needn’t serve to characterize the obtaining of a relation between statements and facts.
184
|
What’s the Point of Knowledge?
Epistemic pragmatism would not obliterate the common distinction, often appropriate and useful, between what someone actually says and what she merely implies or suggests. Nothing I have said (or implied!) commits me to that idea. In many cases there is an intuitive difference be tween what is semantically expressed and what is pragmatically conveyed. For example, if Stewart asks Jessica whether their flight has a stopover in Chicago, and Jessica responds, “Do I look like an airline agent?” she surely implies but does not say that she doesn’t know. Typically, the semantic aspects of an expression are more closely tied to the expression itself, while the pragmatic aspects concern more accidental, removed features. What I am questioning is a very particular way of trying to draw this dis tinction, not the distinction itself. The epistemic pragmatist need not rule out the invocation of such distinctions, for they are often worth drawing.29 But we should not underestimate the difficulty involved in preserving a sharp distinction between the semantics and pragmatics of knowledge ascriptions. There are a variety of cases in which this distinction does not admit of a clear-cut application; indeed, how to draw this distinction is highly controversial (and not at all intuitive) in all the well-worn cases in the contextualism debate (bank cases, airport cases, and so forth). As this literature makes clear, how to distribute the underlying semantics for ap propriate knowledge ascriptions is strongly contested. A theorist in the grip of the traditional paradigm might reply as follows: unless there is a determinate way to distribute the complexity in usage between semantic and pragmatic considerations, our knowledge claims will fail to express complete propositions. As a result, it might follow that we are unable to fully understand what our knowledge claims even mean. Unger, for example, says that if “know” exhibits semantic in determinacy, then “there is no objectively right answer as to what ‘know’ means” (1984: 5). If we adopt the traditional assumption that a sentence only expresses a genuine proposition when it has determinate truth-conditional seman tics, it will follow on my view that many knowledge claims fail to ex press genuine propositions. This might seem like an undesirable result, but
29 Baz also (2012: 120) argues that we can salvage the semantics/pragmatics distinction even if we reject both contextualism and invariantism. Recanati (2004: ch. 5) claims this distinction corresponds to a range of phenomena that constitute a natural class, from a phenomenological point of view. Recanati, Baz, and I all think the distinction is blurrier than many presuppose. Taking this line further, Malcolm (1949), Wittgenstein (1969), and Travis (1991) argue there is no sensible distinction between the semantics and pragmatics of knowledge ascriptions (see Kusch 2009:73).
EPISTEMIC PRAGMATISM
|
185
much of its sting is removed if we realize that “proposition” here is being used in what Unger calls “an especially artful sense, a sense in which no propositions can be at all indeterminate, a sense in which all must be determinately true or else determinately false” (1984: 28). Further, this objection simply assumes the standard picture of truth conditional semantics from which it follows that there is no fact of the matter as to the full semantics of the relevant expressions. Even Unger, who defends semantic indeterminacy, buys into the existing paradigm: he says we can only account for the meaning of “knows” by providing a de terminate truth-conditional answer on this issue. However, there are ways of accounting for meaning in non-truth-conditional terms; for example, one of Alston’s “axioms of meaning” is this: the fact that an expression has a certain meaning is what enables it to play a distinctive role (enables it to be used in a certain way) for communicative purposes (2000: 4). Indeed, Alston says the meaning of an expression is just whatever fits it for a cer tain communicative role. He calls this the use principle: “an expression having a certain meaning consists in its being used to play a certain role (to do certain things) in communication” (2000: 154).30 This principle is not true for all communicative roles of language; it is only appropriate for a subclass, namely speech acts. However, this idea fits nicely with Austin’s own account of knowledge claims. Austin says utterances of the form “I know that such and such” serve a performative and not a descriptive function. We perform the speech act of “giving others our authority” for saying that such and such (1979: 99). On this view, the function of “I know” is similar to the function of “I promise”: both serve as ways of giving one’s word (1979: 97-103). The more general point is that epistemological evaluations are not straightforward descriptions of the world. To say that someone knows something is not just to describe the world in a way that picks out knowl edge; it is to pass a value judgment, to evaluate the agent. As Mark Heller says, “Attributing knowledge is not just describing S’s epistemic condi tion but also asserting that that condition is good enough” (1999: 118). It indicates the belief is good enough to be relied on for practical reasoning or to reasonably terminate inquiry, as I’ve argued in this book. In this way, epistemic evaluations are expressions of a kind of belief whose role in
301 am not putting forward a theory of meaning or an analysis of the meaning of “meaning.” That project goes well beyond my ambitions in this chapter. My point is that once we make room for alternative accounts of meaning, it becomes clear that nothing forces us to accept truth conditional theories of what “knows that” means.
186
|
What’s the Point of Knowledge?
our life is practical but not representational. This type of function-first ap proach is essentially pragmatic and expressivist: we do not merely report epistemically relevant facts when we attribute knowledge; we also commit ourselves to the idea that there is some epistemic position that is good enough to deserve the sort of approval expressed by “knows.” Perhaps my argument will strike the reader as disappointingly unimag inative. Why am I so convinced that this type of semantic indeterminacy is ineliminable? Typically underdetermination of theory by evidence is a passing trait, as Philip Kitcher suggests in The Advancement of Science. Over time, what once seemed undetermined may cease to be so. Have I not looked hard enough for probative ways of furthering our understanding of knowledge claims within the methodology of truth-conditional semantics? There are many ways to attempt to adjudicate the disagreement be tween contextualists and their rivals. As mentioned at the start of this chapter, philosophers have appealed to the ways in which “knows that” functions in comparison with other words, to the plausibility of psycho logical biases that people routinely exhibit, to “semantic blindness,” and to certain tests of the genuine context-sensitivity of a term.31 It remains uncertain, however, whether these tools are sufficient to determine the proper semantics of “knows that.” It is for precisely this reason that many theorists have been motivated to consider the, function of epistemic evalua tion to help adjudicate this dispute (as discussed in §7.1-7.4). But, as I’ve argued, this strategy also fails. It is therefore difficult to see how the cur rent debate between contextualism and invariantism about “knows that” can be described as anything but a stalemate. On these grounds, I have recommended the possibility that this stalemate indicates a worry about a philosophical paradigm.
7.8 Concluding Remarks Several epistemologists have recently appealed to putative facts about the purpose of knowledge ascriptions to provide evidence for this or that
31 Another way one might try to settle this debate is by arguing that contextualism and sensitive invariantism are unorthodox views in epistemology, whereas insensitive invariantism is the orthodox view, and that—all else equal—we should prefer orthodox views to unorthodox views. While this argument is not explicit in the literature, I think it structures a lot of the debate about the semantics of knowledge ascriptions. However, I do not find this argument compelling. Not only is it dubious to think orthodox views should win by default, it is also false that epistemic contextualism is unorthodox. A recent survey confirms that the majority of philosophers are contextualists (Bourget and Chalmers 2014).
EPISTEMIC PRAGMATISM
|
187
semantic theory. This chapter has raised doubts about the viability of this strategy. While the function of knowledge ascriptions does shed some light on their semantics, it does not instruct us as to whether functional role con textualism or insensitive invariantism is true. Are we left at an impasse? Can anything break the tie between these views? A variety of factors must be carefully investigated in order to ade quately determine whether contextualism, invariantism, or some other view is true of our knowledge ascriptions. This of course is a large and complicated task that involves a variety of considerations, to which my discussion is merely a contribution. It is possible that the truth-conditional semantics of knowledge ascriptions will be settled on other grounds. However, it is also possible that there is a deeper mistake about the debate between contextualism and invariantism in its current form. I have cau tiously suggested that there may be no way to apportion the complexity inherent in usage between the semantics of knowledge attributions and pragmatic considerations. No doubt someone who is already committed to the traditional view will not be persuaded. Such a person might think I am too eager to dissolve the debate between contextualism and invariantism rather than advance it. I admit a good deal of uncertainty about the correct answer to this question, but this uncertainty gives no real support to the traditional view. Thus, I offer epistemic pragmatism as a schema for po tentially undercutting the debate. But even if we hold fire on the issue of whether a truth-conditional se mantic account can be given of our knowledge ascriptions, I have argued that ideas about function will not help us resolve the contextualism/ invariantism dispute. As a result, philosophers will have to look elsewhere for evidence about which view offers the correct semantics of our knowl edge ascriptions.
188
|
What’s the Point of Knowledge?
CHAPTER 8
Skepticism and the Point of Knowledge
One must bear in mind that distinction, which I have insisted on in various places, between the actions of life and the search for truth. —rene descartes (V Rep.: VII 350, HR2 206)
skeptical arguments, we know almost nothing. Perhaps we know some simple a priori truths and claims about some aspects of our immediate experience; but many skeptics believe that we know little, if anything, else. I am particularly interested in the skep tical view that we know nothing, or almost nothing, about the physical world around us. This is often called “external world skepticism,” but I will simply call it “skepticism.” What is the relationship between skepticism and the concerns of daily life? A central motif in Hume’s writings is that of an irreconcilable clash between two perspectives: our everyday outlook and the attitude to which we are led by philosophical reflection. This conflict has been a recurrent theme in discussions of skepticism ever since. In his book on Descartes, Bernard Williams (1978: 48) locates the standing invitation to skepticism in the philosophical aspiration to provide an “absolute conception of re ality.” Barry Stroud (1984) claims the road to skepticism is paved as soon as we start philosophizing. Thomas Nagel (1986) argues that skepticism is inherent in any attempt to understand the world from an objective per spective. And Michael Williams (1996: 255) emphasizes the connection between skepticism and the attempt to understand human knowledge in “a distinctively philosophical way.” according to some powerful
I think this view of skepticism is basically right. But what are we to make of this fact? If skepticism is the natural outcome of philosophical reflection, does this mean that philosophy cannot release us from the skeptic’s grip? Many have sought to resolve this clash between philosophy and daily life by appealing to one perspective against the other. For example, J. L. Austin famously insisted that a philosophical theory of knowledge must be true to facts about the circumstances in which it is ordinarily appropriate to make (or challenge) knowledge claims. He catalogs numerous examples in which skepticism seems to lack fidelity to our ordinary practice of ep istemic evaluation, and he takes this to be a significant strike against the skeptical conclusion. The problem with this fine of thought, however, is that it leaves mysterious why skepticism gets any purchase on us. Why do we have a concept of knowledge that offers a standing invitation to skepticism? In The Significance of Philosophical Skepticism, Barry Stroud says we are tempted by the skeptic’s reasoning because her standards for knowledge are the ones to which we are actually committed, contrary to what Austin would have us believe. According to Stroud, we often attribute knowledge to others because the procedures we follow in daily life reflect the “prac tical exigencies” of action. We recognize that for practical purposes we cannot entertain skeptical doubts, but in the theoretical context of philo sophical reflection we come to realize that we actually know little, if an ything, about the world around us. The problem with this line of thought, however, is that it does not explain why the true conditions of knowledge are revealed by philosophical reflection rather than practical life. I will not try to resolve the clash between philosophy and daily life by appealing to one perspective against the other. Rather, I want to see if we can reconcile these perspectives in a way that honors the insights of them both. To accomplish this, we must answer a question that Hume brings to light, namely, why are skeptical doubts so difficult to accept in daily life? Michael Williams (1996: xix) puts this question another way: Why is the conviction that skepticism commands not easily detached from the special context of philosophy reflection? To adequately account for the basis of skepticism, we must explain why skeptical doubts are unstable in this way. To see the shape of my proposal, we need to revisit a distinction that Descartes himself makes, but does not make much of. In the fourth part of the Discourse on Method, Descartes says that in “practical life” it is often “necessary to follow opinions which one knows to be very uncertain,”
190
|
What’s the Point of Knowledge?
while one who is devoted “solely to the search for truth . . . should do just the opposite,” namely, one should “reject, just as though it were abso lutely false, everything in which [one] could imagine the slightest doubt” (VI 31-32). Here Descartes draws our attention to two distinct forms of inquiry. Following Bernard Williams (1978), I will call them “practical inquiry” and “pure inquiry.” By relying on this distinction, I believe we can provide a theory of knowledge that explains how and to what extent knowledge is possible, without assuming there is something drastically mistaken about the skeptical challenge. My main interlocutors in this debate are Descartes, J. L. Austin, Barry Stroud, and Mark Kaplan. I will investigate the Cartesian requirement for knowledge, Austin’s response to the skeptic, Stroud’s rejoinder to Austin, and Kaplan’s recent defense of Austin. I hope to make a new and useful intervention in this debate by appealing to considerations about the pur pose of epistemic evaluation to better understand the nature and force of philosophical skepticism. I beHeve a function-first approach can deepen our understanding of the persuasive power of skepticism while also explaining why skeptical worries do not (and should not) threaten eve ryday knowledge.
8.1 The Possibility of Error Skepticism takes many forms. Here I will focus on just one type of skep tical argument, albeit one that is both historically significant and widely discussed in recent epistemology: the argument from ignorance. The argument from ignorance is characterized by its use of “skeptical hypotheses” (see DeRose 1995: 1). A skeptical hypothesis describes an undetectable cognitively debilitating state such as dreaming, halluci nation, or victimization by an evil demon. Roughly, a hypothesis is “skeptical” if (a) its truth is inconsistent with some propositions we ordinarily take ourselves to know, and (b) the hypothesis is compat ible with all our experience in favor of those ordinary propositions. Consider, for instance, the following brain-in-a-vat version of the skeptic’s argument: 1. I don’t know that I am not a handless brain in a vat. 2. If I don’t know that I am not a handless brain in a vat, then I don’t know that I have hands. 3. Therefore, I don’t know that I have hands.
SKEPTICISM AND THE POINT OF KNOWLEDGE
|
191
Although Descartes never mentions brains in vats, the origin of this argu ment can be traced to his Meditations, if not earlier. Many contemporary epistemologists provide essentially the same formulation of skepticism.1 Suitably articulated, the skeptic’s reasoning will lead us to doubt our puta tive knowledge of the world around us. Skeptical hypotheses may be divided into four broad types. To illustrate these four types, I will start with an example: I cannot find my Siamese cat named Garfield. The most reasonable place to look for him is the kitchen, since he likes hiding there. In order to know whether Garfield is in my kitchen, I must look around. When I explore the kitchen, I eliminate various possibilities—for instance, that Garfield is under the table, behind the door, and so forth. While looking for Garfield, I hear a “meow” come from the living room. I walk over to the living room and see a Siamese cat that looks like Garfield on the rug. Presumably, I now know that Garfield is in the living room.
Yet my experience is compatible with a number of different possibilities. Perhaps a stray Siamese cat has entered my house and is meowing in the living room, while Garfield is still hiding in the kitchen. Or perhaps Garfield has been stolen and replaced with a cat that is perceptually indis tinguishable from him in every way. Or perhaps I am a brain in a vat and never owned a cat at all. Each of these possibilities is highly unlikely, but clearly some possibilities are more outlandish than others. What this helps to show is that error possibilities may differ along at least two dimensions: their scope and falsifiability. These notions op erate tacitly in many discussions concerning skepticism, but they must be made explicit if we are to adequately assess any purported solution to the skeptic’s challenge. Let’s first look at the scope of a skeptical possibility. Some error possibilities are local in scope and others are global.1 2 Local error possibilities do not threaten a wide range of knowledge claims, whereas global error possibilities have the potential to falsify many, or most, of our attributions of knowledge. For example, the possibility that someone has replaced Garfield with an identical cat is a local error pos sibility because it might undermine my claim to know that Garfield is in
1 See DeRose (1995), Schiffer (1996), and Cohen (1999). 2 Cohen (2000: 103) makes a similar distinction. For convenience I will speak of global and local error possibilities as two broad types, but this distinction is really a matter of degree.
192
|
What’s the Point of Knowledge?
the living room, but it does not threaten my knowledge of other facts, such as my country of birth, who won the 1990 World Series, and so forth. In contrast, the brain-in-a-vat hypothesis is a global error possibility be cause it bears on the truth-values of a wide range ofknowledge claims: if I were actually a brain in a vat, virtually all of my knowledge about the world would be lost. (If you think semantic extemalism would make these knowledge claims true, then assume I was recently envatted.) Another important dimension of skepticism is whether or not a given skeptical hypothesis is in principle falsifiable. Nonfalsifiable error possibilities cannot be eliminated on the basis of any evidence because there is nothing in principle available to indicate whether or not such an error possibility obtains. These hypotheses have an inbuilt immunity to falsification because the perceptual experiences that I might take to serve as evidence against such a hypothesis are epistemically indistinguishable from experiences that I would have regardless of whether or not such a error possibility were true. For example, my perceptual experience of having hands does not count as evidence against the hypothesis that I am merely a handless brain in a vat because I would have epistemically indis tinguishable perceptual experiences whether I were envatted or not. This isn’t the case with falsifiable error possibilities, where further investiga tion can confirm or deny the error possibility in question. In the example of Garfield, checking his ID tags and characteristic markings would suffice to rule out the possibility that this is a stray cat that has wandered into my living room. It is always in principle possible to heighten the standards required to count as having falsified an error possibility; for instance, one might claim that checking the cat’s marking and tags does not rule out the possibility that Garfield has been stolen and replaced with a perceptually indistin guishable cat with falsified ID tags. By shifting the evidential standards to a point where the error possibility in question is placed outside of the process of gathering and weighing evidence, we can convert a falsifiable hypothesis into a nonfalsifiable one. Falsifiable error possibilities must fix the relevant epistemic standard at a level that can in principle be met by an inquirer. These two dimensions—scope and falsifiability—give rise to four kinds of skeptical scenarios. First, there are error possibilities that are global in scope and nonfalsifiable, such as the brain-in-a-vat hy pothesis and Descartes’s evil demon hypothesis. I shall refer to error possibilities that are global and non-falsifiable as radical skeptical hypotheses. Second, there are error possibilities that have global scope
SKEPTICISM AND THE POINT OF KNOWLEDGE
|
193
but are in principle falsifiable. This sort of skeptical scenario is not often discussed in epistemology because global error possibilities are often assumed to be nonfalsifiable. In the case of dreaming, however, both Austin (1979) and Jim Stone (1984) think we can tell that we are awake rather than dreaming (when that is how we are) because waking states are “self-recognizing.” They therefore regard dreaming as a global yet falsifiable possibility. I’ll call this type of skepticism broad skepticism. Third, there are local error possibilities that are nonfalsifiable, such as the possibility that Garfield has been replaced with a Siamese cat that is indistinguishable from him in every possible way. I’ll refer to this form of skepticism as moderate skepticism. Finally, there are local error possibilities that are falsifiable (as long as the standards are fixed at a reachable level), such as the possibility that the cat in my living room is actually my neighbor’s Siamese cat. For the sake of simplicity, I will list all four types of skepticism: Radical = global scope + nonfalsifiable Broad = global scope + falsifiable Moderate = local scope + nonfalsifiable Mild = local scope + falsifiable Some people might wish to reserve the term “skepticism” for positions arguing that we do not have any knowledge or that knowledge is impos sible. In his Introduction to Contemporary Epistemology, Jonathan Dancy characterizes skepticism as the view that “knowledge is impossible. No one does know, because no one can know” (1985: 7). Similarly, Marcus Willaschek defines skeptical possibilities as those global error possibilities that “can never be ruled out by empirical investigation” (2007: 253). I do not wish to have a verbal dispute about which views should be classi fied as skeptical. I am trying to be ecumenical by using the word “skep ticism” (and its cognates) to include a variety of error possibilities that might be plugged into the skeptic’s argument. All four versions of skep ticism threaten to undermine ordinary knowledge that we intuitively take ourselves to have, so we ought to be concerned with all four types when answering challenges to our everyday knowledge. By extrapolating away from the details, we can provide a more gen eral formulation of the skeptic’s argument that accommodates all four types of skepticism. Let’s say that O represents some ordinary proposition about the external world that we intuitively know (e.g., I have hands) and SH represents a suitably chosen skeptical hypothesis that is inconsistent
194
|
What’s the Point of Knowledge?
with O (e.g., I am a handless brain in a vat). The general structure of the skeptic’s argument is the following:
1. I don’t know that not-SH. 2. If I don’t know that not-SW, then I don’t know that O. 3. Therefore, I don’t know that O.
DeRose thinks this argument is “clearly valid ... and each of its premises, considered on its own, enjoys a good deal of intuitive support” (1999: 23). Stewart Cohen agrees that “both of these premises are intuitively quite appealing” (1999: 62). The first premise is defended on the grounds that however unlikely or strange it might seem to suppose that I am in a skeptical scenario, it also seems true that I do not know that I am not in one—as DeRose says, “how could I know such a thing?” (1995: 2).3 Further, if I don’t know whether or not I am in a skeptical scenario, then it seems that I do not know many things about the world around me. This claim derives its force from the notion that knowledge transfers across known entailments, and hence that some sort of closure principle holds for knowledge. Roughly: if you know one proposition and know that that proposition entails another, then you know the latter proposition. But if we do not know the falsity of a skeptical scenario, then we can derive a skeptical result from the closure principle in the following way: if we know that O, then we know that not-SH; but we don’t know that not-SH, so we don’t know that O. While there are some problems involved in finding a satisfactory artic ulation of the closure principle, the idea that knowledge is closed under known logical implication is widely accepted.4 Denying this principle would license what DeRose calls “abominable conjunctions” (1995: 2729). An example of an abominable conjunction is “I know where my car is parked, but I don’t know whether it has been stolen and moved.” Another example is “I know that I have hands, but I don’t know that I’m not a
3 Nichols, Stich, and Weinberg (2003) have challenged the alleged intuitiveness of the first premise of the skeptic’s argument. I defend the intuitiveness of skepticism from their argument elsewhere (Hannon 2017c). 4 See Hawthorne (2004) for a discussion of the closure principle. The proposal that we should resolve skeptical worries by rejecting closure has been criticized by Lewis (1996: 564), Vogel (1990: 13), Cohen (1988: 105), Feldman (1995: 487), Schiffer (1996: 320), DeRose (2009: 29), and many others. In contrast, Nozick (1981), Dretske (1970), and Heller (1999) have denied closure.
SKEPTICISM AND THE POINT OF KNOWLEDGE
|
195
handless brain in a vat.” I will assume that some version of the closure principle holds. The skeptic’s argument looks valid and its premises are intuitively plau sible. The problem is that the skeptic’s conclusion seems false: it conflicts with our compelling belief that we do have all sorts of everyday knowl edge. As Lewis puts it, “It is a Moorean fact that we know a lot. It is one of those things that we know better than we know the premises of any phil osophical argument to the contrary” (1996: 549). The result is a paradox:
1. We have all sorts of everyday knowledge. 2. We don’t know that we’re not in a skeptical scenario. 3. If we don’t know that we’re not in a skeptical scenario, then we don’t have all sorts of everyday knowledge. This is a paradox because each of these jointly inconsistent propositions seems true. To escape the paradox, we have to give up something. I maintain that different versions of skepticism require different diagnoses. To anticipate one of my conclusions, I will argue that falsifi able error possibilities that are unlikely to occur will not greatly shape our concept of knowledge; thus, our everyday knowledge attributions tend to be true if the putative knower efiminates all of the relevantly nearby error possibilities. To anticipate another one of my conclusions, I will argue that nonfalsifiable error possibilities are rightly ignored in daily life. These possibilities make no noticeable difference to us because they are outside of the process of gathering and weighing evidence altogether. We therefore should not expect such skeptical worries to exert much pressure on our concept of knowledge. This is in keeping with the “relevant alternatives” account of knowledge that I developed in chapter 3 (and I will say more about this soon).5 Now that I have characterized the skeptic’s argument, I should briefly say something about what skepticism is not.
8.2 The Low Road to Skepticism Some scholars think skepticism merely imposes abnormal requirements on our familiar concept of knowledge. For example, Hans-Johann Glock says
5 Austin (1979), Dretske (1970), Stine (1976), Goldman (1976), Cohen (1988), and many other scholars have developed relevant alternatives views.
196
|
What’s the Point of Knowledge?
the skeptic ... is like someone who claims that there are no physicians in London, since by ‘physician’ he understands someone who can cure any disease within twenty minutes. (Glock 2010: 100)
On this line of reasoning, the imposition of this newer, restrictive standard is not a threat to us because the skeptic simply changes the topic. By illic itly importing unduly high standards, we are no longer talking about the thing we were originally talking about. Paul Moser calls this the “low road” to skepticism. The skeptic takes the low road by resting her argument on “a common but unconvincing philosophical ploy: redefining a key term to gain support for a controver sial claim” (Moser 1990: 127). A variety of scholars accuse the skeptic of taking the low road, including Moser (1990), Hanfling (2000: 125), and Glock (2010). Indeed, Moser targets Descartes, Hume, Stroud, and Unger as the chief offenders here. He says these skeptics advance their skepticism by “redefining” knowledge, specifically, by imposing abnor mally high standards for what knowledge is. This amounts to nothing more than “a low victory by high redefinition—really no victory at all” (Moser 1990: 127). In contrast, the skeptic takes the “high road” by accepting our everyday notion of knowledge and then arguing that on that very notion we cannot (or at least do not) have knowledge about the world around us. Does our skeptic take the low road? Although many have accused the skeptic of a shallow victory, it is rather this diagnosis of skepticism that is shallow. If skepticism rested on a distortion of meaning, then we would be as unsurprised and unperturbed by the skeptic’s denial of knowledge as we are by the claim that there are no “physicians.” But then why have we not already dismissed the skeptic? Presumably it is because skepticism is not obviously violating our ordinary standards for knowledge. Upon reading the Meditations, for example, many are tempted by the thought that Descartes’s meditator must know that he is not dreaming if he is to know that he is sitting by the fire. We find this reasoning tempting and we recognize that we share the meditator’s epistemic shortcomings. An ade quate understanding of skepticism must shed light on the source of this temptation. To put the skeptic in the same boat as the crazed physician denier would render mysterious why anyone has been tempted by skep tical arguments. This pull toward skepticism suggests that deeper and more complex is sues are implicit in the skeptic’s reasoning. As Craig (1990: 107) remarks, it indicates that the skeptic’s argument connects in some way with some
SKEPTICISM AND THE POINT OF KNOWLEDGE
|
197
aspect of the conceptual practice into which we naturally grow up. The skeptic is not merely urging us to adopt different standards that are hardly ever met. Rather, she is arguing that the “new” standards are really the ones we now have, or that they emerge from the ones we have by a natural extension. This is what induces the feeling that we are up against a threat. There is some feature of our concept of knowledge that generates pressure toward skepticism. But what is it about our concept of knowledge that makes skepticism so tempting?
8.3 The Cartesian Requirement for Knowledge In his Meditations, Descartes argues that a general condition of knowing something about the world is that one knows one is not dreaming. When questioning whether he knew that he was sitting by the fire and holding a sheet of paper, Descartes concluded that he must first have evidence or reasons sufficient to rule out the possibility that he was dreaming. If he could not find adequate grounds to exclude this possibility, then he believed he could not know that he was actually sitting by the fire. In the next breath, Descartes considered the possibility that his entire reality was constructed by a powerful demon bent on deceiving him. Again, he concluded that in order to know anything about the world around him, he must have evidence or reasons sufficient to rule out the possibility that he was the victim of sys tematic deception. These two examples indicate that Descartes accepted the following gen eral principle about knowledge: The Cartesian Requirement To know that p one must have evidence or reasons sufficient to rule out all the alternatives to p.6
Admittedly, the Cartesian requirement is not entailed by the claim that one must rule out the possibility that one is dreaming and/or radically deceived in order to know anything about the world. But it would be utterly arbi trary to demand that we must rule out these far-fetched possibilities but not other possibilities to have knowledge. It is far more plausible to think the
6 Being in a position to “rule out” an alternative requires evidence sufficient to entail or imply the denial of the alternative. A possibility q is an alternative to p if and only if it cannot be true that both q and p.
198
|
What’s the Point of Knowledge?
skeptical reasoning in the First Meditation reveals an implicit commitment to the Cartesian requirement.7 Some might prefer to say that knowledge requires one to be certain, on Descartes’s view. The skepticism of the First Meditation is clearly rooted in the demand for certainty: Reason now leads me to think I should hold back my assent from opinions which are not completely certain and indubitable just as carefully as I do from those which are patently false. So, for the purpose of rejecting all my opinions, it will be enough if I find in each of them at least some reason for doubt.
But the quest for certainty is perfectly compatible with the Cartesian re quirement. Certainty requires one to remove all doubt, yet removing all doubt presumably requires one to have evidence or reasons sufficient to rule out all the alternatives to what one claims to know. If I cannot tell whether or not I am dreaming, then I have a reason to doubt that I am ac tually sitting by the fire. The quest for certainty therefore demands that we meet the Cartesian requirement. Anyone who accepts this principle and seeks to show that it can be sat isfied is doomed to fail. It is simply not possible for fallible humans to pos sess the evidence or reasons needed to rule out all the alternatives to what one claims to know. In the case of radical skeptical possibilities, there are no “certain indications” by which we may clearly determine whether we are systematically deceived or not. In the case of mild skeptical possibilities (that are in principle falsifiable), we simply lack the time and resources needed to eliminate all these possible alternatives. In short, we cannot both accept the Cartesian requirement and show that it can be met. Our only hope, as Stroud (1984: 18-19) argues, lies in rejecting this principle. This is not as easy as one might think. After all, we do feel the tug of the skeptic’s reasoning, which provides us with some evidence that the skeptic’s conception of knowledge is the one we already have (I’ll return to this point shortly). Further, it is an everyday procedure for assessing knowledge claims that we are epistemically positioned to rule out alterna tive possibilities. For example, to know that the man approaching my front door is a salesman I need to have evidence or reasons sufficient to rule out the possibility that he has some other occupation (e.g., lawyer, plumber, etc.). Thus, the Cartesian requirement seems like nothing more than an
7 Stroud (19§4), Kaplan (2000), and Lawlor (2013) argue the skeptical conclusion rests on this requirement.
SKEPTICISM AND THE POINT OF KNOWLEDGE
|
199
instance of a familiar procedure with respect to knowledge claims. We are all aware that, even in very ordinary circumstances, we cannot know that p unless we have ruled out certain possibilities that we recognize are incompatible with p. In this way, the skeptic merely invites us to do more of what we do anyway. Suppose we are committed to the Cartesian requirement in grasping the concept of knowledge. Why, then, do people so often speak of knowing things? There are a variety of possible explanations. First, it could be that people reasonably regard their justification for a belief as conclusive when in fact deeper philosophical reflection reveals it is not. Second, it might seem unnecessary or needlessly fussy to insist, in daily fife, that a befief isn’t knowledge when the believer is close enough to meeting the ideal standard (i.e., we exaggerate or engage in “loose talk”). Third, it might be that one who claims to know that p is really making a conditional or elliptical claim, namely: one knows p assuming that some alternative possibilities which are not worth taking seriously are indeed false. These are just a few possible explanations.81 mention them simply to illustrate that the skeptic has several ways to explain why people ordinarily, but mis takenly, speak of knowing.
8.4 The Austinian Requirement for Knowledge To speak of knowing is to speak of something arrived at by removing doubt. On this point, the skeptic is surely correct. Yet it takes only a moment’s reflection to recognize that we rarely insist on the removal of all possible doubts. For example, I know there is milk in my fridge even though a milk thief might have robbed me. You know that you are reading these words even though you might be a brain in a vat. It seems the Cartesian requirement is far too demanding when compared to the circumstances of so-called everyday life. It is even too strict for situations where the standards are typically higher than normal, such as scientific endeavors and courts of law. As Stroud (1984: 57) points out, scientists are not ex pected to attach dream elimination tests to their laboratory reports, and the eyewitness of a crime is not typically expected to rule out the possi bility that she was merely dreaming that she saw the defendant holding a gun. Our inquiries tend to be framed with certain presuppositions, which
8 See Davis (2007) and BonJour (2010) for a detailed set of explanations.
200
|
What’s the Point of Knowledge?
suggests the Cartesian requirement is not a necessary condition for knowl edge about the world. This does not mean that skeptical possibilities are never relevant. We can imagine certain contexts in which it is relevant to insist that someone rules out the possibility that she was dreaming. Consider the following example: If I am lying half-awake in bed early in the morning after a late night and seem to hear someone calling my name from outside the window, I might not be sure whether there really is someone out there or I am only dreaming that I hear the call. (Stroud 1984: 50)
This illustrates that in special circumstances certain kinds of possibilities are relevant to assessing claims to know; thus, a failure to rule them out will imply that one does not know. We find this idea in Austin’s “Other Minds.” Austin invites us to im agine what it takes to know there is a goldfinch in our garden. If asked, “How do you know it’s a goldfinch?” we may reply “From its behavior,” “By its markings,” “By its red face,” and so forth. However, one might still object to the claim that the bird is a goldfinch without disputing any of these facts. One might say, “But that’s not enough: plenty of other birds have red faces and similar markings. What you say doesn’t prove it. For all you know it may be a woodpecker.” Several important points come out here, as Austin (1979: 83-4) notes:
(a) If you say “That’s not enough,” then you must have in mind some more or less definite lack. “To be a goldfinch, besides having a red head it must also have the characteristic eyemarkings”: or “How do you know it isn’t a woodpecker? Woodpeckers have red heads too.” If there is no definite lack, which you are at least prepared to specify on being pressed, then it’s silly (outrageous) just to go on saying “That’s not enough.” (b) Enough is enough: it doesn’t mean everything. Enough means enough to show that (within reason, and for the present intents and purposes) it “can’t” be anything else, there is no room for an alter native, competing, description of it. It does not mean, for example, enough to show it isn’t a stuffed goldfinch. Austin maintains that certain error possibilities are not in question in the ordinary case when one says, “I know it’s a goldfinch.” In daily life,
SKEPTICISM AND THE POINT OF KNOWLEDGE
|
201
“reasonable precautions only are taken.” We do not normally expect our informants to rule out the possibility that they are the victim of some illu sion or deception. Such a demand is legitimate only when we have a “spe cial reason” to think one is the victim of error or deceit. A special reason is, roughly, a reason that we do not have in the normal cases in which people claim perceptual knowledge (Kaplan 2000: 283). But without any special reason to think we are the victims of error or deceit, it would be “outra geous” demand that our informants prove they are not mistaken in these ways. That would demand more than what normally counts as enough. This suggests that we should replace the Cartesian requirement with a weaker principle, such as the following: The Austinian Requirement To know that p one must have evidence or reasons sufficient to rule out the relevant alternative possibilities to p.
Austin recommends a “relevant alternatives” theory of knowledge, which is similar to the view I defended in chapter 3. The rationale for a rele vant alternatives approach is fairly uncontroversial: there seems to be an important connection between our discriminative capacities and knowl edge, coupled with a desire to avoid skepticism. In “Discrimination and Perceptual Knowledge” (1976), Alvin Goldman makes the fink between knowledge and discrimination clear by reflecting on hypothetical cases and our knowledge-attributing behavior. For example, to know that a par ticular object in our visual field is a bam, we must be able to distinguish it from a house, a boat, a car, a tractor, and many other ordinary objects.9 The central idea, following Goldman, is that a person knows that p only if the actual state of affairs in which p is true is distinguishable by him from the relevant possible states of affairs in which p is false (1976: 87-88). We can think of an alternative q as “relevant” if one must rule out q in order to know thatp (Cohen 1988:101). What is it about these alternatives that make them such that we must exclude them? As I’ve argued, a knower must rule out those possibilities that arefitting or reasonable to the members of the epistemic community. In this way, the standards for knowledge are high enough for people who might (in the relevant sense) appeal to the subject’s belief. Relevant alternatives are the kind of possibilities that we
9 Moreover, if we are in an area that is populated by numerous fake bams, it seems to many people that (in these circumstances) we must distinguish the barn-looking object from barn facades in order to know that the building is a barn.
202
|
What’s the Point of Knowledge?
ordinarily take to be the likely alternative possibilities to what the subject claims to know. The idea that knowing p requires the ability to distinguish p from all the relevant alternatives to p has been endorsed by many scholars. I do not claim any originality for this idea. My aim is to motivate this view by providing a function-first explanation of why the concept of knowledge functions in the way outlined by my relevant alternatives account. This has been one of the major obstacles to the idea that knowledge requires the ability to rule out the relevant alternatives. One must say why knowledge requires the ruling out of some alternatives and not others, as well as which alternatives must be ruled out. Before developing this idea in more detail, however, we must first examine a well-known reply to the Austinian view outlined above.
8.5 Stroud’s Reply to Austin If Austin is right, the skeptical conclusion is reached only by violating our ordinary standards and requirements for knowledge. While the skeptic claims we must meet the Cartesian requirement for knowledge, the Austinian requirement is all we really need to satisfy. Thus, Austin’s observations seem to put the skeptic on the low road to her conclusion. But the issue is more complicated than this. In his superb book, Barry Stroud argues that all of Austin’s insights are compatible with the Cartesian requirement for knowledge. Stroud admits “We do not expect, and we would be astonished to find, dream elimination tests appended to laboratory reports” (1984: 57). Further, he says it is perfectly appropriate to credit people with knowledge in everyday life. But Stroud nevertheless recommends the fol lowing skeptical view: although we appropriately credit people with knowl edge all the time, there really is no such thing as knowledge of the external world (or at least a lot less than we suppose). Why think this? According to Stroud, Austin merely reminds us of facts of speech or linguistic usage. What Austin illustrates is that it is often appropriate in the context of everyday life to say that people have knowledge. The important epistemological question, however, is whether this reveals the hue conditions for knowledge. Stroud argues it does not. We must distinguish between two questions that we can ask about what someone says: first, “Is it true?” and second, “Was it appropriately or reasonably said?” As Paul Grice (1989) made clear, all the conditions for warranted assertion can be met even though what is said is not literally true. Even Austin recognizes that it is often appropriate to say that someone knows even when this is false. He writes, “we are often right to say we know even in cases where we turn out subsequently to have been mistaken” (1979: 98).
SKEPTICISM AND THE POINT OF KNOWLEDGE
|
203
With this familiar distinction in place, Stroud creates the space needed to deny that facts about our linguistic practices indicate that the Cartesian requirement is false. On the conception of knowledge that Stroud has in mind, the Austinian requirement is seen as a condition for appropriately ascribing knowledge to oneself or to others, whereas knowledge is actu ally subject to the Cartesian requirement. Stroud writes, the requirement that there must be some “special reason” for thinking a certain possibility might obtain in order for that possibility to be relevant to a particular knowledge-claim would be seen as a requirement on the ap propriate or reasonable assertion of knowledge, but not necessarily as a re quirement on knowledge itself. In the absence of such a “special reason,” one might perhaps be fully justified in saying “I know that p” even though it is not true that one knows that p. (1984: 63)
I do not wish to deny Stroud’s claim that the conditions sufficient for appropriate assertion can be met even though what is said is not literally true—that is uncontroversial. What is controversial is why we should ac cept his claim that the Cartesian requirement is the correct (real, true) standard for knowledge rather than the Austinian requirement. Why should we maintain, as Stroud suggests, that it is inappropriate but nev ertheless true to say that we do not know anything (or almost anything) about the world around us? The mere intelligibility of a logical gap be tween the conditions for warranted assertability and the conditions for truth does nothing to support skepticism. At best, we are left with a stale mate. What could motivate us to think the Cartesian standard is really the right one, in spite of all the linguistic evidence to the contrary? Why should we believe, for instance, that the witness of a crime must be able to rule out the possibility that she is a brain in a vat in order to truly know that the defendant was holding a gun? Stroud provides a twofold reply. First, he says that practical exigency prevents us from applying the Cartesian requirement in real-life situations. We recognize that for “practical” purposes we cannot entertain skeptical challenges, since we have urgent needs for communicating information and it would be too cumbersome to couch virtually all our assertions in probabilistic terms. We therefore ignore the Cartesian requirement in daily fife. In contrast, when we engage in a “philosophical investigation of human knowledge,” we realize that strictly speaking we must be able to meet the Cartesian requirement “if we are to know anything about the world around us” (Stroud 1984: 71).
204
|
What’s the Point of Knowledge?
Other scholars have taken a similar line of reasoning. For example, C. I. Lewis writes: To quibble about such doubts will not, in most cases, be common sense. But we are not trying to weigh the degree of theoretical dubiety which common sense practicality should take account of, but to arrive at an accurate analysis of knowledge. (1946:180, emphasis mine)
Likewise, BonJour claims that meeting the strict standards of knowledge is “obviously intolerable from a practical standpoint.” This is because there is “practical pressure in the direction of reasoning from and acting upon claims whose justification does not fully meet the standard for knowledge” (2010: 75). In particular, BonJour says that knowledge is the norm governing assertion and practical reasoning, but the result is that assertions can almost never properly be made and practical reasoning would be paralyzed. So he says there is strong practical pressure to make unqualified assertions and act upon claims even when “this is not really (fully) warranted by the epistemic situation” (2010:74). This line of reasoning connects to the second part of Stroud’s reply. He argues that our temptation toward skepticism is itself “evidence that the conception of knowledge employed in the [skeptic’s] argument is the very conception we have been operating with all along” (1984: 71). Philosophical reflection is what brings to fight constraints on knowledge which, though they may be unobserved or overruled for pragmatic reasons, apply to knowledge of the world as such. Hume defends a similar view. He claims that skepticism results when reason acts “alone,” that is, when the pressures to act and our natural inclination to believe what one sees are suspended in the pursuit of reflective understanding. Michael Williams (1996: 172) likewise says, “pursuing everyday practical goals ... involves taking many things for granted or accepting them on less than adequate ev idence.” It is because of limited time and resources that “we are bound to act as if we knew all sorts of things that we may not really know at all.” In contrast, when we step back from the procedures that guide our everyday inquiries and ask whether the beliefs that inform them really amount to knowledge, we come to realize that the full conditions for knowing are in fact much stricter than those we settle for in daily life.10
10 An interesting consequence of this idea is that purists (those who claim the standards for knowledge are not shaped by practical considerations) should probably be skeptics, or at least infallibilists. I argue for this view in “Why Purists Should Be Infallibilists” (2019).
SKEPTICISM AND THE POINT OF KNOWLEDGE
|
205
8.6 Kaplan’s Defense of Austin (and Its Limitations) The line of thought just rehearsed has a long history. As Mark Kaplan (2011: 73-74) points out, it has its origins in the early reception of Austin’s work by Paul Grice, Roderick Chisholm, and Stanley Cavell in the 1960s, and it can be found in books on skepticism from the 1980s and 1990s by Barry Stroud, Marie McGinn, and Michael Williams.11 But while this line of argument has a venerable history, it does not offer a forceful critique of Austin’s view. In a series of papers, Kaplan (2000, 2006, 2008, 2011) has been clarifying the epistemological significance of Austin’s work and defending it from these critiques. I am deeply sympathetic with Kaplan’s interpretation of Austin, but I think Kaplan’s argument is limited in impor tant ways. This section will explain why. At the heart of Kaplan’s view is the following idea: once we take all of the evidence into account, our temptation toward skepticism is insufficient to support a skeptical bottom line. According to Kaplan, the allure of the skeptic’s reasoning does provide us with some evidence for the view that we lack knowledge of the world around us. But appealing to our tempta tion toward skepticism is not enough to show that our everyday concept of knowledge is equipped with the Cartesian requirement. After all, we are equally (if not more) tempted to say that we know a lot—hence the par adox. Thus, the evidence provided by our initial reaction to the skeptic’s argument is, at best, equivocal. Moreover, our ordinary practice of epi stemic evaluation is on the whole quite hostile to the skeptic’s conclusion. In asking us to focus on the temptation we feel toward skepticism, but not the pull we feel away from it, Stroud is basically asking us to focus on aspects that favor skepticism at the exclusion of the rest. According to Kaplan (2000: 301), this violates a fundamental principle of sound in quiry: evaluate all the relevant evidence. In reply, Stroud might argue that the legitimacy of the Cartesian re quirement is not to be judged by reflecting on what we would say in or dinary circumstances. This is because he is doubtful about any attempt to arrive at conclusions about the state of knowledge on the basis of facts about how we ordinarily speak and assess knowledge claims:
11 See Grice (1961), Chisholm (1964: 1), Cavell (1965: 216-217), Stroud (1984: ch. 2), McGinn (1989: 62), and Williams (1996: 147-148).
206
|
What’s the Point of Knowledge?
As long as it is even intelligible to suppose that there is a logical gap be tween the fulfilment of the conditions for appropriately making and assessing assertions of knowledge on the one hand, and the fulfilment of the conditions for the truth of those assertions on the other, evidence from usage or from our practice will not establish a conclusion about the conditions of knowledge. (Stroud 1984: 64)
But this move is deeply problematic. In claiming that facts about ordinary usage do not reveal the true conditions of knowledge, Stroud isolates the question about the utility of attributing knowledge from the conditions for knowledge. If, however, we cannot establish any facts about the conditions of knowing from facts about the way we ordinarily speak of knowing, then how are we to determine the nature ofknowledge? To what else could we legitimately appeal in order to decide on the correctness of Stroud’s view? By insulating the conditions of knowing from all the facts of ordinary usage, Stroud “insulate[s] his account of the nature of knowledge from the apparently disconfirmatory evidence provided by our ordinary practice of knowledge attribution” (Kaplan 2000: 285). But to what else might an epistemology be true? Stroud maintains that knowledge is to be judged by “reflection on what we would say were we free from the pressure of practical exigency” (1984: 71). This reply is unconvincing. It might very well be true, as Stroud claims, that our ordinary standards for appropriate knowledge at tribution are less stringent as a result of practical pressures; but it does not follow that the true standards of knowledge are those revealed by philo sophical reflection, once all practical considerations have been set aside. Why not instead think the nature of knowledge is itself sensitive to these practical considerations? Why should we think our epistemology must be truer to the reflections of philosophy than the conditions of practical life? As I’ve been arguing throughout this book, the circumstances of practical life are precisely what give rise to our practice of epistemic evaluation in the first place. On my pragmatic account of knowledge, we may grant all of the fol lowing points to Stroud (and the skeptic): our ordinary practice of knowl edge attribution is shaped by prudential concerns; this practice would be different if we were unfettered by practical considerations; and our knowl edge attributions would reflect the Cartesian requirement if the search for truth were free of practical constraints. However, none of this goes any distance to showing that our ordinary knowledge attributions are therefore false. We need not assume, as Stroud clearly does, that our sensitivity to
SKEPTICISM AND THE POINT OF KNOWLEDGE
|
207
practical considerations leads our ordinary practice of epistemic evalua tion into error.12 While I think this is correct as far as it goes, this argument does not go far enough. I will outline four limitations and then attempt to resolve them. First, this interpretation of Austin leaves mysterious why we are tempted toward skepticism in the first place. On Kaplan’s view, the skeptic claims that knowledge requires something that it does not in fact require. But if this were true, why have we not simply dismissed the skeptic long ago? Lawlor raises a similar objection against Kaplan. According to Lawlor, Kaplan maintains that “nothing about our ordinary practice reveals a com mitment” to skepticism (2013: 193). Now, this is not entirely fair because Kaplan acknowledges the pull of the skeptic’s reasoning. But Lawlor is right that Kaplan leaves unexplained why we would be tempted by skep ticism at all.13 Both Kaplan and Austin simply insist on the certainty of everyday common sense, but this provides no real insight into how we (or the skeptic) go astray. Second, by insisting on the propriety of our ordinary practice of ep istemic evaluation, Kaplan implies that the skeptic imposes artificially strict conditions on knowing. This puts the skeptic on the low road to her conclusion. As I’ve argued, however, we cannot banish the skeptic so easily. Even Austin believed that what counts as “enough” justifica tion for knowledge is determined by our “present intents and purposes” (1979: 84). So why not think we require more justification according to the skeptic’s own intents and purposes, which have nothing to do with practical decision-making? Perhaps stricter than usual standards are sometimes called for. Third, this interpretation of Austin does not explain why our concept of knowledge would be equipped with the Austinian requirement rather than the Cartesian requirement. What considerations can explain why our con cept has certain standards and requirements rather than others?
12 Stroud does provide an extended analogy involving plane-spotters to explain why the skeptic would think the Cartesian requirement is a condition for knowledge (1984: 67-74), but Kaplan (2000: 289-293) convincingly discredits this argument by analogy. 13 At one point, Kaplan (2000: 294) suggests we fall for skeptical arguments because we are vulnerable to cognitive errors, but he does not elaborate on this idea. More recently, Hawthorne (2004: 164) and Williamson (2005: 226) provide a detailed defense of the claim that skepticism’s allure is a result of certain cognitive errors, specifically the “availability heuristic.” But Nagel (2010a) explains why this strategy is highly problematic.
208
|
What’s the Point of Knowledge?
Fourth, even if Austin and Kaplan are right to combat skepticism by pointing to our ordinary practice of epistemic evaluation, Kaplan mistak enly thinks that “what it is appropriate to say and do in ordinary life is the only thing available for an epistemology to be true to” (Kaplan 2000: 301, emphasis mine). This exaggerates the importance of facts about how we ordinarily speak and assess knowledge claims. Kaplan is right to criticize Stroud and other skeptics for turning their back on data from ordinary usage, but this is not the only thing to which an epistemology must be true. As I have been arguing in this book, we should also consider the pur pose of our epistemically evaluative practices. Function-first epistemology thus provides an important and underappreciated item for our toolbox; it offers an independent source of evidence to assess the requirements for knowledge. In the next three sections, I will appeal to considerations about the func tion of epistemic evaluation to help explain the rational intelligibility of skeptical doubt. It is by reflecting on the purpose of knowledge that we get crucial leverage to explain why we are tempted toward skepticism, why our concept of knowledge is equipped with the Austinian requirement, and why linguistic usage is not the only thing to which an epistemology must be true.
8.7 Skepticism and the Point of Knowledge Following Craig, I have argued that the concept of knowledge is rooted in some basic human needs and interests—particularly our need to pool information and share it with others. To achieve these goals, we must identify informants who are in an epistemic position sufficiently strong to allow them to fittingly serve as sources of actionable information for others. This calls for a high-quality epistemic position. But how high must the standard be set? What does it take to count as a good informant in the objectivized sense? If knowledge required absolute objectivi zation, meaning the informant’s evidence must be good enough to guarantee the right answer, then skepticism would promptly follow. Craig himself some times writes as though he endorses this view. At one point he says, All this is going to edge us towards the idea of someone who is a reliable informant as to whether p whatever the particular circumstances of the in quirer, whatever penalties hang over him, and whatever his attitude to them. (1990: 91)
SKEPTICISM AND THE POINT OF KNOWLEDGE
|
209
On this conception, a good informant must be reliable enough to suit the purposes of any inquirer whatsoever, including radically skeptical inquirers. Does our pragmatic account suggest that the standards for knowledge should be raised to skeptical heights? Our function-first approach can help us resist skepticism by explaining why there is downward pressure on the epistemic standards for knowl edge. To illustrate how, we must return to our earlier distinction between the various types of skepticism (see §8.1). I will first show that my prag matic approach can allay worries generated by nonfalsifiable skeptical hypotheses (I’ll focus specifically on radical skeptical possibilities); then I will argue that my view alleviates doubts generated by falsifiable skep tical hypotheses. Much of what I’ll say is indebted to Craig’s own diag nosis of skepticism, but my view goes beyond his in some important ways. For example, Craig focuses solely on radical skepticism, whereas I will extend this approach to nonradical forms of skepticism. In this way, my analysis has wider scope. I will also combine my function-first approach with a theory of relevant alternatives to provide a plausible explanation for what makes a particular alternative relevant. Unlike most relevant alter native theories, which do not provide an explanation for why possibilities are classified this way, my view gives a plausible practical story about why certain possibilities need not be eliminated. Let’s start with radical skepticism. The basic strategy of radical skep ticism is to invent a hypothesis with two properties: first, it would make no noticeable difference to us if it were true; second, if it were true, virtu ally all our present beliefs would be false. The radical skeptic argues that, given the second property, we must either show that the hypothesis is false or abandon virtually all claims to knowledge. But given the first property we cannot show that it is false. Therefore, we must abandon virtually all claims to knowledge.14 The skeptical threat arises if we think the principle of objectiviza tion (as outlined in §2.2) will result in a standard so high that in order to qualify as a good informant one must be reliable enough to satisfy the demands of the skeptic. However, most inquirers are not concerned with finding informants that can distinguish radical skeptical scenarios from nonskeptical scenarios. Why not? Presumably because no such people exist. We are all epistemically handicapped in this way, so there can be no practical purpose in seeking out informants who can distinguish between
14 Craig (1990: 104) characterizes skepticism this way.
210
|
What’s the Point of Knowledge?
radical skeptical situations and ordinary situations. This is not to suggest that it would make no difference to us if we found out that we were rad ically deceived. Surely our interpersonal relationships would be much less meaningful if we were just brains in vats. Still, there is no compel ling reason to worry about such hypotheses because we are not (and pre sumably never will be) epistemically positioned to distinguish between these states of affairs. Whatever differences there are between these two situations, they are not ones that can make any difference to us as inquirers because the two scenarios are epistemically indistinguishable. From the point of view of advancing inquiry, there can be no practical advantage in trying to distinguish such situations because it exceeds our capacities to do so. We are therefore exempt from paying attention to such possibilities. It is by reflecting on the practical role of epistemic evaluation that we can see what is bad about standards so high they lead us to skepticism. It is not that skepticism is itself an unacceptable result, but rather that skeptical standards do not draw the distinctions that are important to us (see also Heller 1999: 119). Suppose I am inquiring into the whereabouts of my cat Garfield. It might be that neither you nor I (nor anyone else) can rule out the possibility that we are brains in vats, but your epistemic position might still be better than mine in multiple ways. Perhaps you have already thor oughly checked in the kitchen while I have not. Or perhaps my wife told you that Garfield is at the vet. These are distinctions that we care about, but they get lost if too many possibilities are selected as ones we must rule out. Now, if we had some other term of epistemic praise to mark the epi stemic conditions or properties we care about, then skepticism wouldn’t be a problem. No one would “know” anything, but knowledge wouldn’t be our word for identifying people who have the properties that matter to us. As it happens, however, “knowledge” is our word for that. Skepticism pulls against our pragmatic approach to knowledge because expecting informants to satisfy the Cartesian requirement would frustrate our communal epistemic practices. To deny knowledge to a potential in formant on this basis would suggest she is not in a strong enough epistemic position to appropriately contribute to the stock of information on which others can draw for their own projects and purposes. In everyday life, how ever, nobody considers whether a potential informant can rule out radical skeptical possibilities in order to qualify as a good informant on some issue.15 This puts downward pressure on the epistemic standards required
15 Something similar can be said for the debate about whether or not our sensory experiences are reliable guides to the world beyond the veil of perception. Having access only to our ideas and
SKEPTICISM AND THE POINT OF KNOWLEDGE
|
211
for knowledge. A function-first account seeks to explain how, given some set of facts about humans, their environment, and their interests or aims, we have a need that our knowledge concept satisfies. The skeptical standard is not one that would adequately serve our needs and interests, so we should not expect the skeptic’s demand for infallibility to exert much pressure on shaping our concept of knowledge. Are there any practical factors that might push the concept of knowl edge to radically skeptical heights? One might argue there are compelling practical reasons for wishing to avoid falsehoods; and this might lead to skepticism because widespread doubt is perhaps the best way to avoid false beliefs. (Recall that the Meditations opens with the thoroughly practical intention of avoiding falsehood.) In addition, we have a tendency in our everyday practice of knowledge ascription to improve the reliability of our information by eliminating error possibilities to what we claim to know. In this way, the skeptic might be extrapolating on what is encouraged when seeking information; thus, she is not obviously following an unacceptable practice. Craig recognizes the danger of slipping into skepticism. He writes, We therefore adopt a strategy designed to succeed whatever ... is this case: set a standard high enough to face the worst with. . . . The possi bility of the worst compels observance of the highest standard, and this is what the concept of knowledge certifies ... [However], I am now in danger of making things much too easy for myself, for what these superlatives (“worst” and “highest”) mean is still unclear, and they may mislead. The danger is of slipping into saying that for knowledge, given that the concept is governed by objectivization in the way just envisaged, “absolute” objec tivization is called for, meaning by that the demand that the informant be certain to have the right answer in any conceivable world, including those “skeptical worlds” of demons and brains in vats. (1990: 108-109)
The worry is that the principle of objectivization has the danger of pushing knowledge to the limit of theoretical possibility, which means that every conceivable way in which the belief that p could be wrong must be taken into consideration. Let your paranoid fantasies rip.16
impressions (in Hume’s sense), we are left with a gap between ourselves and the world around us. If, however, an imperceptible “reality” is forever inaccessible to us, what concern can it be of ours? See Stroud (1984: 34). 161 take this phrase from Lewis (1996: 549).
212
|
What’s the Point of Knowledge ?
But objectivization need not bring skepticism in its wake. No practical motives could lead us to prefer imposing requirements that would keep out falsehoods by rejecting everything. Thus, the search for practical factors that might push us into the arms of the skeptic is unpromising. Our con cept of knowledge is needed to convey information about which claims we take ourselves to have sufficient reason to believe and act on (it is pre cisely for this reason that we have a knowledge concept). The skeptic’s argument, however, invites us to abandon the very purpose of speaking of knowledge (Craig 1990: 119). If knowledge required us to satisfy the skeptic, it would be surprising why such a concept should have entered into our vocabulary at all. In daily life we do not expect knowers to eliminate radical skeptical possibilities because there is no practical advantage (from the point of view of advancing inquiry) in trying to distinguish between scenarios that are epistemically indistinguishable to us. This explanation also applies to moderate skeptical possibilities (which have a narrower scope but are equally unfalsifiable). However, this explanation is inadequate for cases where we can in principle tell whether an error possibility obtains or not. Consider the following example (adapted from Lewis 1996): Garfield is missing again. To know whether Garfield is in my kitchen, I must look around. When I look around the kitchen, I thereby eliminate various possibilities in which Garfield is in the kitchen (under the table, behind the door, etc.). Had I not eliminated these possibilities, I would not qualify as knowing that Garfield is not in the kitchen. Nevertheless, some possibilities are rightly ignored; for instance, I am not required to look for him in closed and locked drawer at the top of the cabinet.
In this case, the possibility that Garfield has miraculously sealed him self in a locked drawer cannot be disqualified by appealing to the idea that there is no test we can perform to successfully tell whether he is in a drawer or not. After all, we can simply open the drawer and look. Thus, we cannot ignore this possibility on the grounds that it lies beyond the limits of our discriminatory capacities. Does function-first epistemology have any power to resist these falsifiable error possibilities? Following Lewis, we can distinguish between possibilities that are eliminated by our evidence and possibilities that are properly ignored. In general, whenever a subject S claims to know thatp, her evidence will elim inate some not-/? possibilities and others will be properly ignored (Lewis 1996: 562). More specifically, there are four types of error possibilities:
SKEPTICISM AND THE POINT OF KNOWLEDGE
|
213
(a) Falsifiable, relevant alternatives', possibilities that can be eliminated but cannot be properly ignored (b) Falsifiable, irrelevant alternatives', possibilities that can either be eliminated or ignored (c) Nonfalsifiable, irrelevant alternatives: possibilities that cannot be eliminated, but can be properly ignored (d) Nonfalsifiable, relevant alternatives: possibilities that cannot be eliminated and cannot be properly ignored
In daily life, the possibility that Garfield is sitting on the floor in my kitchen falls into category (a); the possibility that Garfield has miracu lously shut himself in a drawer falls into category (b); and the possibility that Garfield is on the table but has been made invisible by a deceiving demon falls into class (c). (Admittedly, there may be special contexts in which skeptical possibilities are not properly ignored. I will discuss this idea shortly.) Possibilities can either be relevant or irrelevant. Radical skeptical possibilities are typically irrelevant because they make no noticeable dif ference to us and therefore work against the purpose of having a con cept of knowledge, which is rooted in our need to distinguish reliable informants from unreliable informants. But what about the falsifiable error possibilities that we do not bother eliminating in daily life? What explains why I have to look behind the table but not inside the locked drawer in order to know whether Garfield is in the kitchen? To explain the irrelevance of certain falsifiable error possibilities, the mechanism to which we should appeal is the (un)likelihood that such possibilities obtain (coupled with what’s at stake). I need not check the locked drawer in my kitchen in order to know that Garfield is not in the room because the likelihood that Garfield has managed to seal himself inside that drawer is far too low—too unlikely to have occurred—for it to be necessary to rule out in order for me to qualify as a reliable informant on that matter. (Things may change if the stakes are incredibly high.) If we always required our informants to rule out improbable possibilities, then very few people would count as reliable. We want inquirers to be able to rule out certain error possibilities that we ourselves are not in a position to rule out, but we do not require informants to pay the informa tional costs needed to rule out every possibility (or even every falsifiable possibility). We want people to have a good success ratio, but how good a success ratio is required to meet the commonly shared standard for knowledge? As
214
|
What’s the Point of Knowledge?
suggested in chapters 3 and 4, we are in no position to demand infallibility. As Sosa writes, A concept of knowledge requiring a perfect success ratio would not be very discriminating, and would not help us to keep track of the facts regarding epistemic dependability, our own or others’. Nor would that concept of per fect knowledge aid intercommunication of such facts amongst members of the group. (1991: 275)
To insist that people must satisfy the Cartesian requirement by ruling out all possibilities would exclude a vast number of cases where a particular informant is perfectly suitable for the purposes of many inquirers. Our withholdings would suggest to others that these highly reliable informants do not have sufficiently good information on which we can base our beliefs and actions. However, there is a clear benefit to treating individuals as good informants if they are likely enough to be correct for our usual purposes. Thus, we almost never require our informants to have the highest quality of epistemic position. A typical situation for an inquirer is that she could improve her truth ratio somewhat without giving up a large class of beliefs; however, this would take time and effort. The cost of doing so might be too high relative to other activities (including other activities of inquiry), so it might not be sensible to expend such effort in a given case. The investment of ef fort into inquiry typically depends on what is at stake given our interests. In the bank cases (§4.4), for instance, Keith seeks to confirm the bank’s opening hours when financial ruin is at stake, but not when nothing sig nificant turns on promptly depositing his paycheck. Similarly, Mary and John (§5.2) seek information from the airline agent if they have to meet an important business client during a layover in Chicago, but they gladly rely on the printed itinerary in other circumstances. Thus, I am suggesting a two-pronged reply to the skeptic within one general framework. On the one hand, we have strong inductive evidence against the likelihood of error in many cases involving falsifiable alterna tive possibilities. This provides us with sufficient justification to qualify as knowing in many situations. On the other hand, we cannot extend this ap proach to radical skeptical scenarios because we have no evidence against such alternatives, so we cannot claim that our evidence is good enough to rule them out. Rather, I maintain there can be no practical advantage in trying to distinguish between two scenarios that are epistemically indis tinguishable, so we should not expect the skeptic’s demand to exert much pressure on shaping our concept ofknowledge.
SKEPTICISM AND THE POINT OF KNOWLEDGE
|
215
8.8 No “Hard Boundary” Does this mean skeptical challenges are never legitimate? I hesitate to go that far. On the one hand, the principle of objectivization pushes us to ac cept stricter standards that might logically end at skepticism. On the other hand, there are practical considerations that put downward pressure on our epistemic standards. According to Craig, this tension creates an “area of indeterminacy” that “hosts the controversy about skepticism” (1990: 117). Further, we cannot eliminate the tension because neither side is doing something disallowed; on the contrary, each reflects a view that certain aspects of the concept of knowledge encourages. I want to highlight the radical nature of this proposal. At present we are confronting the following question: will our pragmatic account put an upper boundary on the concept of knowledge, or will it make room for skeptical doubts? According to the skeptic, there are practical factors that push the concept of knowledge to skeptical heights. According to the nonskeptical invariantist, practical considerations will put a hard upper bound on the epistemic standards for knowledge; thus, anyone who insists on confronting a knowledge claim with skeptical fantasies would be misusing the concept of knowledge. According to the contextualist, there might be purposes and circumstances that make skeptical doubts appro priate in some contexts, but this needn’t threaten our everyday knowledge claims. Which view is correct? I reject all three options. We should instead regard our practically explicated concept as open-ended at this point. Put differently, we should not expect our knowledge concept to be loaded with all, strictly all, the conditions needed to determine some “hard boundary” that will allow us to definitively settle the issue (see Craig 1990: 114). On the one hand, the process of objectivization pushes us closer and closer to skepticism, which makes this view tempting. On the other hand, to fully endorse skep ticism would mark the end of any contact with the practical requirements that gave rise to our concept of knowledge. We can acknowledge both these facts without insisting that our concept of knowledge acquires a hard boundary that would allow us decide whether skeptical challenges are permissible or not. As Craig (1990: 116-117) puts it, “the operative fact is precisely that nothing happens here, so we neither have a positive boundary nor the positive absence of one. The resultant area of indetermi nacy hosts the controversy about skepticism.” This fits with my rejection of both contextualism and invariantism in the previous chapter. My proposal also leaves open the possibility that there
216
|
What’s the Point of Knowledge?
may be contexts in which it is appropriate to mount skeptical challenges. The present argument only shows that such contexts are not common in the circumstances of daily life, since there are few (if any) practical contexts in which a good informant would have to meet the Cartesian requirement to qualify as reliable enough.17 If there are circumstances in which skeptical challenges are appropriate, which are they? In the next section, I will suggest that something about the nature of the philosophical investigation of human knowledge pushes us toward skepticism. We thus return to where we began this chapter, namely the clash between ordinary life and the attitude to which we are led by philosophical reflection. I will explain our temptation toward skepticism in some contexts (but not others) in the following way: what we may take for granted, and what needs to be ruled out, in order to qualify as knowing that p is inquiry-bound. For instance, Descartes’s methodological doubt is tied to an unusual form of inquiry dedicated to eliminating all falsehoods from our belief set. Outside of this investigative context, however, prac tical considerations render skeptical doubts inappropriate. Skepticism is contextually constrained because it is connected with a special type of inquiry.
8.9 Pure Inquiry and Practical Inquiry In Descartes: The Project of Pure Inquiry, Bernard Williams introduces the idea of a “pure inquirer” as someone who engages in the “project of pure inquiry.” This project, like other projects of inquiry, is concerned with trying to find the truth (and avoiding error); however, unlike eve ryday inquiry, this task is undertaken “from the very beginning.” In eve ryday contexts we presuppose bits of knowledge about the world, but in the context of pure inquiry we do not make these same presuppositions. According to Williams, A philosophical inquiry differs from others ... in being a specially radical kind of inquiry, and, as a consequence of that, peculiarly free both from assumptions and from certain sorts of constraints which apply in general to the search for knowledge. (Williams 1978: 20)
17 The lottery worry is an exception, for the reasons mentioned in chapter 3.
SKEPTICISM AND THE POINT OF KNOWLEDGE
|
217
This radical kind of inquiry gives rise to skeptical worries because a pure inquirer will not assent to any belief in which he detects the possibility of doubt. Pure inquiry may be contrasted with “practical inquiry.” Williams no where uses this term, but it will be useful for our purposes. As mentioned earlier, Descartes alludes to the notion of practical inquiry in the fourth part of the Discourse'. so far as practical life is concerned, it is sometimes necessary to follow opinions which one knows to be very uncertain, just as though they were indubitable . . . but because I wanted to devote myself solely to the search for truth, I thought it was necessary that I should do just the opposite, and that I should reject, just as though it were absolutely false, everything in which I could imagine the slightest doubt, so as to see whether after that anything remained in my belief which was entirely indubitable. (VI31-32, HR 100-101)
While both the pure inquirer and the practical inquirer want to maxi mize their true beliefs, the pure inquirer goes to great lengths to avoid admitting any false beliefs into her belief set. For example, the meditator of Descartes’s Meditations begins his reflections by proposing to follow one distinctive rule: reject all ordinary knowledge and trust only what cannot be doubted. In contrast, the practical inquirer will almost certainly have false beliefs because she sticks to methods of belief acquisition that are generally reliable but fallible. As I’ve argued, one of the easiest and most efficient ways to acquire a true belief on some issue is to ask someone who is likely to know the answer. If someone else is better positioned than me to know whether p, then it is prudent to rely on her testimony. This situation is captured in the story of the truth-seeking inquirer. She wants true beliefs but her inquiry is constrained by practical considerations such as time commitment, the acces sibility and importance of information, the need to act, her dependence on others, and the need to satisfy other goals. For the pure inquirer, however, none of these considerations apply. The pure inquirer engages in a form of inquiry that is distinctively philosophical, so the considerations that ration ally weigh on us in practical contexts of inquiry bear no relevance. These considerations are laid aside in the search for a method whose correct appli cation guarantees truth. This is not to say that practical inquiry is concerned exclusively with beliefs that are instrumentally valuable (i.e., useful for some practical
218
|
What’s the Point of Knowledge?
purpose). The project of practical inquiry can also be directed toward truths that satisfy our sheer intellectual curiosity, such as wondering why the dinosaurs went extinct. Practical inquiry simply has to do with bal ancing our need for truth (and the avoidance of error) with the constraints and limitations that we fallible humans encounter in daily life. In contrast, these concerns are set aside when we engage in pure inquiry. Pure inquiry is not only different from everyday inquiry; it is also a mode of inquiry that should not be imported to practical contexts. As Descartes says, we are to make use of this doubt only when we are engaged in contemplating the truth. For, as regards the conduct of our life, we are frequently obliged to follow opinions which are merely probable, because the opportunities for action would in most cases pass away before we could deliver ourselves from our doubts. (HR 219-220)
Bernard Williams also emphasizes this point. He says the method of doubt in the Meditations is an “instrument of reflective inquiry . . . not to be brought into practical matters: equally, no values drawn from those matters affect the inquiry” (1978: 32). In short, radical doubt has a far less significant role to play in daily practical life. Consequently, any problems that arise within the context of pure inquiry will be problems ofphilosophy. I think we should grant that skeptical doubts are sometimes apt, but only in the philosophical project of pure inquiry. Many fallibilists are unwilling to countenance even this limited victory to the skeptic. In On Certainty, Wittgenstein claims there is no context in which we can ra tionally doubt the “obvious truisms” of common sense (see §155, §231, §234, §613). And on Kaplan’s interpretation of Austin, we cannot legit imately keep “two sets of books” (2011: 60 n. 8); rather, what we say when doing epistemology must be true to what we say and do in daily life. However, any person who takes this line of argument must provide a more convincing account of the source of philosophical skepticism and also in dicate precisely where the skeptic’s argument goes wrong. I doubt that any such view is in the offing. This proposal raises further topics that are worth exploring, such as the mechanisms by which we move from one form of inquiry to another, as well as the worthwhileness of adopting the perspective of the pure in quirer. I will conclude with some brief remarks about these questions, but more investigation is called for.
SKEPTICISM AND THE POINT OF KNOWLEDGE
|
219
If David Lewis (1996) is right, then any philosophical inquiry into knowledge is inevitably skeptical. Many doubt the skeptic wins so easily, and I include myself in this camp (see Hannon 2017b for an overview). If skepticism were the inevitable outcome of the quest for a theory of knowl edge, than my pragmatic account would also lead to skepticism—but it doesn’t. My account explains why the skeptic’s argument is plausible and yet not a threat to ordinary knowledge. Further, my account can be used to explain why it is much more difficult to take skeptical doubts seriously in some contexts rather than others. In a criminal trial, for example, we cannot easily get people to take seriously radical skeptical possibilities be cause courtrooms are decidedly practical inquiries. In other settings, such as a philosophy classroom, the focus on practical life is often less clear. There may also be circumstances in which it is not clear which type of inquiry we are engaging in, and the result will be that people feel the pull of skeptical reasoning (on the one hand), while also wanting to reject the skeptic’s conclusion (on the other hand). We puzzle over skepticism be cause we are confused about the nature of our inquiry. What about the significance of the project of pure inquiry? It might be true that if one engages in pure inquiry, then concerns other than guaranteeing truth and eliminating falsehood drop out of the picture. However, an important question is whether engaging in this project is it self worthwhile. Why should we care about it? What is the wellspring of such philosophizing? We should not view the project of pure inquiry as a mere pastime invented by epistemologists who set forth the supererogatory ideal that knowledge requires infallibility.18 Pure inquiry can be defended as a worthwhile en deavor in several ways. For example, this project may be valuable because meeting the demands to succeed at pure inquiry will presumably enhance the epistemic quality of our target befief (Sosa 2009: 58). It may also re veal better methods of inquiry even when we are constrained by practical considerations. Moreover, suspending common assumptions might reveal certain truths that previously went unnoticed or improve our reasoning and intellectual powers. Finally, engaging in this project might aid our under standing of the nature, conditions, and extent of human knowledge (Sosa 2009: 50).
18 Recall that the pyrrhonists sought the suspension of belief to further the practical goal of achieving ataraxia (Mates 1996: 92-93).
220
|
What’s the Point of Knowledge?
8.10 Concluding Remarks This chapter has explored the clash between two perspectives: our eve ryday outlook and the attitude to which we are led by philosophical reflec tion. I have argued that skepticism is a natural outcome of philosophizing, but this does not mean that we must live with the truth of skepticism. Rather, I have attempted to reconcile these perspectives in a way that honors the insights of them both. The position I have outlined leaves room for the idea that there may be limited contexts in which it is appro priate to make skeptical challenges; moreover, I have tried to explain why we should expect there to be such contexts in the first place. A virtue of my proposal is that it does not force us to choose either the side of the skeptic or the side of common sense, thereby ignoring our inclination to find both views tempting. While the skeptic’s demands may seem out of the ordinary, this does not make them inappropriate in certain contexts. As Michael Williams puts it, “when unusual questions are on the table, unusual possibilities may have to be considered” (1996: 8). What we can legitimately presuppose when engaging in inquiry depends on the nature and scope of our inquiry.
SKEPTICISM AND THE POINT OF KNOWLEDGE
|
221
chapter 9 | What’s the Point of Understanding?
Any fool can know, the point is to understand. —ALBERT EINSTEIN
on what knowledge is and why it matters. But knowledge is not the only cognitive achievement at which humans aim and it might not be the most valuable. We also seek to under stand a great many things. Some of these things are perplexing and mag nificent, such as the evolution of fife and the nature of the cosmos. We may want to understand other languages, how to play a musical instrument, why the ancient Egyptians mummified their dead, and much more. The things we aim to understand range from the awe-inspiring to the utterly trivial. And yet the epistemic state of understanding is poorly understood. This chapter will extend my function-first approach to the topic of human understanding to see whether new light can be thrown on this cognitive achievement. Some questions I will ask are these: Why do humans think and speak of understanding? What role (or roles) does this concept play in human life? What needs does it answer to? I intend to answer these questions and, in doing so, make progress on the nature and value of un derstanding, the role of explanation in understanding, epistemic luck, and the relationship between understanding and knowledge. our investigation has centered
9.1 Understanding and Knowing Philosophical interest in understanding has expanded in recent years. Those familiar with this body of literature will know that philosophers exploring the nature of understanding have largely done so by comparison
with knowledge. This isn’t all that surprising. Knowledge has been a focal point of modem epistemology and the language of “knowing” and is closely related to the language of “understanding.” For example, “knowing” and “understanding” come in the same linguistic forms: just as one can know-that, know-how, know-who, know-what, know-where, know-when, and know-why, one can also understand-that, understandhow, understand-who, and so forth. Moreover, “knows” and “understands” are sometimes interchangeable; for instance, “I know that two plus two equals four” seems to express the same thought as “I understand that two plus two equals four.” Knowing and understanding seem like closely re lated cognitive achievements. But despite these commonalities, ordinary parlance also points to some important differences between knowing and understanding. For one thing, we often contrast knowing and understanding. Here are two examples: I know that my toaster works, but I don’t understand how it works; I know that some people enjoy the film Battlefield Earth, but I don’t understand why. More generally, we often contrast knowing that so-and-so with un derstanding how or why so-and-so.1 Further, it is commonly said that we know more than we understand. As Einstein put it, “Any fool can know, the point is to understand.” These contrasts do not rule out the possibility that understanding is ultimately reducible to knowledge (I’ll return to this idea in §9.7), but they do suggest that understanding might require a level of intellectual sophistication not necessarily demanded by knowledge. This way of talking isn’t peculiar to English. As Alison Hills (2016) reports, many languages draw a similar distinction (e.g., French, German, Russian, Hebrew, Danish, and Irish). This gives some prima facie support to the idea that knowing and understanding are distinct. Presumably, we wouldn’t find this distinction in many languages if knowledge and under standing were simply interchangeable or referred to the same epistemic concept. As a consequence of comparing knowing and understanding, philosophers have tended to concentrate on the following questions: Is understanding factive? Is it immune to epistemic luck? Is it transparent? Does it come in degrees? Is it transmittable via testimony? Is it reducible to knowledge? This cluster of questions has largely set the research agenda for the study of understanding in epistemology. Philosophers have sought
11 am not suggesting the distinction between knowing that and understanding how or why reveals that understanding is different from knowledge (see Brogaard 2005). My point is just that we often contrast these notions.
what’s the point of understanding?
|
223
to improve our understanding of understanding by comparing it with what we know about knowledge. Unsurprisingly, philosophers disagree about the answers to these questions. It is widely accepted that knowledge is factive, but it is less clear whether understanding is factive.2 While pretty much everyone thinks cer tain types of epistemic luck undermine knowledge, there is little agreement about whether understanding is immune to the same forms of luck.3 Almost everyone agrees that understanding comes in degrees, but there is disagree ment about whether knowledge admits of degrees.4 It has also been said that understanding, unlike knowledge, is transparent; but this, too, is disputed.5 Finally, it is debatable whether understanding is transmittable via testimony and yet testimony is a basic way in which knowledge gets around.6 Interesting though these questions may be, I want to temporarily set them aside (I’ll return to them later). These debates are focused on getting the conditions under which one has understanding just right, leaving aside for the most part questions about the role of understanding in our lives more broadly. Let me explain. Suppose, however optimistically, that these conceptual issues were one day resolved. This would certainly be a considerable achievement, but answering these questions would not tell us why our epistemic concepts have (or lack) these features. For example, let’s assume that understanding turned out to be nonfactive and compatible with epistemic luck. We might ask: why do we have a concept demarcated by those conditions? This question does not naturally arise in the current debates on understanding. Further, the usual attempts to analyze understanding tell us little, if any thing, about why this concept might differ from nearby epistemic notions, such as knowledge. Assume, for instance, that truth is a necessary
2 Zagzebski (2001), Elgin (2007), and Riggs (2009) argue that understanding is not factive, while Pritchard (2009), Strevens (2013), Greco (2014), and Hills (2016) claim that at least some types of understanding, such as understanding-why, are factive. 3 Zagzebski (2001), Kvanvig (2003), Morris (2012), and Hills (2016) argue that understanding is immune to knowledge-undermining luck. Pritchard (2009) says understanding is incompatible with Gettier-type luck but not environmental luck. Brogaard (2005), Grimm (2006), Khalifa (2013), and Greco (2014) argue that understanding is incompatible with the same type of epistemic luck as knowledge. 4 Kvanvig (2003) and Hills (2016) argue that understanding, but not propositional knowledge, admits of degrees, whereas Hetherington (2001) says knowledge-that is also gradable. Brogaard (2005) and Riaz (2015) argue that knowledge-why is gradable. 5 Zagzebski (2001) says understanding is transparent in the sense that there is no gap between seeming to understand and understanding. Trout (2002) and Hills (2016) deny this. 6 Pritchard (2009) and Hills (2016) argue that understanding cannot be transmitted by testimony, but Sliwa (2015) and Boyd (2017) deny this.
224
|
What’s the Point of Knowledge?
condition for knowledge but not for understanding. Why would that be? What social role does a factive epistemic concept play and how does it differ from the social role played by a nonfactive epistemic concept? This question is wide open for philosophical investigation. Thus, many of the standard epistemological debates about understanding should be regarded as a prolegomenon to a future (and I think more interesting) philosophical inquiry. I hope to foreground these broader questions about the role of understanding in human life, as well as to throw new light on epistemic value.
9.2 The Point of Understanding Why do we attribute understanding? What role does this concept play in human life? The hypothesis I want to test is the following: attributions of understanding are primarily used to identify individuals who can provide us with explanations. Put more briefly, I hypothesize that “understands” identifies explainers. This isn’t just wild speculation. In his Posterior Analytics, Aristotle discusses the close connection between explanation and understanding, and this conceptual link is widely acknowledged in the contemporary lit erature. Here are a few representative statements: [U]nderstanding, as Salmon puts it, results from “our ability to fashion explanations.” That is almost tautological. (Kim 1996: 61)
Understanding without explanation? Impossible, or so I will argue—in the case of science, at least. (Strevens 2013: 510)
[I]f you understand why p, you can give an explanation of why p and you can do the same in similar circumstances. (Hills 2016: 662)
While the link between explanation and understanding can be filled out in different ways, several philosophers have argued that one must be able to give a correct explanation to have understanding. Khalifa (2011: 108), for instance, says it is impossible to understand why a phenomenon took place without believing a correct explanation for it.7 And de Regt (2009: 25) 7 Must the understander believe her explanation? Khalifa (2011: 108) thinks so, but Wilkenfeld (2017) presents a case in which an agent seems to understand why p and yet does not fully believe the explanation for why p. The relationship between belief and understanding is important for at least two reasons. First, //understanding is a species of knowledge and //knowledge requires belief, then it follows immediately that understanding requires belief. Thus, if it could
what’s the point of understanding?
|
225
claims that understanding a phenomenon just is having an explanation of that phenomenon from an intelligible theory. Thus, the idea that explana tion plays an indispensable role in understanding is not all that new, and certainly not mere speculation. You might worry that my hypothesis gets things backward. Shouldn’t we characterize an explainer in terms of understanding rather than the other way around? Put differently, doesn’t one need to have understanding to qualify as a good explainer? This is a natural concern. But despite the allure of this idea, I will try to cast doubt on it by developing a more plausible alternative. I maintain that understanding is most plausibly understood in terms of the idea of a good explainer; viz., we get to the nature of understanding by reflecting on the criteria for being a good explainer. It is by starting with the idea of an explanation-seeking inquirer that we are ultimately led to a plausible account of the nature and value of understanding. In contrast, anyone who claims that a good explainer should be characterized in terms of understanding would still owe us an account of what understanding is. I hope to shed light on the nature of understanding by reflecting on the criteria for being a good explainer, thereby providing a fuller account. (Anyone who is unconvinced that we can get at the nature of understanding by investigating the role of the concept of understanding can still take my argument to illuminate the nature, purpose, and value of the concept of understanding in human life.) This hypothesis—that attributions of understanding are used to certify explainers—makes the connection between explanation and understanding be shown that belief is necessary for knowledge but not understanding, this would entail that understanding is not a species of knowledge. Second, if understanding why p requires believing the explanation for why p, this raises the question of why understanding has this requirement. What is so important about flagging people who believe their explanations as opposed to identifying people who simply are good explainers (whether they believe their explanations or not)? In short, why would a good explainer have to believe what she tells her inquirer? I think our function-first approach is equipped to explain why we should come to expect people to believe their explanations in order to understand why p. The explanation is basically that outlined in §2.3. In general, informants will not provide inquirers with explanations they themselves do not believe. Cases involving people who do not believe their explanations and yet somehow manage to insincerely give good explanations are rare, and people who do this regularly are probably nonexistent. In the vast majority of cases, a good explainer will not only believe the information she gives, she will give it because she believes it. Inquirers will thus tend to seek explainers who either believe that p when p, or believe that not-/? when not-p. Nevertheless, we can concede some ground to people, like Wilkenfeld, who want to play down the belief requirement (because it may not be necessary); and yet belief need not drop out of the picture altogether (as it would on traditional analyses). Our functionalist story would explain why belief is an important component of understanding even if it is not absolutely necessary. The prototypical cases of understanding would include belief because they “enshrine the features that effect realization of the purpose when things are going on as they nearly always do” (Craig 1990: 15).
226
|
What’s the Point of Knowledge?
quite clear. But why do we care about identifying explainers? This ques tion needs an answer if we are to clarify the importance of understanding in epistemic evaluation. The answer is simple: we care about identifying explainers because we value explanations. Humans seek to acquire and provide explanations in a variety of domains. We seek to understand and to explain the origins of the universe, why some actions are morally wrong, the cause of a rash, and so on. Our quest for explanations is strikingly domain general, as Gopnik points out: We seek and are satisfied by explanations of physical objects, animate beings, psychological agents, and even social groups. We seek and are satisfied by explanations in terms of physical laws, biological processes, reasons, or rules. (2000: 311)
Even children ask “Why?” within months of uttering their first words, and this question cuts across domains for them as well (Wellman, Hickling, and Schult 1997). Humans, it seems, are natural explanation-seekers. The unmet need for an explanation can leave us irritated or unsatisfied in a way that compels us to seek a resolution. Humans seek explanations, but what are explanations? They have been characterized in a variety of ways, with different theories offering different constraints on explanations. For example, Hempel and Oppenheim (1948) require an explanation to make essential use of a law of nature. Unificationist theories state that explanations are deductions that are tokens of the most unifying types of argument (Kitcher 1989 and Gijsbers 2013). Causal theorists maintain that events can only be explained by their causes (Salmon 1984, Strevens 2008, and Woodward 2003). While there is surely some truth in each of these views, there is no compelling reason to think all explanations must have a common structure or function. The research on explanations spans multiple kinds of judgments and distinct cognitive mechanisms. Thus, we should be permissive about what counts as an “explanation.” This pluralist view does not prevent us from offering a general char acterization of explanations. Like any family resemblance term, we can pick out a cluster of overlapping similarities, even if there is no set of necessary and sufficient features that will account for all the things we call “explanations.” For example, explanations are typically the answers to “why” or “how” questions. These answers will often refer to causal relations (Keil 2006), but not all explanations must pick out causal re lations; for instance, explanations given in mathematics or morality are
WHAT’S THE POINT OF UNDERSTANDING?
|
227
unlikely to be causal. We should expect different kinds of explanations from different subject matters. Thus, an attractive strategy proposed by Kim (1994), Greco (2014), and Grimm (2014) is to set aside the narrow focus on causation and appeal to the more general notion of dependence. On this view, dependence is the genus category, with different kinds of dependence—causal, mereological, grounding—playing the role of spe cies (Grimm 2014: 341). Unlike much of the literature on explanation, which has been firmly grounded in the philosophy of science, this sort of explanatory pluralism allows us to make sense of our pervasive use of nonscientific explanations in daily life. I have argued that identifying explainers is important because seeking explanations is central to human inquiry, but I have not yet said much about why we value explanations. What is so great about them? I think our interest in explanations has two general sources: we seek them for prac tical reasons as well as purely epistemic reasons. There are many practical incentives for wanting to explain our envi ronment. For example, the process of seeking explanations facilitates the confirmation of instrumentally valuable theories, which contributes to predicting future events. This, in turn, enhances our ability to control changes in our environment. As Craik puts it, the power to explain involves the power of insight and anticipation, and this is very valuable as a kind of distance-receptor in time, which enables organisms to adapt themselves to situations which are about to arise. (1943: 7)
Similarly, Hempel refers to man’s persistent desire to improve his strategic position in the world by means of dependable methods for predicting and, whenever possible, con trolling the events that occur in it. (1962: 9)
As a means to anticipating and accommodating the future, explanations have clear survival value. Thus, explanations serve an important practical function in guiding our interactions with the world. Explanations also play a crucial role in justifying or rationalizing action. Why did the Ancient Egyptians mummify their dead? Why is Tom being such a jerk lately? Questions that call for explanations, and explanations for human action in particular, are among the commonest questions we have. Why did Othello kill Desdemona? To answer this question, we need
228
|
What’s the Point of Knowledge?
an explanation for his action, namely, the fact that he believed Desdemona was unfaithful to him (among other facts, presumably). An emerging body of research from the cognitive sciences demonstrates that both explanation (the process) and explanations (the products) have many other practical benefits for everyday cognition. I do not have space to discuss this literature in detail, but Lombrozo (2011) nicely summarizes these findings. She discusses how engaging in explanation can facilitate learning, guide exploration, and influence decision-making. Explanations also play a role in motivating the construction of causal theories, resolving inconsistencies, calibrating metacognition, and diagnosing and repairing things. Thus, explanations clearly have prudential value. But a purely prudential account of explanatory inquiry would be too limited. Beyond our practical concerns, we also have a desire to make sense of the world rooted in what Hempel (1965) calls “sheer intellectual curiosity.” We are fascinated by why the dinosaurs became extinct even though knowing its cause would have little, if any, practical impact on our lives. Some questions attract our curiosity even though their answers have no practical value for us. This point is widely acknowledged. Goldman talks about the desire for truth 'for its own sake, not for ulterior ends” (1998: 98). Grimm distinguishes “epistemic curiosity,” which responds to our sense of puzzlement, from “prudential curiosity,” which responds to some basic prudential concerns of ours (2008: 737). And Michael Strevens opens his book Depth: An Account of Scientific Explanation with the following lines: If science provides anything of intrinsic value, it is explanation. Prediction and control are useful, and success in any endeavor is gratifying, but when science is pursued as an end rather than a means, it is for the sake of understanding—the moment when a small, temporary being reaches out to touch the universe and makes contact. (2008: 3)
In other words, our second basic motivation for explanatory inquiry is our deep concern to make sense of the world we live in, to explain the un ending flow of our experiences. Our interest in explanations thus has two sources: we have many prac tical incentives for wanting to explain our environment, but explanations can also be rewarding in their own right, independently of whatever prac tical benefits they might confer. Both of these facts explain why humans need to identify explainers. We value explanations, but as finite beings with limited cognitive resources we cannot figure out everything on our
what’s the point of understanding?
|
229
own. Instead we often rely on the testimony of others. Thus, we have an interest in identifying people who can provide us with explanations. I hy pothesize that the concept of understanding is used to pick out explainers. In this way, the concept of understanding plays an important role in the general economy of our concepts.
9.3 Refinements You might worry this hypothesis is too narrow. Aren’t there many possible situations in which a person can have understanding and yet not be able to provide an explanation? For example, explanation seems to be linguistic and explicit, but someone who cannot speak or write can surely have un derstanding. Also, a person like Robinson Crusoe might understand many things and yet not function as an informant or explainer for anybody. So why should we think attributions of understanding are used to identify people who can provide us with explanations? Further, inquirers will not settle for just any explanation: they want good (or true) explanations. In all of these cases, a gap seems to open up between our natural ascriptions of understanding and the concept we have arrived at by considering the practical situation of an explanation-seeker. I will be the first to admit that we may need to supplement or modify our initial hypothesis, especially if it doesn’t issue in a concept that is at least very similar to our everyday notion of understanding. But this is no reason for despair. It is a good policy to test the explanatory powers of the simple before resorting to the more complex, even if a more complex hypothesis is ultimately needed. Thus, I will now introduce some refinements to stave off objections and hopefully enhance my proposal’s plausibility. First, my hypothesis is best construed as an account of understandingwhy. Understanding-why is at the center of a lot of recent work in episte mology, as well as the philosophy of science, and many scholars regard it as the paradigmatic form of understanding. Pritchard (2010), Hempel (1965), Kitcher (1989), Grimm (2008), de Regt (2009), Khalifa (2012), Strevens (2013), and Hills (2016) all focus on understanding-why, and Pritchard and Hills say it is the paradigmatic form of understanding.8 Some examples of understanding-why include these: I understand why the earth’s average temperature is increasing; I understand why that driver ran a red light; I un derstand why the sky is blue. My hypothesis is meant to provide an account
8 This is not uncontroversial (see Kvanvig 2003 and Elgin 2007).
230
|
What’s the Point of Knowledge?
of understanding-why p, where p is some proposition. I am especially inter ested in this type of understanding because the social function of epistemic evaluation seems deeply linked to the transmission of information, and propositions are the most common vehicle of transmittable information. Pritchard takes paradigm uses of “understands” to be statements like “I understand why p is the case,” but he says understanding usually isn’t concerned with propositions. He thinks this because “it is rare to talk of understanding that p” (Pritchard 2010: 74, emphasis mine).9 However, it is plausible that to understand why p one’s understanding must be directed toward a set of propositions, namely, those reasons as to why p is the case (see Hills 2016 and Boyd 2017). Thus, Pritchard’s reasoning does not necessarily speak against the idea that the information content of understanding-why is propositional. Understanding-why is often contrasted with two other kinds of under standing: practical and objectual. Practical understanding, or understandinghow, is more closely tied to abilities or skills than explanations. 10 For example, a player who understands how to catch a fly ball might be un able to explain how he can do this, and explanations seem inadequate to imbue such a skill. Objectual understanding, in contrast, is more holistic. It is usually attributed using the verb “understands” followed directly by a noun; for example, “Elizabeth understands American history” and “Lana understands quantum physics.” According to Khalifa (2013), objectual understanding can be reduced to understanding-why. Similarly, Grimm (2017) argues that understandingwhy and objectual understanding differ only in degree, not in kind. To the extent that objectual understanding can be reduced to understanding-why, my claims about understanding-why will translate to the objectual case. This would increase the explanatory power of my account. Further, we invoke explanations not only to shed light on why something happened but also to explain how to do things, what to do in a given situation, and what happened. This provides some reason to think our need for good explainers isn’t limited to cases of understanding-why. Nevertheless, there
9 This is actually false. Peter van Elswyk surveyed instances of “understanding that” in the Corpus of Contemporary American English (COCA) and found they outnumber instances of “understanding why” more than two to one. 10 Not all cases of understanding-how are incompatible with my account. Some “how” questions demand explanations; for example, “How do seahorses reproduce?” Thus, we cannot distinguish understanding-why from other types of understanding solely in terms of their grammatical form. Nevertheless, I hope to make use of a distinction between various forms of understanding that is widely accepted in the epistemology of understanding in order to delimit the focus of my inquiry.
what’s the point of understanding?
|
231
is likely a nonreducible variety of uses of the word “understands,” not all of which can be assimilated to my explanatory account. So while I am open to the possibility that my account may shed light on other types of under standing, I shall restrict all subsequent remarks about “understanding” to cases of understanding-why. A second refinement is needed. Our current hypothesis is that attributions of understanding are used to identify explainers, but not just any explainer will do. A street psychic might tell me that my marriage is struggling because my wife has a conflicting zodiac sign, but this is not a good explanation of our unfortunate situation. Even if I believed in astrology and would thus get the “sense” of understanding from this explanation, that does not suffice to make it a good explanation. A good explanation must be genuinely explana tory. Astrology, however, does not correctly explain the cosmic order. Thus, our initial hypothesis needs a slight modification, namely, attributions of understanding are used to mark out people who can provide us with good explanations. In short, “understands” identifies good explainers.11 Three clarifications are in order. First, it is implausible that one can understand why p if it is not the case that p. For example, you cannot understand why Jesse James robbed the bank if he did not rob the bank. Likewise, you cannot understand why Brad Pitt dumped Jennifer Aniston if he did not dump her. Thus, understanding-why turns out to be factive in basically the same way that knowledge is factive. It is widely accepted that S fails to know p if p is false. Likewise, we should accept that S fails to understand why p if it is not the case that p. The explanandum must be true if one is to have understanding-why. This sheds some light on what our concept of understanding is like. If understanding-why were not factive, it likely wouldn’t satisfy our need to identify good explainers.11 12
11 You might wonder why we need the concept of understanding if we already have the concept of a good explainer. However, I am using “good explainer” as a theoretically fertile philosophical notion, not as a term that has lay currency. It is designed to be a limiting concept that might be different from any folk conception of this expression. I will outline my criteria of a good explainer in the next section. 12 A far more complicated issue is whether the explanans must also be true. According to some scholars, a factive conception of understanding “cannot do justice to the cognitive contributions of science” (Elgin 2007: 32). For example, scientists purport to understand the behavior of actual gases by reference to so-called ideal gas, even though there is no such gas. Elgin calls these “felicitous falsehoods” (2007: 39). I will not pretend to resolve this contentious matter here, but I will offer two brief remarks. First, I find it plausible that some cases of understanding are not factive (or only quasi-factive); however, these are not typically cases of understanding-w/zy. Elgin seems to grant this point in her new book (2017: 40). A nonfactive conception of understanding is far more plausible in cases of objectual understanding than understanding-why. Second, Hazlett (2017) has convinced me that understanding-why is a species of correct representation.
232
|
What’s the Point of Knowledge?
Second, one might argue that an explanation must by definition be genu inely explanatory; in other words, a false explanation would be no explana tion at all. However, I think conspiracy theorists can provide explanations for events even when these explanations are false. They simply do not provide good explanations (because they are not genuinely explanatory). Third, it might be the case that what counts as a good explanation (and perhaps even an explanation at all) varies contextually, both between speakers and for the same speaker at different times. I will return to the issue of context sensitivity at the end of this section. On the table now is the following modified version of our initial hypoth esis: “understands” certifies good explainers. But this hypothesis still faces objections. There are situations in which a person seems to count as an “understander” even though that person is unable to provide a good explanation. Suppose Barney is unable to speak and write, so he cannot provide us with an explanation for why the floor is covered in milk. Still, Barney might understand why there is milk on the floor because he saw what happened. Thus, attributions of understanding do not seem to certify good explainers—or so the objection goes. This objection can be straightforwardly answered. Barney’s lack of verbal and linguistic abilities might render him unable to provide us with a good explanation, but if Barney truly understands why there is milk on the floor, then he must be able to adequately explain this to himself. If he cannot formulate an explanation to himself, he lacks understanding.13 As John Searle once said, “If you can’t say it clearly, you don’t understand it yourself.” Thus, there is a sense in which Barney can provide a good explanation, even if he cannot explain it to us due to certain communi cative obstacles. To say Barney understands but cannot explain would be pedantry. Even if he cannot express it, he has a good explanation (Gijsbers 2013: 518). Further, Barney’s inability to communicate an explanation would not imply the explanation is inherently incommunicable, but merely that he is presently the wrong person for the job.14 Another way of putting As such, understanding-why p will require “correctly representing E as the explanation for the fact that p, which entails that E is the explanation of the fact” (2017: 138-139). 13 One might prefer to say an understander must grasp a correct explanation (see Kvanvig 2003, Elgin 2009, Grimm 2011, Strevens 2013, and Hills 2016). My hypothesis could be reformulated thus: attributions of understanding are used to identify individuals who grasp a correct explanation of the phenomenon to be understood. This might be the right way to understand the epistemic relation between an agent and an explanation, but a lot turns on what grasping amounts to. Until we get clearer on the nature of grasping, I wish to avoid postulating the existence of a novel and poorly understood mental state. 14 Hills (2016) thinks understanding-why typically requires the ability to explain, but she says it might not be necessary. Perhaps there are some (unusual) circumstances in which one
what’s the point of understanding?
|
233
this point is that we are interested in identifying people who have good explanations. Presumably we are interested in identifying such people be cause it is in our collective interest to mark out individuals who can share their explanations with us and with others. A consequence of this view is that infants lack understanding-why. According to some cognitive psychologists, infants can show tacit un derstanding even though they are unable to provide an explanation (and also a self-explanation). For example, Clements and Pemer (1994) claim that infants implicitly understand that some agent holds a false belief about the location of an object on the grounds that these children will look at the area where the agent mistakenly believes the object is located (as opposed to where the children, but not the agent, saw the object was moved). However, this phenomenon can easily be described as cases in which young children show some awareness of false beliefs without really having understanding-why. As Pemer and Ruffman (2005: 214) argue, the apparent success on these theory-of-mind tasks can be explained without supposing that infants have any understanding of the belief at all. (I set aside animal understanding because I am providing an account of human understanding.) Why, then, do we sometimes attribute understanding to young chil dren? As Elgin (2017: 30) points out, we can use epistemic success terms in ways that reflect a “courtesy usage”; for instance, we might say that a young child has some understanding of photosynthesis because she thinks that sunlight is the flower’s food. I find it feasible to count such attributions of understanding as mere courtesy usage—or what Kvanvig (2003) calls “honorific” uses. If you are unconvinced by this move, I ask you to reconsider whether these children really have understanding-w/zy. Perhaps there are plausible cases in which children have objectual under standing or understanding-that, but it is less plausible to say that a child who cannot formulate any explanation for why a phenomenon occurred (even roughly and to herself) has understood why the phenomenon occurs. Acknowledging this does not prevent us from saying something about what makes the view of one child better than the view of another child. The child who thinks that sunlight is the flower’s food may still know more, or be closer to understanding photosynthesis, than the child who thinks that flowers survive on magical powder.
understands but cannot articulate an explanation, even to oneself. In these cases, we might follow Hills and distinguish explicit understanding from tacit understanding.
234
|
What’s the Point of Knowledge?
It is also worth pointing out that what counts as “providing an expla nation” can be filled out in a number of ways, not all of which might be linguistic. Perhaps Barney can show me what happened, even if he cannot tell me or write it down. But even if there is no way for Barney to provide me (or anyone but himself) with an explanation, this obstacle is really no more worrying than a situation in which a person cannot provide us with an explanation because she is nowhere to be found. Someone who cannot be located can still, in an important sense, provide a good explanation— she simply cannot provide it to us right now. As information-dependent beings, it is in our interest to identify people as good explainers even if they are not functioning as a good explainer at a specific moment. After all, we can often rely on such people for communicating explanatory in formation under the right circumstances (e.g., if we can find them and if they are willing to tell us). Potentially worrying cases involving isolated individuals, such as Robinson Crusoe, can be dealt with in a similar way. You might worry that the social function of identifying good explainers leaves mysterious why there are cases of understanding that do not involve any interper sonal relations. But there are at least two reasons to regard someone as an explainer even if no one else is around. First, we can explain things to ourselves, as already mentioned. Second, it is important that we have a practice whereby people can declare themselves to be good explainers, since they themselves will often be the only people in a position to tell whether they are qualified or not. An understander is someone who meets a sufficiently high-quality epistemic position such that potential inquirers could in principle rely on this person’s information, even if nobody actu ally does ever seek such an informant. This explains why we might want to say that someone understands why p even though, as it happens, that person does not actually function as an explainer for anybody. This account is also compatible with the highly plausible idea that un derstanding comes in degrees, thereby casting additional light on what our concept of understanding is like (Elgin 2007, Grimm 2014, and Hills 2016). The quality of explanations comes in degrees, so attributions of understanding will track these differences in quality. I will not provide a detailed account of explanatory quality here, but it will suffice to note that the quality of explanations varies along at least two dimensions: depth and breadth.15 The comprehensiveness of the body of information contained in
15 See Elgin (2007), Greco (2014), and Strevens (2013) for more discussion of this distinction.
what’s the point of understanding?
|
235
an explanation can increase its breadth, while a causal model that is more abstract is deeper (Strevens 2008). Further, Railton (1981) says an expla nation that traces an event’s causal history back farther in time is better for it, and Hills (2016) argues that one’s understanding can be better or worse depending on how good one is at explaining things in one’s own words. How deep or broad an explanation must be for the explainer to qualify as an understander will likely depend on context. Thus, the concept of un derstanding will in some way be context sensitive. I do not have space to delve into this complicated matter here, but I will point out that context can affect the epistemic standard for understanding in at least two ways (I will set aside the pragmatist view outlined in chapter 7). On the one hand, the threshold for understanding might be invariant even though our will ingness to attribute understanding is affected by context. According to this view, it might be inappropriate to attribute understanding to someone who meets the (fixed) threshold for understanding because it would generate the false implicature that her explanation is good enough for the context. On the other hand, the threshold for who truly qualifies as an understander (and not just when it is appropriate to attribute understanding) might itself vary with context. I want to remain neutral on this issue, but I will reg ister that I incline toward the latter view. This is because who truly counts as a good explainer will likely depend in part on the circumstances and one’s audience. For example, one might qualify as a good explainer in the context of a high school classroom but not at a professional meeting of experts in the relevant field. Thus, in some contexts one might not meet the required standard for understanding while in other contexts the epistemic bar might be lower. Still, there will likely be a floor to how low the epistemic standard may drop, regardless of context. If we assume that our concept of under standing will undergo a process of “objectivization,” then attributions of understanding will be used to assess the adequacy of explainers for people and purposes beyond one’s immediate context. As with knowledge, our epistemic needs are more socially and temporally extended than requiring information for oneself here and now. We want to identify people who can provide adequate explanations for a community of individuals with diverse needs and interests. This will edge us toward the idea of someone who is a sufficiently good explainer as to why p quite generally. In any case, settling this question about the truth conditions of under standing attributions is not essential for my purposes because ascriptions of understanding will likely serve their function regardless of whether invariantism or contextualism is true (for reasons similar to those mentioned
236
|
What’s the Point of Knowledge?
in chapter 7). At the very least, there is no obvious reason to think one of these two views about the semantics of understanding attributions is in compatible with my hypothesis about their function.
9.4 Do Understanding and Explanation Come Apart? A function-first epistemologist aims to construct a concept that not only functions in the manner suggested by his or her hypothesis, but also one that fits our intuitions. To this end, we must examine the extent to which our hypothesis matches the everyday concept of understanding. There are two ways to drive a wedge between explanation and under standing. The first is to show there are cases in which a person can explain why p even though she does not understand why p. The second, which I briefly discussed above, is to argue that an agent might have genuine un derstanding and yet be unable to provide a good explanation. Building on our refined hypothesis from the previous section, I’ll now deal with addi tional objections along these fines. I argue that my hypothesis does issue in conditions of application that match the intuitive extension of the word “understands.” The most obvious case in which a person can provide an explanation without having understanding is that of a parroter. A parroter is someone who mindlessly repeats what he or she has heard. Imagine a presidential candidate who knows little about climate change but who memorizes the following explanation of the phenomenon for his stump speech: Our climate is undergoing dramatic changes as a result of human activity. Although greenhouse gases like carbon dioxide play an important role in keeping the earth warm, human activities like the burning of fossil fuels have produced too much carbon dioxide. We need to reduce the emission of greenhouse gases in order to limit anthropogenic climate change.
The candidate might understand some aspects of this description, but let’s suppose he does not understand what greenhouse gases are, how they re late to climate change, or what role carbon dioxide plays in heating the earth. Nevertheless, the candidate seems to successfully provide an expla nation of climate change. It is, however, compatible with my hypothesis that people can provide explanations without themselves understanding. As you’ll recall, I argue that ascriptions of understanding pick out good explainers. The notion
what’s the point of understanding?
|
237
of a “good explainer” can be filled out in multiple ways, but one highly plausible characteristic of such a person is the ability to reliably evaluate explanations. Following Khalifa (2013), we may characterize a good ex plainer as someone who can typically discriminate between correct and incorrect explanations, and thus is not susceptible to believing incorrect explanations. Further, a good explainer can normally provide elabora tion, answer closely related questions, give explanations in his or her own words, follow an explanation of why p given by someone else, and will often be able to answer “what if things had been different” questions.16 The parroter, in contrast, will lack these abilities and will therefore not qualify as an understander. This also illustrates that knowing a good ex planation does not suffice for being a good explainer, since it is plausible to say the candidate knows a good explanation (via testimony) and yet he does not qualify as a good explainer. Another potential gap between explanation and understanding is the possibility of nonexplanatory understanding. You might think it is possible that, for all correct explanations E of p, a person might not know that E explains p and yet still understand p. But is understanding without expla nation possible? In “Understanding without Explanation,” Lipton (2009) rejects the idea that wherever we find understanding we also find explanation. He defends the thesis that understanding without explanation is possible by appealing to several examples, one of which is the following: Suppose that a boxing match between Able and Baker is rigged so that Baker—though in fact the far better boxer—would take a dive in the tenth round. Imagine, however, that as a matter of fact Able floors Baker with a lucky uppercut in the fifth. (Lipton 2009: 51)
Lipton uses this example to argue that potential explanations may provide understanding without approximating an actual explanation; thus, under standing needn’t pass through actual explanation. But how plausible is Lipton’s boxing example? Suppose we know that if the fight had lasted until the tenth round, Baker would have taken the dive, but we do not know that Abie’s fifth-round knockout actually caused his victory. Do we under stand why Baker lost the match?
16 See Woodward (2003), Grimm (2012), and Hills (2016). More abilities might be needed, such as those required for what Hills calls “cognitive control.”
238
|
What’s the Point of Knowledge?
According to Lipton, knowing that the match was rigged helps us un derstand why Able won, even though the explanation of the win depends entirely on the lucky punch. It is because we possess knowledge of salient counterfactual scenarios that we understand, even though we do not know the actual explanation. Khalifa agrees with Lipton that this is a case of understanding, although he says we do not understand why Able won as well as someone who knows about the actual course of events (2012: 12). I find these judgments surprising. It seems quite intuitive to me that we would only think we understood why Able won, when in fact we would not actually understand why. Imagine someone who is surprised by Abel’s victory asks us: “How did Baker lose to Abel?” If our answer is “Because Baker took a dive,” then we clearly do not understand why Baker lost the match. If we do not know that Abie’s fifth-round knockout was actually due to a lucky uppercut, then we have an incorrect explana tion for the match’s result and, consequently, we fail to understand why Baker lost (see Strevens 2013: 514). Thus, Lipton’s example does not illustrate the possibility of understanding without explanation. Rather, it seems consistent with my hypothesis that understanding requires a good explanation. Lipton also considers Galileo’s thought experiment that demonstrates that gravitational acceleration is independent of mass. Suppose that heavier things accelerate faster than light things, and consider a heavy and a light mass connected by a rope. Considered as two masses, the light one should slow down the heavier one, so the system should ac celerate slower than the heavier mass alone. But considered as one mass, the system is of course heavier than the heavy mass alone and so should accelerate faster than the heavy mass alone. But the system cannot both accelerate slower and faster, so acceleration must be independent of mass. (Lipton 2009: 47)
According to Lipton, Galileo’s argument is not an explanation but is a case of understanding-why. He says we understand why acceleration is independent of mass once we appreciate the contradiction revealed by Galileo’s thought experiment. However, I think a more plausible expla nation is this: appreciating Galileo’s thought experiment leads us to un derstand that acceleration must be independent of mass, but it does not lead us to understand why it is. In order to understand why acceleration is independent of mass, we must possess an explanation of that fact, which turns on the equivalence of gravitational and inertial mass, accounted
WHAT’S THE POINT OF UNDERSTANDING?
|
239
for by general relativity. So this case does not support the possibility of understanding-why without explanation. Perhaps, contra Lipton, there is a sense in which Galileo’s argument is an explanation. The explanation is this: to suppose that gravitational acceleration is not independent of mass leads to a contradiction, but contradictions are impossible, so this explains why gravitational accelera tion is independent of mass. This might qualify as an explanation; but if it is an explanation, it is not a very illuminating one, so it would give us very limited understanding-why. Lipton also says we can acquire understanding through demonstrations rather than explanations. Drawing on his own experience, Lipton writes: I never properly understood the why of retrograde motion until I saw it demonstrated visually in a planetarium. A physical model such as an orrery may do similar cognitive work. These visual devices may convey causal information without recourse to an explanation. And people who gain un derstanding in this way may not be left in a position to formulate an expla nation that captures the same information. Yet their understanding is real. (2009: 45)
This passage expresses two important ideas: first, that understanding can be attained in a nonexplanatory way (i.e., via demonstration); second, that understanding can be nonlinguistic. The first point is perfectly compatible with my hypothesis. I claim that ascriptions of understanding are used to pick out good explainers, but I do not argue that demonstrations cannot put one in such a position. In fact, it seems highly plausible that a visual demonstration of the sort Lipton describes might put an individual in a position to provide a good explana tion. Thus, I grant that understanding might be attained through demon stration (or some other nonexplanatory route). Lipton’s second point is that understanding can be nonlinguistic. We can reconcile this point with my hypothesis in two ways, both of which I have already suggested. First, we might expand what counts as an “ex planation” to include some nonlinguistic behaviors. I leave this possibility open. Second, we may acknowledge the possibility of agents, like Barney, who are unable to formulate an explanation for others and yet are able to engage in self-explanation. Still, I want to insist on my earlier claim that any agent who is unable to formulate a self-explanation does seem to lack understanding. In Lipton’s case, the reason the outer planets occa sionally seem to reverse their motions is that the earth’s orbit around the
240
|
What’s the Point of Knowledge?
sun is tighter than the outer planets, so it will occasionally overtake them, thereby causing the appearance of retrograde motion. Anyone who cannot formulate this idea, even roughly, has not understood why the phenom enon occurs. Thus, it seems impossible for someone to understand why p unless she can formulate an explanation as to why p. A consequence of this view is that we probably understand less than we think we do. I find this perfectly acceptable. It is not counterintuitive to say that people often think they understand something that they do not ac tually understand. Indeed, our phenomenological sense of understanding is known to be an unreliable indicator of genuine understanding (see Trout 2002). My account would simply reveal that people sometimes claim to understand things that, if pressed for an explanation, they turn out not to understand. This phenomenon is known as “the illusion of explana tory depth” and it is well documented in psychology (see Rozenblit and Keil 2002). Lipton provides other putative examples of understanding without ex planation, but I do not have space to discuss them all here.17 Many of them can be set aside as examples of practical understanding or objectual under standing, not understanding-why. As mentioned earlier, an explanationist conception of understanding is probably too narrow to account for the full range of cases in which we attribute understanding. But this cuts no ice with my view because I am analyzing understanding-why. In lieu of an exhaustive critical discussion of each of Lipton’s examples (and many other possible examples), let’s suppose for the sake of argu ment there are some genuine cases of understanding-why without expla nation. Would this put my view in peril? Not necessarily. Even if there were instances of understanding-why without explanation, these situations are presumably quite rare (I haven’t come across any such cases). This is consistent with my claim that the primary function of attributing understanding-why is to identify good explainers. As a function-first epistemologist, the goal of my inquiry is not to enumerate the necessary and sufficient conditions for understanding, but rather to investigate the nature and origins of our present practice. I am interested in identifying what might be called the “core” of the concept of understanding-why, which allows me to specify conditions that hold only for the most part, but not always.
17 See Khalifa (2012) and Strevens (2013) for additional criticisms of Lipton’s examples.
what’s the point of understanding?
|
241
An advantage of this approach is that we need not ignore important char acteristics of understanding. A theory that proposes necessary and/or suf ficient conditions will be rejected upon encountering any counterexample. If, for example, it can be argued that a case of knowledge without belief is possible, then belief will make no appearance in our theory of knowl edge (see §1.2 for a discussion of this point). Similarly, if a case of under standing without explanation were possible, then explanation would not be a necessary condition for understanding. Must we therefore abandon our theory in the face of such “counterexamples”? I suggest not. Instead, we should be open to the possibility that explaining the nature and value of understanding will come from certain contingent characteristics. Even if there are some cases of understanding without explanation, the possibility remains that explanation is something we value about understanding. In contrast to traditional conceptual analysis, my approach permits us to include characteristics that, while not necessary, are important to under standing and important to us.18
9.5 Does “Understand” Identify Experts? I am not the first person to explore why humans attribute understanding. Daniel Wilkenfeld, Dillon Plunkett, and Tania Lombrozo (2016) have recently investigated the folk concept of understanding to determine its function. Their study consists of a series of experiments that contrast people’s use of the locution “X understands why” with their use of the lo cution “X knows why.” Wilkenfeld and his colleagues (henceforth WPL) look at how people actually use the word “understands,” and then they draw conclusions about the function of understanding attributions from this data. In other words, they attempt to “reverse-engineer” the concept of understanding, so their methodology will face the objections mentioned in §1.4. But instead of rehashing the limitations of this approach, I want to dis cuss the empirical findings reported by WPL and evaluate their hypoth esis about understanding. I will argue that they mischaracterize what attributions of understanding are for; moreover, their data are compatible with my hypothesis that understanding-why identifies good explainers. WPL tested for potential differences in attributions of knowledge and understanding as a function of explanatory depth. What they found
18 Jones (1997) makes this point with respect to Craig’s theory of knowledge.
242
|
What’s the Point of Knowledge?
is that people attribute knowledge more willingly than they attribute un derstanding in cases where someone has a true explanatory belief that is nevertheless fairly shallow. In their own words, “participants are more in clined to attribute knowledge than understanding when the character has beliefs about an explanatory connection, but minimal explanatory depth” (2016: 381). This is consistent with Pritchard’s (2009: 38) claim that attributions of understanding have higher demands for explanatory depth than attributions of knowledge. To illustrate this point, Pritchard describes a case in which it seems possible for a child to know why his house has burned down (e.g., he is told it was because of the faulty wiring) even though he does not understand why his house burned down, for he has no conception of how faulty wiring might cause a fire. Why do knowledge and understanding diverge in this way? According to WPL, it is because the epistemic aim of understanding (but not knowl edge) is to pick out experts to whom we should defer. They write, understanding attributions are used to identify experts to whom we should defer, and as a result we are differentially hesitant to ascribe understandingwhy (relative to knowledge-why) when the beliefs constituting that un derstanding are not supported by access to deep explanatory information. (2016: 383)
To test this hypothesis, WPL designed an experiment to examine whether people’s ratings of how much someone understood why correlated with their ratings of how valuable that person would be as a source of informa tion about closely related processes. All participants read the same eight vignettes. Each vignette described a person who sits down at a computer, accidentally presses some keys on the keyboard, and causes an “fl” char acter to appear on the computer screen. These vignettes varied in the ex planatory depth of the character’s grasp of the cause of the fl character appearing on the screen. WPL found that when participants were asked whether the protagonist would be a good resource for a person to consult if he or she needed help resolving a technical problem with the computer, positive responses were significantly more correlated with understanding attributions than knowledge attributions. Thus, understanding attributions correlate with a participant’s willingness to defer to the understander re garding issues from the same domain. WPL take this connection to reveal the functional role of the concept of understanding. We now have two hypotheses on the table. I have hypothesized that understanding attributions certifies good explainers. According to WPL,
what’s the point of understanding?
j
243
however, ascriptions of understanding are used to demarcate skillful experts to whom we should defer. Who is right? Before arguing that my hypothesis is more plausible, let me high light three commonalities between these two proposals. First, the data collected by WPL reveal a close conceptual link between understanding and explanation, which favors both hypotheses equally. Second, both hypotheses correctly predict that people will withhold attributions of understanding when the subject of the attribution has minimal explan atory information. (This is because good explainers and experts must both attain a certain degree of explanatory depth.) Third, people’s ten dency to ascribe understanding to a subject S will, on both accounts, significantly correlate with a willingness to defer to S on issues in the same domain. Given these commonalities, you might wonder whether these hypotheses are just two sides of the same coin. I argue they are not. While under standing seems deeply connected to identifying good explainers, there is a large conceptual gap between understanding and expertise. Let me explain. While understanding is almost certainly necessary for expertise in a domain (i.e., there are no experts who have no understanding), it is not sufficient. Indeed, there are many possible cases in which a person has un derstanding but does not qualify as an expert in the relevant domain. For example, when I see someone accidentally hit a table with his knee and a cup falls from the table, I understand why it falls; but it seems false to say that my judgment reflects expertise. There are many similar cases of “easy understanding” that create a large gap between expertise and under standing. Consider the following case by Grimm: when I start chopping onions and my eyes begin to water, I think I under stand why my eyes are beginning to water, namely, because I am chopping the onions. I don’t think it is because of the time of day, or the color of the shirt I am wearing, or anything like that; it’s because of the onions. (2014: 337)
Admittedly, this understanding is not as deep or extensive as the under standing had by someone who is able to say what it is about the onions that causes this reaction (i.e., the sulfur compounds). But understanding comes in degrees, so a fairly shallow understanding is still genuine under standing. Nevertheless, it would be odd to say that Grimm’s understanding makes him an “expert” on the relevant issue.
244
|
What’s the Point of Knowledge?
By tying understanding to expertise, WPL make understanding a rare achievement. Few people are experts across domains, but many have some degree of understanding across domains. So if “understands” picks out ex pertise, than understanding is a much rarer achievement than we thought. While this sort of skepticism about understanding is not necessarily false, it contradicts WPL’s goal of studying the folk concept of understanding. Thus, it is doubtful that ascriptions of understanding play “a fundamental role in identifying experts who should be heeded with respect to the gen eral field in question” (Wilkenfeld, Plunkett, and Lombrozo 2016: 373). There is, however, a way to interpret WPL’s view about the purpose of understanding attributions that makes it more plausible. Perhaps WPL are not concerned with the everyday notion of expertise. Rather, they might be using “expert” to refer to people whose beliefs attain a certain level of explanatory depth and who can be relied on to solve difficult problems within the relevant domain. This is a broader category of people than is captured by the folk notion of an “expert.” On this interpretation, under standing would not be an all-too-rare achievement. Many people can attain a certain degree of explanatory depth in a domain without qualifying as experts in that domain (e.g., I know a fair bit about Hume’s Enquiry, but I am no expert). This interpretation of WPL’s hypothesis brings it much closer to my own view, but a crucial gap must still be filled. WPL admit they “have not offered an analysis of how attributions of knowledge function in our epistemic lives” (2016: 391). Instead, their aim is to characterize under standing. The problem, however, is that knowledge ascriptions are com monly held to identify people to whom we should defer (Craig 1990 and Keil 2006). But, as WPL observe, attributions of knowledge and attributions of understanding often diverge, which suggests these concepts play different roles in epistemic evaluation. So if the role of understanding is to identify people to whom we may reasonably defer, what role is left for knowledge? We need an account of epistemic evaluation that is sensitive to at least three facts: first, attributions of knowledge and attributions of un derstanding often diverge, and this seems to be a function of explanatory depth; second, these differences in attribution suggest that understanding is often a stronger epistemic state that knowledge; third, understanding and knowledge are both intimately related to deference. In the next sec tion, I will argue that understanding and knowledge play different social roles. Further, I will discuss the importance of these roles and argue that knowledge is more central than understanding to human cooperation.
what’s the point of understanding?
|
245
9.6 The Social Roles of Knowing and Understanding It is useful for humans to have a range of predicates that serve to express a variety of evaluations of the epistemic position of other people. As I have been arguing in this chapter, one useful predicate has to do with identifying good explainers. This is because humans are natural explana tion seekers and yet cognitively dependent on others. Another useful pred icate, such as “rational,” might be used to influence our audience to follow the endorsed befief-forming rules (Dogramaci 2012). A third predicate, namely “knows,” serves the function of identifying reliable informants. As argued in §5.2, we often terminate inquiry when we find a suffi ciently reliable informant. This is because spending all our time and re sources to continue inquiring would be impractical. Further inquiry is not always worth the reduced risk of being wrong. We therefore need a point at which people may reasonably stop inquiring. How do we signal when inquiry has gone on long enough? I have argued that we use the concept of knowledge to certify information as being such that it may, or even should, be taken as settled for the purposes of one’s deliberations. Pritchard denies that inquiry reasonably ends at knowledge. He argues that inquiry reasonably terminates at understanding: Now ask yourself whether inquiry that resulted in knowledge but not in the corresponding understanding would be deemed a successful inquiry (and thus a “closed” inquiry, at least as regards the original question under inves tigation). I suggest not. (Pritchard 2010: 85)
To support this idea, Pritchard considers a case in which someone finds his house has burned down and is led to wonder what caused the fire. Pritchard argues that this inquiry will not be properly closed until that person comes to understand why his house burned down. While I agree with Pritchard about this example, his case does not il lustrate that understanding is the aim of inquiry. That view is too strong to be plausible. Why think all inquiries must have the same aim? While some inquiries might demand understanding in order to be legitimately closed, it is unlikely that all do. Upon visiting the computer repair shop, I am satisfied with knowing that my laptop works without needing to un derstand why or how it works. This inquiry would successfully end at knowledge because I am not always interested in underlying explanations or the causes of certain events. The reason why inquiry aims at under standing in Pritchard’s example is rooted in the fact that the inquirer wants
246
|
What’s the Point of Knowledge?
an explanation for the event. But our curiosity is very often directed at nonexplanatory information (see Kelp 2014). Suppose I am looking for my car keys. In this situation, I might want to know whether my wife has them or not. But it is of no interest to me that my wife has them because she needed to get the groceries out of the car and she forgot to put the keys back on the table. My inquiry would reach its goal and properly end even if I did not attain an understanding of why the relevant proposition is true. Thus, while understanding sometimes legitimately closes inquiry, mere knowledge will often suffice.19 In fact, knowledge is probably more centrally linked to inquiry closure than understanding. Acquiring a deep understanding of the world might be worthwhile from a “purely epistemic point of view” (or within the proj ect of “pure inquiry”), but such ratified concerns have little to do with the vast majority of our everyday inquiries. When I need to find my way to Yankee Stadium, my judgment about whether a random passerby knows its location has little to do with a concern for understanding. I do not re quire my informant’s belief to attain a high level of explanatory depth; all I need is to acquire a true belief from a reliable source of information. What I am looking for is a reliable informant on the truth of whether p, not a good explainer as to why p. Ascriptions of knowledge and ascriptions of understanding often come apart because “knows” typically picks out reli able informants, whereas “understands” identifies good explainers. Since our ordinary inquiries are often satisfied by identifying reliable informants as to whether p (without necessarily acquiring a good explanation as to why p), inquiry will often end at knowledge. It might be true that if we could set aside various practical considerations, then we would aim for understanding. But given our need for actionable information, our epi stemic dependence on others, and the fact that we can rarely afford to pay the higher “informational costs” needed to acquire understanding, it is far more likely that everyday inquiry will terminate at knowledge.20 In this respect, ascriptions of knowledge play an important role in everyday life that ascriptions of understanding do not.
19 In more recent work, Pritchard seems to grant this point (2016: 34). However, he argues that our inquiries ought not to be sated by mere knowledge of the answer to one’s question. An agent who is generally willing to regard inquiries as legitimately closed even though the relevant understanding has not been attained lacks intellectual virtue, according to Pritchard. I think this is also too strong. 20 It is also plausible that inquirers sometimes have to settle for less than knowledge due to limitations of time and other resources.
what’s the point of understanding?
|
247
Linguistic data support this argument. As mentioned in the introduction of this book, “know” is one of the 10 most commonly used verbs in English, the first cognitive verb that children learn, and the most prominently used term in epistemic assessment. It has also been argued that “know” is unlike almost every other word because it finds a precise meaning equivalent in every human language. In contrast, “understands” is used far less frequently, learned later in life, and features less often in our practice of epistemic as sessment. These facts suggest that knowledge-talk plays a more crucial role in epistemic evaluation than understanding-talk. Knowledge-talk is also more closely tied to tracking norms of assertion and practical reasoning than understanding-talk. This is because knowledge ascriptions (and denials) align with natural assessments of assertion and practical reasoning in ordinary language. As argued in §5.3, it seems ap propriate to challenge assertions by asking the asserter, “How do you know that?” In contrast, we are far less inclined to challenge assertions by asking about understanding. Similarly, we can rightfully criticize a person’s actions or reasoning when that person acted without knowledge; for instance, “You shouldn’t have gone down this street, since you didn’t know that the restau rant was there.” Again, we are less likely to criticize action or practical rea soning by appealing to understanding.21 Thus, knowledge ascriptions seem to play valuable social roles that attributions of understanding do not. Specifically, knowledge ascriptions serve the interrelated functions of identifying reliable informants, typically signaling the appropriate end of inquiry, and providing a threshold-marker that indicates that the epistemic standard that is usually necessary and suffi cient for assertion and practical reasoning has been met. The same cannot be said for attributions of understanding.
9.7 Is Understanding Reducible to Knowledge? If attributions of knowledge and attributions of understanding play different social roles, does this imply that understanding-why is not reducible to knowledge-why? Many scholars have argued that understanding-why is merely a species of knowledge-why, although this
21 Cf. Hills (2016) for the idea that understanding-why has at least as important a role in action as knowledge.
248
|
What’s the Point of Knowledge?
idea is controversial.22 The reductionists claim that to understand some thing is just to know more about it; for instance, to know the reasons why p is true, or to know how it relates to other aspects of the relevant do main. The antireductionists deny this on various grounds; for example, it has been argued that understanding (but not knowledge) is immune to certain forms of epistemic luck,23 and that understanding is more diffi cult to transmit via testimony than knowledge.24 What bearing does my proposal have on this contentious issue? While I think understanding-why is probably reducible to knowledgewhy, I will not argue for this view here. Instead, I will briefly explain why my proposal is compatible with both reductionism and nonreductionism. Proponents of either camp can therefore accept my hypothesis. Suppose you think understanding-why is not reducible to knowledgewhy. If this type of antireductionism were true, it would be unsurprising that understanding attributions and knowledge attributions could play different social roles. In contrast, if understanding-why were nothing more than knowledge-why, then you might doubt whether attributions of knowledge and attributions of understanding really do play different roles in epistemic evaluation. Does the idea that knowledge and under standing play different social roles go hand-in-hand with a commitment to antireductionism? It does not. Even if reductionism were true, a distinction between “knowing” and “understanding” might be drawn because the latter picks out a special kind of knowledge. The point I want to emphasize here is not about the reductive relationship between knowledge and understanding, but rather about the difference in focus when we think and speak in terms of understanding rather than knowing. Knowledge attributions often refer to knowledge-that, whereas the type of knowledge needed for understanding includes, at a minimum, knowledge of the explanation as to why p, knowing
22 Defenders of reductionism include Achinstein (1983), Woodward (2003), Brogaard (2005), Grimm (2006,2014), Greco (2014), Kelp (2014), Sliwa (2015), and Riaz (2015). Antireductionists include Zagzebski (2001), Kvanvig (2003), Elgin (2004, 2007, 2009), Pritchard (2008,2009), Hills (2009, 2016), Riggs (2009), Moms (2012), and de Regt (2015). 23 Kvanvig (2003: 197-198), Pritchard (2008: 37), and Hills (2009: 204) claim that one’s . knowledge can be undermined by luck even though one retains the corresponding understanding. In reply, Sliwa (2015) argues that utterances of the form “I understand why Stalin was evil but I don’t know why Stalin was evil” are clearly infelicitous. She also provides cases of epistemic luck that are structurally identical to the standard cases and yet it is not intuitive that the agent lacks knowledge. This throws into question the robustness of the intuitions in the cases presented by Kvanvig, Pritchard, and Hills (see also Gendler and Hawthorne 2005). 24 Kvanvig (2003), Elgin (2007), Pritchard (2008), Hills (2009), and Strevens (2013) all describe cases of testimony as a challenge to reductionism.
what’s the point of understanding?
|
249
how to elaborate and explain things in one’s own words, and seeing ex planatory connections. It is having this type of knowledge that makes one a good explainer (in the sense outlined in §9.4).25 We attribute understanding when an agent meets a sufficiently high standard of explanatory depth. Meeting this standard could simply require more knowledge (of causes, their relation, etc.), but we mark this standard with the word “understands.”
9.8 Epistemic Luck My proposal might also throw some light on the relationship between knowledge, understanding, and epistemic luck. I will merely gesture at a way to elucidate this issue, however, because the intuitions driving this debate are highly unstable. Indeed, there is no clear consensus on whether there even are cases in which one understands but does not know. On the one hand, Pritchard (2009) and Hills (2016) argue that under standing (but not knowledge) is compatible with “environmental luck.” On the other hand, Gendler and Hawthorne (2005) and Sliwa (2015) show that our intuitions about cases with the same structure are flimsy. Further, Grimm (2006) puts forward some cases against the claim that under standing can be lucky, and he argues that the cases where understanding isn’t undermined by luck are also cases where epistemic luck does not un dermine knowledge.26 For these reasons, the plausibility of my functionfirst account should not turn on its ability to explain the relationship between knowledge, understanding, and luck. Nevertheless, I’ll try to briefly indi cate how my hypothesis might be developed to explain why understanding is compatible with certain forms of knowledge-undermining luck. Imagine Winston is living in an Orwellian state in which the govern ment attempts to falsify the past. Winston is in a room full of elaborately falsified history books, but he luckily picks up the one accurate book that was not destroyed. As he reads the book, he comes to believe many of
25 Kvanvig (2009: 99) makes a similar point. He says attributions of knowledge and attributions of understanding point to two kinds of intellectual achievements, even if the language of understanding can be replaced with the language of knowing. On Kvanvig’s view, “knowledge” focuses on elements such as quality of evidence, ruling out alternative hypotheses, and not being right merely by accident; in contrast, “understands” focuses on “the grasped relatedness of the items that constitute a body of information possessed by the individual in question.” 26 Morris (2012) claims the intuitions that rule against lucky understanding can be explained away. However, Sliwa (2015: 60-61) argues that sentences of the form “I understand why x, but I don’t know why x” are infelicitous, and there is no obvious way to explain their infelicity by appealing to pragmatic considerations.
250
|
What’s the Point of Knowledge?
its true claims about the past. To focus on one claim in particular, sup pose Winston comes to believe that the Comanche dominated the southern plains of North America during the eighteenth century because of their superior horsemanship.27 Let’s suppose Winston grasps this explana tion, it makes sense to him, and so on. Does he now understand why the Comanche dominated the southern plains during this period? According to Kvanvig, Pritchard, and Hills, it seems that Winston does have under standing. Does he also know why the Comanche dominated the southern plains? Some philosophers have been tempted to say that he does not. His beliefs might easily have been mistaken, and one of the main lessons from the Gettier literature is that knowledge requires a more secure connection to the truth. Why might certain forms of epistemic luck undermine knowledge but not understanding? Here’s one possible answer: it might be that certain types of luck undermine an informant’s reliability without necessarily undermining her status as a good explainer. And given the different social roles played by knowledge attributions and understanding attributions, it would follow that in such cases we will be inclined to ascribe under standing but not knowledge. To illustrate, let’s return to our Orwellian scenario. We arguably have a good reason to regard Winston as an unreliable informant: the correla tion between Winston’s justification and his being right about why p was merely accidental. This will affect our attitude toward his reliability. As Craig notes, relying on Winston would “produce the retrospective feeling of having run a risk, of having done something that one would not have done had one been just a little better informed at the time” (1990: 49). It is in our interest as truth-seeking inquirers to want beliefs that are not just luckily or accidentally true. We certainly cannot recommend as a general epistemic policy that people rely on reasoning that luckily culminates in a true conclusion. In standard cases of environmental luck, the way in which the agent reaches a belief impugns his reliability, despite the fact that he holds a true belief. Such a person’s reliability is understood counterfactually: he is unreliable because he could have easily been wrong
27 This example is adapted from Kvanvig (2003: 197-198). It is structurally analogous to Pritchard’s case of the fire officer (2009), which is also used to support the conclusion that understanding is compatible with certain forms of knowledge-undermining luck. Both cases are assumed to be structurally similar to fake-barn cases where one has a true belief that is lucky in the sense that it is unsafe (Goldman 1976). Pritchard classifies these as examples of “environmental luck.”
what’s the point of understanding?
|
251
had things been slightly different. Winston might easily have picked up a book with false claims. In contrast, the same type of luck might not threaten the agent’s status as a good explainer. After all, the lucky reader can still correctly answer a wide range of questions about the Comanche dominance; for instance, by pointing to particular aspects of their horsemanship that brought about the result, and so on.28 To be able to provide a good explanation, one need not worry much about how one came to believe the explanatory proposition. Our main concern, as Kvanvig says, is whether the ability in question is actually present, and we are less concerned with whether the ability came to exist in a chancy way. Thus, what we are focusing on when evaluating an individual as a reliable informant (a knower) differs from what we focus on when we evaluate a person as good explainer (an understander). In the former case, we tend to focus on “external” facts about the eti ology of the belief. Consequently, a more secure connection to the truth seems required. Acquiring the truth by chance, through history books or otherwise, is not enough. But when we focus on understanding, we tend to focus on “internal” elements such as an individual’s ability to see how different causal elements depend upon one another in our representation of the world (Grimm 2014: 343). This is all fairly speculative. I will not pretend to have resolved any issues about the nature of epistemic luck. Rather, I am simply indicating a possible way in which my hypothesis might be developed to illuminate this issue.
9.9 Transparency My hypothesis is also amenable to the idea that understanding is best un derstood along epistemically internalist lines. As Pritchard observes, it is hard to make sense of how an agent could possess understanding and yet lack good reflective accessible grounds in support of that understanding. Understanding thus cannot be “opaque” to the subject in the way that knowledge, by epistemically externalist lights at least, can sometimes be. (2010: 76)
28 If we conceive of understanding as a kind of ability, as suggested by de Regt (2004), Pritchard (2009), and Hills (2016), then we would say this ability is not lost as a result of the luck in question.
252
|
What’s the Point of Knowledge?
Someone who understands why p must be able to say something on behalf of p—to offer reasons for one’s understanding. As Aristotle would put it, one who understands does not just know that such-and-such is the case but also knows “the why” of the thing. A good explainer undoubtedly meets this requirement. Our sense of understanding is of course fallible. One might feel as though one understands why p and yet not really have understanding. A believer in astrology might think he understands why his relation ship failed (i.e., because his wife had a conflicting zodiac sign), but this is not a correct explanation and so does not produce any real under standing. It is possible for one to have more or less understanding than one thinks. An excessively humble or modest individual might believe she understands less than she actually does, yet the true extent of her understanding would be revealed were she encouraged to provide an explanation. Grimm (2017) argues that understanding is not as internalist-friendly as many have supposed. Indeed, he rejects the idea that one needs to have reflectively accessible grounds in support of one’s understanding: the main reason to reject this claim when it comes to understanding is closely tied to—perhaps just identical to—the main reason many have rejected this idea when it comes to knowledge: namely, that understanding (like knowl edge) seems to be available to animals and young children, even though animals and young children apparently lack the metacognitive ability to identify the grounds of their understanding. (2017: 220)
If Grimm is right, then the very same sorts of cases that led epistemologists to conclude that knowledge could sometimes be opaque, unreflective, or inarticulate also show the same thing about understanding. However, the cases on which Grimm focuses are not clearly cases of understanding why. A source of confusion is that Grimm attempt to an alyze understanding without distinguishing the various kinds of under standing. Once we distinguish understanding-why from other forms of understanding, it becomes clear that his examples are not instances of understanding-why. In the case of children, Grimm appeals to recent work by psychologists studying infant cognition. In one study, infants who observed a bean bag being tossed from behind a screen would act surprised if the screen were subsequently removed and nobody was there (Carey 2009: 234-240). The children therefore seemed to have inferred a dependence between
what’s the point of understanding?
|
253
the appearance of the bean bag and human agency. Psychological studies aside, Grimm also asks us to imagine an 11-month-old toddler who sees a toy on top of a stand and who learns through experience that if he pushes the stand, the toy will drop to the floor. According to Grimm, these examples provide a picture of how children “come to understand the world” (2017: 222). These children certainly grasp some dependence relations and make causal inferences, but do they possess understanding-why? It is not clear what the object of such understanding-why would be in these cases. The child who pushes the table might understand that pushing the table is a way to get the toy, but it is doubtful that he understands why this is so. Further, children who are confused by how the bean bag was moving through the air do not seem to understanding “the why” of the situation. The same can be said for Grimm’s examples involving rats, crows, and other animals that seem able to infer causal relationships in ways that go beyond mere condi tioning. In all of these cases, it is plausible that neither young children nor animals understand why such-and-such is the case. Grimm anticipates this reply. He says one might argue there is no real understanding in the cases of young children and animals because (a) “real understanding requires the ability to identify causal mechanisms, or the causal processes actually responsible for bringing about events” (2017: 223), or (b) “it might be thought that genuine understanding requires the ability to subsume events under laws, and that the ability to formulate laws is beyond children and animals” (2017: 223). Grimm rejects (a) on the grounds that it is too demanding to say under standing requires the identification of causal mechanisms. On this point, I wholeheartedly agree. One can understanding why one’s eyes are wa tering (e.g., because one is chopping onions) on the basis of past expe rience, even though one has no clue about the particulars of onion and eyeball chemistry that cause the watering. A really good explainer might be able to identify the relevant causal mechanism, but this is not an es sential component of understanding on my account. In fact, there may be some instances of understanding-why where causal mechanisms are not part of the explanation at all (e.g., in morality or mathematics). Interestingly, Grimm does not reject (b) on the grounds that it is too demanding. Rather, he argues that children and animals can often appeal to generalizations that are relatively invariant or stable across contexts and can subsume particular events under these generalizations. So if we think of these generalizations as laws, then it seems young chil dren and animals are capable of understanding even if they are unable
254
|
What’s the Point of Knowledge?
to grasp the dependency relationships at a high level of abstraction. (And, presumably, if we think that agents must grasp the dependency relationships at a high level of abstraction, then the requirement would be too demanding.) I am happy to concede both of these points to Grimm. Genuine under standing does not require the ability to identify causal mechanisms or grasp dependency relationships at a high level of abstraction. However, Grimm does not consider the possibility that young children and an imals lack understanding-why on the grounds that they are not good explainers.
9.10 Concluding Remarks I make two proposals in this chapter. First, I suggest that we can make headway in the epistemology of understanding by taking a function-first approach. Second, I hypothesize that humans think and speak in terms of “understanding” because it allows us to certify good explainers, which is an important dimension of epistemic evaluation. As cognitively interde pendent explanation-seekers, we need a way to identify informants who can provide us with good explanations for both practical and theoretical purposes. This hypothesis throws light on the nature and importance of under standing, the role of explanation in understanding, and the relationship between understanding and knowledge. I have argued that understanding and knowledge play different social roles, since the latter is not neces sarily geared toward an explanation of why such-and-such is the case. Examining these two cognitive achievements from the point of view of their function also sheds light on epistemic value. Specifically, we see that understanding is valuable and yet knowledge might play a more important role in human survival and flourishing. Roughly, knowledge is closely tied to answering our need for true beliefs, whereas under standing answers our need for good explanations. Inquiry is typically aimed at true beliefs, which is why knowledge matters, but sometimes we need more than just true beliefs to get by in the world. We want to grasp a variety of connections; we want to anticipate what would have happened had things been different; we want to see how things hang together. Thus, understanding is important even though the concerns of practical life will often prevent us from reaching for the highest epistemic fruit.
what’s the point of understanding?
|
255
BIBLIOGRAPHY
Achinstein, P. 1983. The Nature of Explanation. Oxford University Press. Adams, F. 2006. “Intentions Confer Intentionality upon Actions: A Reply to Knobe and Burra.” Journal of Cognition and Culture 6 (1-2): 255-268. Adams, F. and A. Steadman. 2004a. “Intentional Action in Ordinary Language: Core Concept or Pragmatic Understanding?” Analysis 64 (2): 173-181. Adams, F. and A. Steadman. 2004b. “Intentional Action and Moral Considerations: Still Pragmatic.” Analysis 64 (3): 268-276. Ahlstrom, K. 2008. “Epistemology and Empirical Investigation.” Grazer Philosophische Studien 76 (1): 109-134. Alexander, J. and J. Weinberg. 2014. “The ‘Unreliability’ of Epistemic Intuitions.” In E. Machery and E. O’Neill (eds.), Current Controversies in Experimental Philosophy. Routledge. Alston, W. 2000. Illocutionary Acts and Sentence Meaning. Cornell University Press. Anderson, C. 2015. “On the Intimate Relationship of Knowledge and Action.” Episteme 12 (3): 343-353. Arico, A., B. Fiala, R. Goldberg, and S. Nichols. 2011. “The Folk Psychology of Consciousness.” Mind and Language 26 (3): 327-352. Austin, J. L. 1946. “Other Minds.” Proceedings of the Aristotelian Society 148: 148-187. Austin, J. L. 1975. How to Do Things with Words. Clarendon Press. Austin, J. L. 1979. Philosophical Papers. Clarendon Press. Axelrod, R. 1984. The Evolution of Cooperation. Basic Books. Ayer, A. J. 1956. The Problems of Knowledge. Macmillan. Bach, K. 2005. “The Emperor’s New ‘Knows.’” In G. Preyer and G. Peter (eds.), Contextualism in Philosophy: Knowledge, Meaning, and Truth. Oxford University Press. Bach, K. 2010. “Knowledge In and Out of Context.” In J. K. Campbell, M. O.’Rourke, and H. S. Silverstein (eds.), Knowledge and Skepticism. MIT Press. Baz, A. 2012. When Words Are Called For: In Defense of Ordinary Language. Harvard University Press.
257
Beebe, J. 2012. “Social Functions of Knowledge Attributions.” In J. Brown and M. Gerken (eds.), Knowledge Ascriptions. Oxford University Press. Beebe, J. 2013. “A Knobe Effect for Belief Ascriptions.” Review of Philosophy and Psychology 4 (2): 235-258. Beebe, J. and W. Buckwaiter. 2010. “The Epistemic Side-Effect Effect.” Mind and Language 25 (4): 474-498. Benton, M. A. 2016. “Expert Opinion and Second-Hand Knowledge.” Philosophy and Phenomenological Research 92 (2): 492-508. Berry, D. 2011. Infallibilism and the State of Nature. Master’s thesis, University College London. http://discovery.ucl.ac.Uk/id/eprint/l 332885. Block, N. 1978. “Troubles with Functionalism.” In C. W. Savage (ed.), Minnesota Studies in the Philosophy of Science, 9: 261-325. University of Minnesota Press. BonJour, L. 2006. “Komblith on Knowledge and Epistemology.” Philosophical Studies 127 (2): 317-335. BonJour, L. 2010. “The Myth of Knowledge.” Philosophical Perspectives 24 (1): 57-83. Bourget, D. and D. Chalmers. 2014. “What Do Philosophers Believe?” Philosophical Studies 170 (3): 465-500. Boyd, K. 2017. “Testifying Understanding.” Episteme 14 (1): 103-127. Boyd, K. and J. Nagel. 2014. “The Reliability of Epistemic Intuitions.” In E. Machery andE. O’Neill (eds.), Current Controversies in Experimental Philosophy. Routledge. Brandom, Robert. 2000. Articulating Reasons: An Introduction to Inferentialism. Harvard University Press. Bridges, J. 2008. “Pulling Semantic Contextualism Out by Its Roots.” Unpublished manuscript. Brogaard, B. 2005. “I Know, Therefore I Understand.” Unpublished manuscript. Brown, J. 2005. “Comparing Contextualism and Invariantism on the Correctness of Contextualist Intuitions.” Grazer Philosophische Studien 69 (1): 71-100. Brown, J. 2006. “Contextualism and Warranted Assertability Maneuvers.” Philosophical Studies 130 (3): 407^35. Brown, J. 2008a. “Knowledge and Practical Reason.” Philosophy Compass 3 (6): 1135-1152. Brown, J. 2008b. “Subject-Sensitive Invariantism and the Knowledge Norm for Practical Reasoning.” Nous 42 (2): 167-189. Brown, J. 2012. “Assertion and Practical Reasoning: Common or Divergent Epistemic Standards?” Philosophy and Phenomenological Research 84 (1): 123-157. Brown, J. 2014. “Impurism, Practical Reasoning, and the Threshold Problem.” Nous 48 : 179-192. (1) Brown, J. 2018. Fallibilism: Evidence and Knowledge. Oxford University Press. Buckwaiter, W. 2010. “Knowledge Isn’t Closed on Saturday: A Study in Ordinary Language.” Review of Philosophy and Psychology 1 (3): 395-4-06. Buckwaiter, W. and S. Stich. 2011. “Gender and the Philosophy Club.” Philosophers’ Magazine 52 (52): 60-65. Buckwaiter, W. and S. Stich. 2014. “Gender and Philosophical Intuition.” In J. Knobe and S. Nichols (eds.), Experimental Philosophy, vol. 2. Oxford University Press. Butchvarov, P. 1970. The Concept of Knowledge. Northwestern University Press. Cappelen, H. and E. Lepore. 2003. “Context Shifting Arguments.” Philosophical Perspectives 17 (1): 25-50.
258
|
Bibliography
Carey, S. 2009. The Origin of Concepts. Oxford University Press. Cariani, F. 2011. “Judgment Aggregation.” Philosophy Compass 6 (1): 22-32. Carnap, R. 1950. Logical Foundations of Probability. University of Chicago Press. Cassam, Q. 2007. The Possibility of Knowledge. Oxford University Press. Cavell, S. 1965. “Austin at Criticism.” Philosophical Review 74 (2): 204-219. Cavell, S. 1979. The Claim of Reason: Wittgenstein, Skepticism, Morality, and Tragedy. Oxford University Press. Chalmers, D. 2011. “Verbal Disputes.” Philosophical Review 120 (4): 515-566. Chisholm, R. 1957. Perceiving: A Philosophical Study. Cornell University Press. Chisholm, R. 1964. “J. L. Austin’s Philosophical Papers.” Mind 73 (289): 1-26. Chisholm, R. 1977. Theory of Knowledge. Prentice Hall. Chrisman, M. 2010. “From Epistemic Expressivism to Epistemic Inferentialism.” In A. Haddock, A. Millar, and D. Pritchard (eds.), Social Epistemology. Oxford University Press. Clements, W. A. and J. Pemer. 1994. “Implicit Understanding of Belief.” Cognitive Development 9: 377-395. Cocker, M. 1998. Rivers of Blood, Rivers of Gold: Europe’s Conflict with Tribal Peoples. Jonathan Cape. Cohen, L. J. 1995. Essay on Belief and Acceptance. Oxford University Press. Cohen, S. 1988. “How to Be a Fallibilist.” Philosophical Perspectives 2: 91-123. Cohen, S. 1999. “Contextualism, Skepticism, and the Structure of Reasons.” Philosophical Perspectives 13: 57-89. Cohen, S. 2000. “Contextualism and Skepticism.” Nous 34 (1): 94-107. Cohen, S. 2005. “Knowledge, Speaker, and Subject.” Philosophical Quarterly 55 (219): 199-212. Conee, E. and R. Feldman. 2004. Evidentialism. Oxford University Press. Craig, E. 1990. Knowledge and the State of Nature. Oxford University Press. Craig, E. 2007. “Genealogies and the State of Nature.” In A. Thomas (ed.), Bernard Williams. Cambridge University Press. Craik, K. 1943. The Nature of Explanations. Cambridge University Press. Cushman, F. and A. Mele. 2008. “Intentional Action: Two-and-a-Half Folk Concepts?” In J. Knobe and S. Nichols (eds.), Experimental Philosophy. Oxford University Press. Dancy, J. 1985. An Introduction to Contemporary Epistemology. Blackwell. Dancy, J. 1992. “Review of Knowledge and the State of Nature.” Philosophical Quarterly 42 (186): 393-395. Davies, M. and D. Gardner. 2010. Frequency Dictionary ofAmerican English. Routledge. Davis, W. 2007. “Knowledge Claims and Context: Loose Use.” Philosophical Studies 132 (3): 395-438. DePaul, M. 2009. “Ugly Analyses and Value.” In A. Haddock, A. Millar, andD. Pritchard (eds.), Epistemic Value. Oxford University Press. de Regt, H. 2004. “Discussion Note: Making Sense of Understanding.” Philosophy of Science 71 (1): 98-109. de Regt, H. 2009. “Understanding and Scientific Explanation.” In H. de Regt, S. Leonelli, and K. Eiger (eds.), Scientific Understanding: Philosophical Perspectives. University of Pittsburgh Press. de Regt, H. 2015. “Scientific Understanding: Truth or Dare?” Synthese 192 (12): 3781-3797.
Bibliography
|
259
DeRose, K. 1992. “Contextualism and Knowledge Attributions.” Philosophy and Phenomenological Research 52 (4): 913-929. DeRose, K. 1995. “Solving the Skeptical Problem.” Philosophical Review 104 (1): 1-52. DeRose, K. 1999. “Contextualism: An Explanation and Defense.” In J. Greco and E. Sosa (eds.), Blackwell Guide to Epistemology. Blackwell. DeRose, K. 2002. “Knowledge, Assertion, and Context.” Philosophical Review 111 (2) : 167-203. DeRose, K. 2005. “The Ordinary Language Basis for Contextualism, and the New Invariantism.” Philosophical Quarterly 55 (219): 172-198. DeRose, K. 2009. The Case for Contextualism. Oxford University Press. DeRose, K. 2011. “Contextualism, Contrastivism, and X-Phi Surveys.” Philosophical Studies 156(1): 81-110. Descartes, R. 2006. Meditations, Objections, and Replies. Hackett. Dinges, A. 2016. “Skeptical Pragmatic Invariantism: Good, but Not Good Enough.” Synthese 193 (8): 2577-2593. Dogramaci, S. 2012. “Reverse Engineering Epistemic Evaluations.” Philosophy and Phenomenological Research 84 (3): 513—530. Douven, I. 2006. “Assertion, Knowledge, and Rational Credibility.” Philosophical Review 115 (4): 449-485. Dretske, F. 1970. “Epistemic Operators.” Journal of Philosophy 61 (24): 1007-1023. Dretske, E 1981. “The Pragmatic Dimension of Knowledge.” Philosophical Studies 40 (3) : 363-378. Elgin, C. 2004. “True Enough.” Philosophical Issues 14 (1): 113-131. Elgin, C. 2006. “From Knowledge to Understanding.” In S. Hetherington (ed.), Epistemology Futures. Oxford University Press. Elgin, C. 2007. “Understanding and the Facts.” Philosophical Studies 132 (1): 33-42. Elgin, C. 2009. “Is Understanding Factive?” In A. Haddock, A. Millar, and D. Pritchard (eds.), Epistemic Value. Oxford University Press. Elgin, C. 2017. True Enough. Oxford University Press. Fantl, J. and M. McGrath. 2009. Knowledge in an Uncertain World. Oxford University Press. Feldman, R. 1995. “In Defense of Closure.” Philosophical Quarterly 45 (181): 487—494. Feldman, R. 1997. “Review Essay: Knowledge and the Stated of Nature: An Essay in Conceptual Synthesis.” Philosophy and Phenomenological Research 57 (1): 205-221. Feltz, A. and C. Zarpentine. 2010. “Do You Know More When It Matters Less?” Philosophical Psychology 23 (5): 683-706. Fodor, J. 2003. Hume Variations. Oxford University Press. Fogelin, R. 1994. Pyrrhonian Reflections on Knowledge and Justification. Oxford University Press. Fricker, E. 2015. “Know First, Tell Later: The Truth about Craig on Knowledge.” In J. Greco and D. Henderson (eds.), Epistemic Evaluations. Oxford University Press. Fricker, M. 1998. “Rational Authority and Social Power: Towards a Truly Social Epistemology.” Proceedings of the Aristotelian Society 98 (2): 159-177. Fricker, M. 2007. Epistemic Injustice. Oxford University Press. Fricker, M. 2008. “Skepticism and the Genealogy of Knowledge: Situating Epistemology in Time.” Philosophical Papers 37 (1): 27-50.
260
|
Bibliography
Fricker, M. 2012. “Group Testimony? The Making of a Collective Good Informant.” Philosophy and Phenomenological Research 84 (2): 249-276. Gelfert, A. 2011. “Steps to an Ecology of Knowledge: Continuity and Change in the Genealogy of Knowledge.” Episteme 8 (1): 67-82. Gendler, T. and J. Hawthorne. 2005. “The Real Guide to Fake Bams: A Catalogue of Gifts for Your Epistemic Enemies.” Philosophical Studies 124 (3): 331-352. Gentzler, J. 1995. “How to Discriminate between Experts and Frauds.” History of Philosophy Quarterly 12 (3): 227-246. Gerken, M. 2011. “Warrant and Action.” Synthese 178 (3): 529-547. Gerken, M. 2013. “Epistemic Focal Bias.” Australasian Journal of Philosophy 91 (1): 41-61. Gerken, M. 2014. “Same, Same but Different: The Epistemic Norms of Assertion, Action and Practical Reasoning.” Philosophical Studies 168 (3): 725-744. Gerken, M. 2015. “The Roles of Knowledge Ascriptions in Epistemic Assessment.” European Journal of Philosophy 23 (1): 141-161. Gerken, M. 2017. On Folk Epistemology. Oxford University Press. Gettier, E. 1963. “Is Justified True Belief Knowledge?” Analysis 23 (6): 121-123. Gijsbers, V. 2013. “Understanding, Explanation, and Unification.” Studies in History and Philosophy of Science 44 (3): 516-522. Gilbert, M. 1989. On Social Facts. Routledge. Gilbert, M. and D. Pilchman 2014. “Belief, Acceptance, and What Happens in Groups.” In J. Lackey (ed.), Essays in Collective Epistemology. Oxford University Press. Glock, H. 2010. “The Development of Analytic Philosophy: Wittgenstein and After.” In D. Moran (ed.), The Routledge Companion to Twentieth Century Philosophy. Routledge. Goddard, C. 2010. “Universals and Variation in the Lexicon of Mental State Concepts.” In B. Malt and P. Wolff (eds.), Words and the Mind. Oxford University Press. Goldman, A. 1967. “A Causal Theory of Knowing.” Journal of Philosophy 64 (12): 357-372. Goldman, A. 1976. “Discrimination and Perceptual Knowledge.” Journal of Philosophy 73: 771-791. Goldman, A. 1998. Epistemology and Cognition. Harvard University Press. Goldman, A. 2004. “Group Knowledge versus Group Rationality: Two Approaches to Social Epistemology.” Episteme 1 (1): 11-22. Gopnik, A. 2000. “Explanation as Orgasm and the Drive for Causal Understanding: The Evolution, Function and Phenomenology of the Theory-Formation System.” In F. Keil and R. Wilson (eds.), Cognition and Explanation. MIT Press. Greco, D. and B. Hedden. 2016. “Uniqueness and Metaepistemology.” Journal of Philosophy 113 (8): 365-395. Greco, J. 2003. “Knowledge as Credit for True Belief.” In M. DePaul and L. Zagzebski (eds.), Intellectual Virtue: Perspectives from Ethics and Epistemology. Oxford University Press. Greco, J. 2007. “The Nature of Ability and the Purpose of Knowledge.” Philosophical Issues 17 (1): 57-69. Greco, J. 2008. “What’s Wrong with Contextualism?” Philosophical Quarterly 58 (232): 416-136.
Bibliography |
261
Greco, J. 2009. “Knowledge and Success from Ability.” Philosophical Studies 142 (1): 17-26. Greco, J. 2014. “Episteme: Knowledge and Understanding.” In K. Timpe and C. Boyd (eds.), Virtues and Their Vices. Oxford University Press. Greco, J. and D. Henderson. 2015. Epistemic Evaluation: Purposeful Epistemology. Oxford University Press. Grice, H. P. 1961. “The Causal Theory of Perception.” Proceedings of the Aristotelian Society: 121-153. Grice, H. P. 1989. Studies in the Way of Words. Harvard University Press. Grimm, S. 2006. “Is Understanding a Species of Knowledge?” British Journal for the Philosophy of Science 57 (3): 515-535. Grimm, S. 2008. “Epistemic Goals and Epistemic Values.” Philosophy and Phenomenological Research 71 (3): 725-744. Grimm, S. 2011. “On Intellectualism in Epistemology.” Mind 120 (479): 705-733. Grimm, S. 2012. “The Value of Understanding.” Philosophy Compass 7 (2): 103-117. Grimm, S. 2014. “Understanding as Knowledge of Causes.” In A. Fairweather (ed.), Virtue
Epistemology Naturalized: Bridges between Virtue Epistemology and Philosophy of Science. Synthese Library, Springer. Grimm, S. 2015a. “Wisdom.” Australasian Journal of Philosophy 93 (1): 1-16. Grimm, S. 2015b. “Knowledge, Practical Interests, and Rising Tides.” In J. Greco and D. Henderson (eds.), Epistemic Evaluations. Oxford University Press. Grimm, S. 2017. “Understanding and Transparency.” In S. Grimm, C. Baumberger, and S. Ammon (eds.), Explaining Understanding: New Essays in Epistemology and the Philosophy of Science. Routledge. Hakli, R. 2007. “On the Possibility of Group Knowledge without Belief.” Social Epistemology 21 (3): 249-266. Hanfling, O. 2000. Philosophy and Ordinary Language. Routledge. Hannon, M. 2013. “The Practical Origins of Epistemic Contextualism.” Erkenntnis 78 : 899-919. (4) Hannon, M. 2014. “Fallibilism and the Value of Knowledge.” Synthese 191: 1119-1146. Hannon, M. 2015a. “The Importance of Knowledge Ascriptions.” Philosophy Compass 10 (12): 856-866. Hannon, M. 2015b. “Stabilizing Knowledge.” Pacific Philosophical Quarterly 96 (1): 116-139. Hannon, M. 2015c. “The Universal Core of Knowledge.” Synthese. 192 (3): 769-786. Hannon, M. 2017a. “A Solution to Knowledge’s Threshold Problem.” Philosophical Studies 174 (3): 607-629. Hannon, M. 2017b. “Skepticism and Contextualism.” In J. J. Ichikawa (ed.), Routledge Handbook of Epistemic Contextualism. Taylor and Francis. Hannon, M. 2017c. “Skepticism about Meta-skepticism: Meditations on Experimental Philosophy.” Episteme 14 (2): 213-231. Hannon, M. 2019. “Why Purists Should Be Infallibilists.” Philosophical Issues. Harman, G. 1973. Thought. Princeton University Press. Haslanger, S. 1999. “What Knowledge Is and What It Ought to Be: Feminist Values and Normative Epistemology.” Philosophical Perspectives 13 (13): 459-480. Haslanger, S. 2012. Resisting Reality: Social Construction and Social Critique. Oxford University Press.
262
|
Bibliography
Haspelmath, M. and U. Tadmor. 2009. Loanwords in the World’s Languages: A Comparative Handbook. Mouton de Gruyter. Hawthorne, J. 2004. Knowledge and Lotteries. Oxford University Press. Hawthorne, J. and J. Stanley. 2008. “Knowledge and Action.” Journal of Philosophy 105 (10): 571-590. Hazlett, A. 2017. “Understanding and Structure.” In S. Grimm (ed.), Making Sense of the World: New Essays in the Philosophy of Understanding. Oxford University Press. Heider, F and M. Simmel. 1944. “An Experimental Study of Apparent Behavior.” American Journal of Psychology. 57 (2): 243-259. Heller, M. 1999. “The Proper Role for Contextualism in an Anti-luck Epistemology.” Philosophical Perspectives 13: 115-129. Hempel, C. 1962. “Explanation in Science and in History.” In R. G. Colodny (ed.), Frontiers of Science and Philosophy. University of Pittsburgh Press. Hempel, C. 1965. Aspects of Scientific Explanation and Other Essays in the Philosophy of Science. Free Press. Hempel, C. and P. Oppenheim. 1948. “Studies in the Logic of Explanation.” Philosophy of Science 15 (2): 135-175. Henderson,D. 2009. “Motivated Contextualism.” Philosophical Studies 142(1): 119-131. Henderson, D. 2011. “Gate-Keeping Contextualism.” Episteme 8 (1): 83-98. Henderson, D. and T. Horgan. 2015. “What’s the Point?” In D. Henderson and J. Greco (eds.), Epistemic Evaluation: Purposeful Epistemology, 87-114. Oxford University Press. Hetherington, S. 2001. Good Knowledge, Bad Knowledge: Two Dogmas ofEpistemology. Oxford University Press. Hetherington, S. 2006. “Knowledge’s Boundary Problem.” Synthese 150 (1): 41-56. Hills, A. 2009. “Moral Testimony and Moral Epistemology.” Ethics 120 (1): 94-127. Hills, A. 2016. “Understanding Why” Nous 49 (2): 661-688. Huebner, B. 2013. “Cognitive Systems and the Extended Mind.” Philosophical Psychology 26 (2): 315-318. Huebner, B. 2016. “The Group Mind in Commonsense Psychology.” In J. Sytsma and W. Buckwaiter (eds.), A Companion to Experimental Philosophy. Routledge. Ichikawa, J. J. 2012. “Knowledge Norms and Acting Well.” Thought 1 (1): 49-55. Ichikawa, J. J. and C. J. Jenkins. 2017. “On Putting Knowledge ‘First.’” In J. A. Carter, E. C. Gordon, and B. Jarvis (eds.), Knowledge First: Approaches in Epistemology and Mind. Oxford University Press. Jackson, F. 1998. From Metaphysics to Ethics. Oxford University Press. Jackson, F. 2011. “On Gettier Holdouts.” Mind and Language 26 (4): 468-481. Jones, W. 1997. “Why Do We Value Knowledge?” American Philosophical Quarterly 34 (4): 423-439. Kant, I. 1960. Observations on the Feeling of the Beautiful and Sublime. Translated by John T. Goldthwait. University of California Press. Kaplan, M. 1985. “It’s Not What You Know That Counts.” Journal of Philosophy 82 (7): 350-363. Kaplan, M. 2000. “To What Must an Epistemology Be True?” Philosophy and Phenomenological Research 61 (2): 279-304. Kaplan, M. 2003. “Who Cares What You Know?” Philosophical Quarterly 53 (210): 105-116.
Bibliography
|
263
Kaplan, M. 2006. “If You Know You Can’t Be Wrong.” In S. Hetherington (ed.), Epistemology Futures. Oxford University Press. Kaplan, M. 2008. “Austin’s Way with Skepticism.” In J. Greco (ed.), The Oxford Handbook of Skepticism. Oxford University Press. Kaplan, M. 2011. “Tales of the Unknown: Austin and the Argument from Ignorance.” In M. Gustafson and R. S0rli (eds.), The Philosophy of J.L. Austin. Oxford University Press. Kappel, K. 2010. “On Saying That Someone Knows: Themes from Craig.” In A. Haddock, A. Millar, A., and D. Pritchard (eds.), Social Epistemology. Oxford University Press. Keil, F. C. 2006. “Explanation and Understanding.” Annual Review of Psychology 57: 227-254. Kelp, C. 2011. “What’s the Point of ‘Knowledge’ Anyway?” Episteme 8: 53-66. Kelp, C. 2014. “Knowledge, Understanding, and Virtue.” In A. Fairweather (ed.), Virtue Epistemology Naturalized: Bridges between Virtue Epistemology and Philosophy of Science. Synthese Library, Springer. Khalifa, K. 2011. “Understanding, Knowledge, and Scientific Antirealism.” Grazer Philosophische Studien 83 (1): 93—112. Khalifa, K. 2012. “Inaugurating Understanding or Repackaging Explanation?” Philosophy of Science 79 (1): 15-37. Khalifa, K. 2013. “Understanding, Grasping, and Luck.” Episteme 10 (1): 1-17. Kim, J. 1994. “Explanatory Knowledge and Metaphysical Dependence.” Philosophical Issues 5: 51-69. Kim, J. 1996. Philosophy of Mind. Westview Press. Kim, M. and Y. Yuan. n.d. “Cross-Cultural Universality of Knowledge Attributions.” Unpublished manuscript. Kitcher, P. 1989. “Explanatory Unification and the Causal Structure of the World.” In P. ’ Kitcher and W. Salmon (eds.), Scientific Explanation. University of Minnesota Press. Kitcher, P. 1993. The Advancement of Science. Oxford University Press. Knobe, J. 2003. “Intentional Action and Side Effects in Ordinary Language.” Analysis 63 (279): 190-194. Koriat, A. 2012. “The Self-Consistency Model of Subjective Confidence.” Psychological Review 119 (1): 80-113. Komblith, H. 2002. Knowledge and Its Place in Nature. Oxford University Press. Komblith, H. 2005. “Replies to Alvin Goldman, Martin Kusch and William Talbott.” Philosophy and Phenomenological Research 71 (2): 427—441. Komblith, H. 2006. “Appeals to Intuition and the Ambitions of Epistemology.” In S. Hetherington (ed.), Epistemology Futures. Oxford University Press. Komblith, H. 2011. “Why Should We Care about the Concept of Knowledge?” Episteme 8 (1): 38-52.2011. Kukla, R. 2015. “Delimiting the Proper Scope of Epistemology.” Philosophical Perspectives 29 (1): 202-216. Kusch, M. 2009. “Testimony and the Value of Knowledge.” In D. Pritchard, A. Haddock, and A. Miller (eds.), Epistemic Value. Oxford University Press. Kusch, M. 2011. “Knowledge and Certainties in the Epistemic State of Nature.” Episteme 8: 6-23. Kusch, M. and R. McKenna. 2018. “The Genealogical Method in Epistemology.” Synthese: 1-20.
264
|
Bibliography
Kvanvig, J. 2003. The Nature and Value of Knowledge. Cambridge University Press. Kvanvig, J. 2009. “The Value of Understanding.” In D. Pritchard, A. Haddock, and A. Miller (eds.), Epistemic Value. Oxford University Press. LaBarge, S. 1997. “Socrates and the Recognition of Experts.” Apeiron 30 (4): 51-62. Lackey, J. 2007a. “Why We Don’t Deserve Credit for Everything That We Know?” Synthese 158 (3): 345-361. Lackey, J. 2007b. “Norms of Assertion.” Nous 41 (4): 594-626. Lackey, J. 2011. “Assertion and Isolated Second-hand Knowledge.” In J. Brown and H. Cappelen (eds.), Assertion: New Philosophical Essays, 251-276. Oxford University Press. Lackey, J. 2012. “Group Knowledge Attributions.” In J. Brown and M. Gerken (eds.), New Essays on Knowledge Ascriptions. Oxford University Press. Lackey, J. 2016. “Assertion and Expertise.” Philosophy and Phenomenological Research 92 (2): 509-517. Lawlor, K. 2005. “Enough Is Enough: Pretense and Invariance in the Semantics of ‘Knows That’.” Philosophical Perspectives 19 (1): 211-236. Lawlor, K. 2013. Assurance. Oxford University Press. Lawlor, K. 2015. “Precis of Assurance.” Philosophy and Phenomenological Research 90 (1): 194—204. Lehrer, K. 1965. “Knowledge, Truth and Evidence.” Analysis 25: 168-175. Lehrer, K. 1990. Theory of Knowledge. Routledge. Lewis, C. I. 1946. An Analysis of Knowledge and Valuation. Open Court. Lewis, D. 1983. Philosophical Papers. Oxford University Press. Lewis, D. 1996. “Elusive Knowledge.” Australasian Journal of Philosophy (4): 549-567. Lipton, P. 2009. “Understanding without Explanation.” In H. de Regt, S. Leonelli, and K. Eiger (eds.), Scientific Understanding: Philosophical Perspectives. University of Pittsburgh Press. List, C. 2005. “Group Knowledge and Group Rationality: A Judgment Aggregation Perspective.” Episteme 2 (1): 25-38. Locke, J. 1975. An Essay Concerning Human Understanding. Oxford University Press. Lombrozo, T. 2011. “The Instrumental Value of Explanations.” Philosophy Compass 6 (8): 539-551. Longino, H. 1994. The Fate of Knowledge. Princeton University Press. Lycan, W. 2006. “On the Gettier Problem Problem.” In S. Hetherington (ed.), Epistemology Futures. Oxford University Press. Machery, E., S. Stich, D. Rose, A. Chatterjee, K. Karasawa, N. Struchiner, S. Sirker, N. Usui, andT. Hashimoto. 2015. “Gettier across Cultures.” Nous 51 (3): 645-664. Malcolm, N. 1949. “Defending Common Sense.” Philosophical Review 58 (3): 201-220. Malle, B. 2010. “Intentional Action in Folk Psychology.” In T. O’Connor and C. Sandis (eds.), A Companion to the Philosophy ofAction. Wiley Blackwell. Mates, B. 1996. The Skeptic Way. Oxford University Press. May, J., W. Sinnott-Armstrong, J. Hull, and A. Zimmerman. 2010. “Practical Interests, Relevant Alternatives, and Knowledge Attributions: An Empirical Study.” Review of Philosophy and Psychology 1 (2): 265-273. Maynard Smith, J. 1982. Evolution and the Theory of Games. Cambridge University Press.
Bibliography
|
265
Mayseless, O. and A. Kruglanski. 1987. “What Makes You So Sure? Effects of Epistemic Motivations on Judgmental Confidence.” Organisational Behavior and Human Decision Processes 39: 162-83. McBain, J. 2004. “Epistemic Analysis and the Possibility of Good Informants.” Principia 8(2): 193-211. McCann, H. 2005. “Intentional Action and Intending: Recent Empirical Studies.” Philosophical Psychology 18 (6): 737-748. McGinn, M. 1989. Sense and Certainty: A Dissolution of Scepticism. Blackwell. McKenna, R. 2013. “‘Knowledge’ Ascriptions, Social Roles, and Semantics.” Episteme 10 (4): 335-350. McMahon, C. 2003. “Two Modes of Collective Belief.” Protosociology 18:347-362. Medina, J. 2011. “The Relevance of Credibility Excess in a Proportional View of Epistemic Injustice: Differential Epistemic Authority and the Social Imaginary.” Social Epistemology 25 (1): 15-35. Mercier, H. 2011. “On the Universality of Argumentative Reasoning.” Journal of Cognition and Culture 11: 85-113. Millar, A. 2007. “The State of Knowing.” Philosophical Issues 17 (1): 179-196. Millikan, R. 1984. Thought, Language, and Other Biological Categories. MIT Press. Montminy, M. 2009. “Contextualism, Invariantism and Semantic Blindness.” Australasian Journal of Philosophy 87 (4): 639-657. Montminy, M. 2013. “Why Assertion and Practical Reasoning Must Be Governed by the Same Epistemic Norm.” Pacific Philosophical Quarterly 94 (1): 57-68. Moran, R. 2005. “Getting Told and Being Believed.” Philosophers’ Imprint 5 (5): 1-29. Morris, K. 2012. “A Defense of Lucky Understanding.” British Journalfor the Philosophy of Science 63 (2): 357-371. Moser, P. 1990. “Two Roads to Skepticism.” In M. D. Roth and G. Ross (eds.), Doubting: Contemporary Perspectives on Skepticism. Kluwer Academic Publishers. Myers-Schulz, B. and E. Schwitzgebel. 2013. “Knowing That P without Believing That P.” Nous 47 (2): 371-384. Nagel, J. 2010a. “Knowledge Ascriptions and the Psychological Consequences of Thinking about Error.” Philosophical Quarterly 60 (239): 286-306. Nagel, J. 2010b. “Epistemic Anxiety and Adaptive Invariantism.” Philosophical Perspectives 24: 407-435. Nagel, J. 2013. “Defending the Evidential Value of Epistemic Intuitions: A Reply to Stich.” Philosophy and Phenomenological Research 86 (1): 179-199. Nagel, J. 2014. Knowledge: A Very Short Introduction. Oxford University Press. Nagel, J., S. Juan, and R. Mar. 2013. “Lay Denial of Knowledge for Justified True Beliefs.” Cognition 129 (3): 652-666. Nagel, T. 1986. The View from Nowhere. Oxford: Oxford University Press. Neta, R. 2006. “Epistemology Factualized: New Contractarian Foundations for Epistemology.” Synthese 150 (2): 247-280. Neta, R. 2011. “A Refutation of Cartesian Fallibilism.” Nous 45 (4): 658-695. Nichols, S., S. Stich, and J. Weinberg. 2003. “Metaskepticism: Meditations in Ethnoepistemology.” In S. Luper (ed.), The Skeptics. Ashgate. Nozick, R. 1981. Philosophical Explanations. Harvard University Press.
266
|
Bibliography
Partee, B. 2004. “Comments on Jason Stanley’s ‘On the Linguistic Basis for Contextualism.’” Philosophical Studies 119 (1-2): 147-159. Pemer, J. and T. Ruffman. 2005. “Infants’ Insight into the Mind: How Deep?” Science 308: 214-216. Pemu, T. 2009. “Is Knowledge a Natural Kind?” Philosophical Studies 142 (3): 371-386. Pettit, D. and J. Knobe. 2009. “The Pervasive Impact of Moral Judgment.” Mind and Language 24 (5): 586-604. Pettit, P. 2003. “Groups with Minds of Their Own.” In F. Schmitt (ed.), Socializing Metaphysics. Rowman and Littlefield. Pinillos, A. 2012. “Knowledge, Experiments and Practical Interests.” In J. Brown and M. Gerken (eds.), Knowledge Ascriptions. Oxford University Press. Price, H. 2011. Naturalism without Mirrors. Oxford University Press. Pritchard, D. 2008. “Radical Skepticism, Epistemic Luck, and Epistemic Value.” Aristotelian Society Supplementary Volume 82 (1): 19-41. Pritchard, D. 2009. “Knowledge, Understanding and Epistemic Value.” In A. O’Hear (ed.), Epistemology, Royal Institute of Philosophy Supplements, 64: 19-^13. Cambridge University Press. Pritchard, D. 2010. “Knowledge and Understanding.” In D. Pritchard, A. Millar, and A. Haddock, (eds.), The Nature and Value of Knowledge: Three Investigations. Oxford University Press. Pritchard, D. 2012. “Anti-luck Virtue Epistemology.” Journal of Philosophy 109 (3): 247-279. Pritchard, D. 2016. “Seeing It for Oneself: Perceptual Knowledge, Understanding, and Intellectual Autonomy.” Episteme 13 (1): 29^4-2. Pynn, G. 2017. “The Intuitive Basis for Contextualism.” In J. J. Ichikawa (ed.), The Routledge Handbook of Epistemic Contextualism. Routledge. Quinton, A. 1975. “Social Objects.” Proceedings of the Aristotelian Society 76: 1-27. Radford, C. 1966. “Knowledge—by Examples.” Analysis 27 (1): 1-11. Railton, P. 1981. “Probability, Explanation, and Information.” Synthese 48 (2): 233-256. Recanati, F. 2004. Literal Meaning. Cambridge University Press. Reed, B. 2010. “A Defense of Stable Invariantism.” Nous 44 (2): 224-244. Reynolds, S. 2002. “Testimony, Knowledge, and Epistemic Goals.” Philosophical Studies 110(2): 139-161. Riaz, A. 2015. “Moral Understanding and Knowledge.” Philosophical Studies 172 (1): 113-128. Richerson, P.J., R. T. Boyd, and J. Henrich, 2003. “Cultural Evolution of Human Cooperation.” In P. Hammerstein (ed.), Genetic and Cultural Evolution of Cooperation. MIT Press. Riggs, W. 2009. “Understanding, Knowledge, and the Meno Requirement.” In A. Haddock, A. Millar, and D. Pritchard (eds.), Epistemic Value. Oxford University Press. Rorty, R. 1979. Philosophy and the Mirror of Nature. Princeton University Press. Rorty, R. 1982. Consequences of Pragmatism. Cambridge University Press. Rorty, R. 1995. “Is Truth a Goal of Enquiry?” Philosophical Quarterly 45 (180): 281-300. Roth, P. 2003. “Review of Hilary Komblith, Knowledge and Its Place in Nature.” Notre Dame Philosophical Reviews 2003 (12).
Bibliography
|
267
Rozenbfit, L. and F. Keil. 2002. “The Misunderstood Limits of Folk Science: An Illusion of Explanatory Depth.” Cognitive Science 26: 521-562. Rupert, R. 2005. “Minding One’s Cognitive Systems: When Does a Group of Minds Constitute a Single Cognitive Unit?” Episteme 1 (3): 177-188. Russell, B. 1912. The Problems of Philosophy. Williams and Norgate. Russell, B. 1914. Our Knowledge of the External World. Open Court. Russell, B. 1948. Human Knowledge: Its Scope and Limits. George Allen and Unwin. Russell, G. K. and J. M. Doris. 2008. “Knowledge by Indifference.” Australasian Journal of Philosophy 86 (3): 429-437. Rysiew, P. 2001. “The Context-Sensitivity of Knowledge Attributions.” Nous 35 (4): 477-514. Rysiew, P. 2007. “Speaking of Knowing.” Nous 41 (4): 627-662. Rysiew, P. 2012. “Epistemic Scorekeeping.” In J. Brown andM. Gerken (eds.), Knowledge Ascriptions. Oxford University Press. Salmon, W. 1984. “Scientific Explanation: Three Basic Conceptions.” PSA: Proceedings of the Biennial Meeting of the Philosophy of Science Association: 293-305. Schaffer, J. and J. Knobe. 2012. “Contrastive Knowledge Surveyed.” Nous 46 (4): 675-708. Schiffer, S. 1996. “Contextualist Solutions to Skepticism.” Proceedings of the Aristotelian Society 93: 317-333. Schmitt, F. 1992. “Review of Knowledge and the State of Nature: An Essay in Conceptual Synthesis.” Mind 101 (403): 555-559. Schmitt, F. 1994. Socializing Epistemology: The Social Dimensions of Knowledge. Rowman and Littlefield. Seyedsayamdost, H. 2015. “On Normativity and Epistemic Intuitions: Failure of Replication.” Episteme 12 (1): 95-116. Shapin, S. 1995. A Social History of Truth. University of Chicago Press. Shatz, M., H. M. Wellman, and S. Silber. 1983. “The Acquisition of Mental Verbs: A Systematic Investigation of the First Reference to Mental State.” Cognition 14: 301-314. Shope, R. 1983. The Analysis of Knowing: A Decade of Research. Princeton University Press. Siegel, H. 1997. Rationality Redeemed? Routledge. Sliwa, P. 2015. “Understanding and Knowing.” Proceedings of the Aristotelian Society 115(1): 57-74. Smithies, D. 2012. “The Normative Role of Knowledge.” Nous 46 (2): 265-288. Smithies, D. 2015. “Why Justification Matters.” In David Henderson and John Greco (eds.), Epistemic Evaluation: Point and Purpose in Epistemology, 224-244. Oxford University Press. Sosa, E. 1974. “How Do You Know‘1” American Philosophical Quarterly 11: 113-122. Sosa, E. 1988. “Knowledge in Context, Skepticism in Doubt.” Philosophical Perspectives 2: 139-157. Sosa, E. 1991. Knowledge in Perspective: Selected Essays in Epistemology. Cambridge University Press. Sosa, E. 2007. A Virtue Epistemology. Oxford University Press. Sosa, E. 2009. Reflective Knowledge. Oxford University Press. Sosa, E. 2015. Judgment and Agency. Oxford University Press.
268
|
Bibliography
Sripada, S. and J. Stanley. 2012. “Empirical Tests and Interest-Relative Invariantism.” Episteme 9(1): 3-26. Stanley, J. 2004. “On the Linguistic Basis for Contextualism.” Philosophical Studies 119 (1-2): 119-146. Stanley, J. 2005. Knowledge and Practical Interests. Oxford University Press. Sterelny, K. 2012. The Evolved Apprentice. MIT Press. Stich, S. 1992. “What Is a Theory of Mental Representation?” Mind 101 (402): 243-261. Stich, S. 2013. “Do Different Groups Have Different Epistemic Intuitions? A Reply to Jennifer Nagel.” Philosophy and Phenomenological Research 87 (1): 151-178. Stine, G. 1976. “Skepticism, Relevant Alternatives, and Deductive Closure.” Philosophical Studies 29 (4): 249-261. Stone, J. 1984. “Dreaming and Certainty.” Philosophical Studies 45: 353-368. Strevens, M. 2008. Depth: An Account ofScientific Explanation. Harvard University Press. Strevens, M. 2013. “No Understanding without Explanation.” Studies in History and Philosophy of Science 44 (3): 510-515. Stroud, B. 1984. The Significance of Philosophical Skepticism. Clarendon Press. Swain, S., J. Alexander, and J. Weinberg. 2008. “The Instability of Philosophical Intuitions: Running Hot and Cold on Truetemp.” Philosophy and Phenomenological Research 76 (1): 138-155. Thomasson, A. 2003. “Realism and Human Kinds.” Philosophy and Phenomenological Research 67 (3): 580-609. Tomasello, M. and M. Carpenter. 2007. “Shared Intentionality.” Developmental Science 10(1): 121-125. Travis, C. 1989. The Uses of Sense: Wittgenstein’s Philosophy of Language. Oxford University Press. Travis, C. 1991. “Annals of Analysis.” Mind 100 (398): 237-264. Trout, J. D. 2002. “Scientific Explanation and the Sense of Understanding.” Philosophy of Science 69 (2): 212-233. Tuomela, R. 2004. “Group Knowledge Analyzed.” Episteme 1 (2): 109-127. Turri, J. 2011. “Manifest Failure: The Gettier Problem Solved.” Philosophers’ Imprint 11 (8): 1-11. Turri, J. 2013. “A Conspicuous Art: Putting Gettier to the Test.” Philosophers’ Imprint 13 (10): 1-16. Unger, P. 1975. Ignorance: A Case for Skepticism. Oxford University Press. Unger, P. 1984. Philosophical Relativity. Oxford University Press. Vogel, J. 1990. “Are There Counterexamples to the Closure Principle?” In M. D. Roth and G. Ross (eds.), Doubting: Contemporary Perspectives on Skepticism. Dordrecht: Kluwer. Weinberg, J. 2006. “What’s Epistemology For? The Case for Neopragmatism in Normative Metaepistemology.” In S. Hetherington (ed.), Epistemology Futures. Clarendon Press. Weinberg, J. 2007. “How to Challenge Intuitions Empirically without Risking Skepticism.” Midwest Studies in Philosophy 31 (1): 318-343. Weinberg, J., S. Nichols, and S. Stich. 2001. “Normativity and Epistemic Intuitions.” Philosophical Topics 29 (1-2): 429-460. Wellman, H. M., A. K. Hickling, and C. A. Schult. 1997. “Young Children’s Explanations: Psychological, Physical, and Biological Reasoning.” In H. Wellman
Bibliography
|
269
and K. Inagaki (eds.), The Emergence of Core Domains of Thought: Physical, Psychological, and Biological Thinking. Wiley Blackwell. Wierzbicka, A. 1999. Emotions across Languages and Cultures. Cambridge University Press. Wilkenfeld, D. 2017. “Understanding Without Believing.” In S. Grimm, C. Baumberger, and S. Ammon (eds.), Explaining Understanding: New Essays in Epistemology and the Philosophy of Science. Routledge. Wilkenfeld, D., D. Plunkett, and T. Lombrozo. 2016. “Depth and Deference: When and Why We Attribute Understanding.” Philosophical Studies 173 (2): 373-393. Willaschek, M. 2007. “Contextualism about Knowledge and Justification By Default.” Grazer Philosophische Studien 74 (1): 251-277. Williams, B. 1973. Problems of the Self. Cambridge University Press. Williams, B. 1978. Descartes: The Project of Pure Inquiry. Routledge. Williams, B. 2002. Truth and Truthfulness. Princeton University Press. Williams, M. 1996. Unnatural Doubts: Epistemological Realism and the Basis for Skepticism. Princeton University Press. Williams, M. 2015. “What’s So Special about Human Knowledge?” Episteme 12 (2): 249-268. Williamson, T. 1994. Vagueness. Routledge. Williamson, T. 2000. Knowledge and Its Limits. Oxford University Press. Williamson, T. 2005. “Knowledge, Context, and the Agent’s Point of View.” In G. Preyer and G. Peter (eds.), Contextualism in Philosophy: Knowledge, Meaning, and Truth. Oxford University Press. Williamson, T. 2007. The Philosophy of Philosophy. Blackwell. Wittgenstein, U. 1964. Blue and Brown Books. Blackwell. Wittgenstein, L. 1969. On Certainty. Edited by G. E. M. Anscombe and G. H. von Wright and translated by Denis Paul and G. E. M. Anscombe. Harper Torchbooks. Woodward, J. 2003. Making Things Happen: A Theory of Causal Explanation. Oxford University Press. Wray, K. 2003. “What Really Divides Gilbert and the Rejectionists?” Protosociology 18. Wright, J. 2010. “On Intuitional Stability: The Clear, the Strong, and the Paradigmatic.” Cognition 115 (3): 491-503. Wright, J. 2013. “Tracking Instability in Our Philosophical Judgments: Is It intuitive?” Philosophical Psychology. 1-17. Wright, J. 2016. Intuitional Stability. In J. SytsmaandW. Buckwaiter (eds.), A Companion to Experimental Philosophy. Routledge. Wright, J. and J. Bengson 2009. “Asymmetries in Judgments of Responsibility and Intentional Action.” Mind and Language 24 (1): 24-50. Zagzebski, L. 2001. “Recovering Understanding.” In M. Steup (ed.), Knowledge, Truth, and Duty: Essays on Epistemic Justification, Responsibility, and Virtue. Oxford University Press.
270
|
Bibliography
INDEX
actionable information, 6-7, 12-13, 55, 67-69, 81, 87, 88-89, 99-100, 115-16, 123,160-62, 169-71, 209,247 Adams, Fred, 120 Ahlstrom, Kristoffer, 32 airport cases, 108, 169, 170-71, 185 Alexander, Joshua, 147, 154,155 ameliorative epistemology, 151-52 argument from ignorance, 191-93 Arico, Adam, 132-33 Aristotle, 40, 65-66, 225, 253 artifactual kind. See knowledge: social construction of assurance, 7, 103, 121-24,136 Austin, J. L., 8, 22, 56, 68-69, 79-80, 121, 171-72,173-76, 183-84,186, 190,191, 193-94, 200-2, 203^1, 206, 208-9, 219 Ayer, A. J„ 11,16,32-33
bank cases, 83-84, 85, 87-89,107-8, 109, 110, 167-68,170, 172,175, 182, 184, 185, 215, 232 Baz, Avner, 173, 179-80, 185n29 Beebe, James, 117-19, 120-21 BonJour, Laurence, 6, 57-59, 60-63, 64-66, 70, 72-73, 74, 77-78, 7980, 205 Boyd, Kenneth, 143,145n7
brain in a vat, 69-70,71-72,107,147,191-93, 194-96,200-1,204,210-11,212 Bridges, Jason, 179n24 Brown, Jessica, 82-83, 86-87, 90, 92-93, 95,99,100,111-12,115-16 94Buckwaiter, Wesley, 118,119,121 Butchvarov, Panyot, 20, 73
Carnap, Rudolf, 12n2 Cartesian conception of knowledge. See infallibilism Cartesian requirement, 198-202 Cassam, Quassim, 31 Cavell, Stanley, 206 Chalmers, David, 14-15 Chisholm, Roderick, 16, 67, 206 Chrisman, Matthew, 179-80, 181 closure principle, 195 cognitive ethology, 30-31, 33 Cohen, L.J., 130-31 Cohen, Stewart, 31, 78-79, 108, 192n2, 195 conceptual analysis, 4-5, 11-12, 15-20, 25, 26, 33, 37, 242 contextualism, 108n7, 110-11,116, 157-59,163, 164, 165-66, 167, 168-73,174-82, 183, 185, 187-88, 216-17, 233, 236-37 cooperation, 2,4, 5-7, 22, 23-24, 36, 40, 48, 49-52, 55, 117
271
Craig, Edward, 2, ll-12n5, 12-13, 14, 18, 19, 35n20, 37, 39, 42, 43-45, 49-51, 52, 53, 54, 55, 64, 67-68, 76, 80, 87, 104-5, 106, 108-9, 124, 129-30, 153,158-59, 166, 172nl6, 197-98, 209, 210, 212, 216, 251-52 creationist schoolteacher, 42,43-44, 130, 211-12nl5 curiosity, 35,93-94, 148^19,218-19, 229, 246^47
Dancy, Jonathan, 39,42-43,46,194 Deference, 243—44, 245 DePaul, Michael, 27-28 De Regt, Henk, 225-26, 230-31 Descartes, Rene, 189, 190-91, 192, 193-94,197,198, 199, 218, 219 Dogramaci, Sinan, 22 doubt, 56, 66-67,72, 122-23,137-38, 190,192,199,200-1,209, 210, 212, 216, 217,218,219,220 dreaming, 107,189,193-94,197,198-99, 200-1 Dretske, Fred, 142
Elgin, Catherine, 137n35, 232nl2,234 epistemic community, 41, 67, 68-69, 81, 91, 93-94, 98-100, 115-16, 123-24, 150, 162, 164, 166,167,168, 181, 202-3 dependence, 4, 36, 51,103, 215, 247 diversity, 3^1,7-8,139-56 injustice, 133-36 luck, 9, 14, 16-17n5, 58, 61, 73-74, 75-77, 111, 145-46, 147, 222, 223-24, 248-49, 250-52 naturalism, 4-5,11-12,15, 26, 27-33 pluralism, 7,23-24,54,103-36,151-52 relativism (see epistemic: diversity) side-effect effect, 118-21 standard, 40^-1,46 value, 2,12,13,18, 20-21nll, 21, 22, 27-28, 33, 57-58, 60,67, 71 epistemology descriptive, 4-5, 25,151-52
272
|
Index
knowledge-first, 4-5,15, 20-22 normative, 4—5, 25, 33, 151-52 error possibility, 79-80, 88, 89, 93, 100, 107, 108-26, 162-63, 191-96, 201-2,212,213,214 ethno-epistemology, See epistemic: di versity examiner situation, 37,74-75 experimental philosophy, 7-8,11-12, 117-20,140-41,146-47,151,152 experts, 26-27, 65, 242-45 explainer, 9, 225, 226-27,228, 229-30, 231-32, 233, 235-36, 237-38, 241, 242, 243-44, 246, 247, 249-50, 251, 252, 253, 254, 255 explanatory depth, 241, 242-43, 244, 245, 247, 249-50
factivity, 14, 21,107n6,182, 223-24, 232 fake bams, 77, 114, 125, 154nl6, 202, 251n27 fallibilism, 6,7-9, 56-58,59-63, 64-65, 67, 68-69, 70-71, 72-74, 77-80, 116-17, 196, 218, 219, 253. See also infallibilism Fantl, Jeremy, 82n2, 83,110-11 Feldman, Richard, 46-49, 50, 66-67 Fodor, Jerry, 17-18 Fogelin, Robert, 17n5 folk epistemology, 7-8, 25, 26, 136-37, 143,144, 145^47,151,152-54, 155-56, 242, 245 Fricker, Elizabeth, 21nl2, 52 Flicker, Miranda, 3n3, 35nl, 51,133-34 function-first epistemology definition, 2,12 methodology, 2, 3-5, 9, 12, 52-53,136, 186-87,211-12, 241
Galileo, 124, 239-40 Gelfert, Axel, 3n3, 52 Gendler, Tamar, 77nl8, 250 genealogy, 2, 5-6, 39-41, 50, 52-53,153. See also state of nature Gerken, Mikkel, 45, 106,113-15 Gettier Edmund, 16,61
problem, 2, 16-17, 33-34, 37, 61, 62-63, 73-77, 111, 113, 140-41, 142, 143-44, 146-41, 152-53, 251 Gilbert, Margaret, 128,131 Glock, Hans, 197 Goddard, Cliff, 149 Goldman, Alvin, 106-7, 202, 229 good informant. See reliable informant Greco, John, 157-58,159-60,161, 164, 227-28 Grice, Paul, 48,165-66, 203, 206 Grimm, Stephen, 68-69, 82, 91-92, 93, 94-95, 157-58, 162-64, 229, 230-31, 244-22, 250, 253-55 group belief nonsummative account of, 126,128-29 summative account of, 126-28, 129 Haslanger, Sally, 135,151-52 Hawthorne, John, 82,110-11,161-62, 250 Hazlett, Allan, 232n 12 Heller, Mark, 186-87 Hempel, Carl, 227, 228, 229 Henderson, David, 26, 54-55, 68-69, 91-92, 93,157-58, 160-61,164 Hetherington, Stephen, 57-58, 224n4 Hills, Alison, 223, 224n2, 224n3, 224n4, 224n5, 230-31, 234n6, 235-36, 238nl6,248n21, 250-51 Huebner, Bryce, 132-33 Hume, David, 150,189,190, 197, 205, 211-12nl5
Ichikawa, Jonathan 27nl 1,112 Impurism, 6-7, 81-83, 84-85, 86-93, 97, 9598,100-1 communal, 6-7, 82-83, 86-101 indicator property, 38-39, 129-30 infallibilism, 6,46-47,56,59,62-57,64,65, 72-74,78,206nll, 212,214-15,220 benefits of, 58,61, 63,73-80 informant, unrecognizable 42 invariantism insensitive, 78-79, 80, 89-90, 157-58, 161-62, 163-64, 165-73, 174-75, 177-78, 181-82, 183, 187-88, 216-17, 236-37
subject-sensitive, 158, 161-64 Jackson, Frank, 1 lnl justification conclusive, 59, 61, 62-63, 64, 67, 73-74, 77-79 qualitative model of, 66-67
Kant, Immanuel, 51,134 Kaplan, Mark, 70, 75-76,191, 2069,219 Kappel, Klemens, 23, 42-43, 108-9, 117 Kelp, Christoph, 42-43, 80,108-9, 114, 157-58 Khalifa, Kareem, 225-26, 230-32, 237-38, 239 Kitcher, Philip, 187 Knobe, Joshua, 118, 119nl7,120 ‘know’ cross-cultural, 1, 139-56 frequency of, 1,139-56 semantics of, 8, 33-34,157-88 knowledge animal, 30-31, 33,104, 137, 253 blame, 7, 117-21, 124, 136 communal threshold of, 6-7, 67-70, 88-89,91,100,115-16 end of inquiry, 7, 71-72, 103,107-9, 116, 122-23, 124, 136,186-87, 246-47, 248 group, 33-34, 104, 125-33, 136, 137 (see also group belief) indeterminacy of, 8-9, 86, 90-91, 92, 97, 96158, 174, 175-76, 177-78, 182, 185-86, 187, 216 myth of, 9-10, 57 natural kind, 28-33 norms (see norm of assertion; norm of practical reasoning) primitive, 20-21 social construction of, 4-5, 31-32, 33, 125 as universal, 7-8, 139-40, 142-56 without belief, 42, 43-44, 129-31, 242 (see also creationist school teacher)
Index |
273
Koriat, Asher, 155-56 Kornblith, Hilary, 27-29, 30-32, 33 Kukla, Rebecca, 7-8 Kusch, Martin, 52, 80 Kvanvig, Jonathan, 107, 109, 234, 249-50n23, 250-51, 252
Lackey, Jennifer, 42, 125-26, 127, 128-31,133 Lawlor, Krista, 69n9,121-24, 208 Lehrer, Keith, 141 Lewis, David, 173,196,213, 220 Lipton, Peter, 238—41 Lombrozo, Tania, 229,242-45 Longino, Helen, 135 lottery paradox, 61-63, 73, 77-79, 109-10, 217nl7 Lycan, William, 17-18,74
Machery, Edouard, 143-44 McGrath, Matthew, 82n2, 83,110-11 McKenna, Robin, 105n2,157-58 Mercier, Hugo, 147 Millar, Alan, 16-17 Moran, Richard, 4 Moser, Paul, 197 Nagel, Jennifer, 30,143 Neta, Ram, 150 Nichols, Shaun, 141,146,151,195n3 norm of assertion, 7, 22, 33-34,10916, 248 norm of practical reasoning, 7, 22, 33-34, 109-16, 248 objectivization, 5-6, 39-47, 53, 54, 67, 76,104—5, 209, 210-11,212-13, 216,236 ordinary language philosophy, 11-12, 23-24, 25, 109-10, 171, 173, 248 parroter, 237 Pettit, Philip, 126-27,128-29 Plato, 16, 57, 65,140n2 Plunkett, David, 242-45 practical explication. See Carnap, Rudolf
274
|
Index
practical inquiry, 217-20 practical interests, 6-9, 31-32, 35, 50, 89-90, 91, 92,94,98,99,101,124, 159,160-62, 163,164-65 practical reasoning, 2,6-7,33-34,81-82, 83-85,86-87, 88-89,90-91,92,93, 98,99-102,124,159,161-59,162-63, 164,165,181,186-87,205,248 pragmatism, 8, 21-22, 82,125, 158, 172-58,177-78,179-88 Price, Huw, 177-78,180 . Pritchard, Duncan, 58n3, 224nl, 224n2, 224n5, 230-31, 242-43, 246, 250 probability, 6,46, 57, 58, 59-60, 61-62, 64, 77-78, 79-80, 82 problem of easy knowledge, 85-86,90-91,92 problem of practically irrelevant propositions, 86,90-92,93, 94,95, 96-97,100 pure inquiry, 217-20 purism. See impurism Railton, Peter, 235-36 Recanati, Francois, 185n29 Reed, Baron, 78,111 relevance approach, 83, 85-86 relevant alternatives, 67-70,93, 98-99, 100,101,123-24,150,196, 198, 199-203, 210-15 reliable informant, criteria for, 3739, 42^-5 reverse engineering, 4-5, 11-12, 15, 22-27, 242 Robinson Crusoe problem, 44, 230, 235 Rorty, Richard, 157, 180 Russell, Bertrand, 12,16n8, 70 Rysiew, Patrick, 45, 70, 106,108-9, 168n8
Schmitt, Frederick, 21n8, 29-30,46,106 Shapin, Stephen, 38-39 skepticism, 2, 5-6, 7-9, 33-34, 46-49, 59, 62, 189-221. See also infallibilism and the point of knowledge, 209-21 types of, 191-95,196-97
Sliwa, Paulina, 248n23, 250 Socrates, 11-12,16 Sosa, Ernest, 36-37, 214-15 stakes, 6-7,47-36,68-56, 81-82,83-85, 87-89,90-91,93,94,95,96-97,98, 100-1,107-8,113,114nl5,115-16, 162nll7,162-66,167-68,169,170, 172,214 Stanley, Jason, 161-62 state of nature, 2, 39, 52-53 Sterelny, Kim, 4 Stich, Stephen, 17-18, 141,143,145-46 Stone, Jim, 193-94 Strevens, Michael, 229 Stroud, Barry, 158-59, 190,191,197, 200-1, 203-7 survival, 3-4,12, 30-31, 35,71,117, 118-19,148nl 1,228, 255 Swain, Stacey, 147,154,155
testimony, 2, 3-4,22, 36,44, 51, 68-69, 78-79, 96, 104-5, 120,134, 218, 223-24, 229-30, 248-49 thought experiments, 11-12, 28, 146-47, 239-40 threshold problem, 6, 57-58,59, 60-61, 63, 65-66, 69, 70, 71, 80, 82-101, 115,116 trivial truths, 91-92, 94-95, 97, 98 Truetemp, 141, 146^-7, 152-53,154,155 trust, 51, 96, 99
understanding, 1, 2, 9-10,12,14, 25 end of inquiry, 246—47, 248 infant, 234, 253-29 objectual, 231, 234, 241 practical, 231, 241 reductionism, 248-49 role of, 2, 9, 246-^8 transparency, 252-29 value of, 2,9,14-15,226-29,242, 247—48, 255 vs knowledge, 2, 9, 242-50 Unger, Peter, 177-78, 185-86, 197 unity approach, 83-85 vagueness, 63-64, 67-68, 70
Weinberg, Jonathan, 31, 52, 141, 145-46, 147,154, 155 Wierzbicka, Anna, 149 Wilkenfeld, Daniel, 225n7, 242-4-5 Willaschek, Markus, 194 Williams, Bernard, 26-27, 36, 37, 52, 74, 189, 190-91,217-18,219 Williamson, Timothy, 16-18, 19, 20-22, 27-28, 30-31, 164-65 wisdom, 1,9-10,180n26 Wittgenstein, Ludwig, 18, 180, 219 Wright, Jennifer, 153-54,155
Zagzebski, Linda, 224n2, 224n3, 224n5, 249n22
Index
|
275