328 12 2MB
English Pages 256 [249] Year 2021
Neue Perspektiven der marktorientierten Unternehmensführung Ruth Stock-Homburg · Jan Wieseke Hrsg.
Moritz Merkle
Humanoid Service Robots Customer Expectations and Customer Responses
Neue Perspektiven der marktorientierten ¨ hrung Unternehmensfu Reihe herausgegeben von Ruth Stock-Homburg, Marketing und Personalmanagement, Technische Universität Darmstadt, Darmstadt, Deutschland Jan Wieseke, Lehrstuhl für Marketing, Ruhr-Universität Bochum, Bochum, Deutschland
Der Reihe „Neue Perspektiven der marktorientierten Unternehmensführung“, die sich Konzepten des erfolgreichen Umgangs mit aktuellen und zukünftigen Entwicklungen in der Unternehmenspraxis widmet, liegt eine interdisziplinäre Perspektive zugrunde. Der Interdisziplinarität wird dadurch Rechnung getragen, dass verschiedene Disziplinen innerhalb der Betriebswirtschaftslehre beleuchtet werden (insbesondere Marketing, Innovationsmanagement und Personalmanagement). Darüber hinaus erfährt die Schnittstelle zwischen verschiedenen Facetten der Betriebswirtschaftslehre und der Psychologie (insbesondere Arbeits- und Organisationspsychologie) besondere Bedeutung. Die in der Reihe „Neue Perspektiven der marktorientierten Unternehmensführung“ erscheinenden Arbeiten orientieren sich inhaltlich und konzeptionell an internationalen wissenschaftlichen Standards. Ausgehend von einer stringenten theoretischen Fundierung erfolgt die qualitative bzw. quantitative empirische Untersuchung des jeweiligen Forschungsgegenstands.
Weitere Bände in der Reihe http://www.springer.com/series/12610
Moritz Merkle
Humanoid Service Robots Customer Expectations and Customer Responses
Moritz Merkle Department of Law and Economics Technical University of Darmstadt Darmstadt, Germany Ph.D. Thesis at Technical University of Darmstadt in 2020
ISSN 2626-1499 ISSN 2626-1529 (electronic) Neue Perspektiven der marktorientierten Unternehmensführung ISBN 978-3-658-34439-9 ISBN 978-3-658-34440-5 (eBook) https://doi.org/10.1007/978-3-658-34440-5 © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Fachmedien Wiesbaden GmbH, part of Springer Nature 2021 This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, expressed or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Responsible Editor: Marija Kojic This Springer Gabler imprint is published by the registered company Springer Fachmedien Wiesbaden GmbH part of Springer Nature. The registered company address is: Abraham-Lincoln-Str. 46, 65189 Wiesbaden, Germany
Acknowledgments
My time as Ph.D. student at the Technical University of Darmstadt in close cooperation with the leap-in-time research center was a truly extraordinary experience. During this time, I had the opportunity to work on a variety of exciting research topics and corporate projects, to participate in the development of the department and to experience exciting technologies. When I started my Ph.D., the department was mainly focused on strategic marketing, innovation research and HR topics. Now that I am completing my Ph.D. as the first research assistant in HRI, robotic research is the strongest research focus of the department. I am proud of this development and grateful for the opportunity to participate. Therefore, my first and greatest thanks goes to my supervisor Prof. Dr. Dr. Ruth Stock-Homburg for providing the opportunity of such an exciting and eclectic Ph.D. period, for her constant support, thrilling ideas, and for opening up opportunities for personal and professional development. Furthermore, I would like to thank Prof. Dr. Peter Buxmann, as expert in artificial intelligence and digital transformation, who agreed to be my second supervisor for his time and his interest in my work. The experimental studies with face-to-face HRIs were very complex as well as expensive. It would not have been possible to conduct these experiments without the help of my colleague and robotics expert Johannes Völker, as well as the support of the Förderverein Innov@tion4 Future e. V., Dietmar Eidens of Merck KGaA, Oliver Klink of Taunus Sparkasse, the German Research Foundation, the Dr. Hans Riegel Foundation and the leap-in-time research center. I would also like to thank my colleagues for their support during my time at the department. Together we created a highly productive, inspiring and pleasant atmosphere. Special thanks to Carmen Lukoschek, Christian Holthaus, Inés Scherer and Katharina Schneider, who have accompanied me throughout my entire
v
vi
Acknowledgments
Ph.D. time and to my office colleague Mai Anh Nguyen, for supporting me during my time at the department. I would also like to thank the student assistants who supported me during the experimental studies: Adriana, Eileen, Fiona, Marius, Ramona, and Tobias. Further, I want to express my sincere personal gratitude to my family and my friends who have always supported me. Thank you for believing in me and your boundless support. Darmstadt October 2020
Moritz Merkle
Summary
Nowadays, service firms are increasingly relying on service robots at the customer service encounter in many fields, such as retailing, healthcare, tourism and hospitality. Service robots are on the rise and current developments in artificial intelligence are further supporting this trend. This thesis addresses customer expectations and customer responses toward humanoid service robots that have gained remarkable attention in research and practice. On the one hand, it contributes to the latest scientific research by conducting experimental face-to-face human-robot interaction (HRI) studies in service settings, transferring knowledge from service marketing to HRI and demonstrating benefits of integrating different theoretical perspectives. On the other hand, it is of managerial relevance providing knowledge about customer reactions and identifying preferable robotic behaviors at the service encounter. Thereby, this thesis takes a customer-centric approach by focusing on customer attitudes and perceptions of robotized services. Thus, the first major goal relates to the analysis of customer attributions and expectations towards service robots. To achieve that goal, two studies were designed based on established theories and concepts. The first study theoretically develops and tests a service robot acceptance model (SRAM), relying on expectation disconfirmation paradigm, role theory and a qualitative study with n = 63 customers. The model assumes that customers intuitively rely on alternative references to assess the interaction with service robots. This in turn affects customers’ service robot acceptance. The experimental laboratory study with n = 90 participants and the humanoid robot of the type Nao (study 1) shows that the acceptance of service robots depends on the reference category the customer relies on prior to the interaction with a service robot. Furthermore, robot anxiety affects the relevance of different reference categories for service robot acceptance.
vii
viii
Summary
Study 2 was designed with an international focus to identify customers’ expectations toward service robots and to identify cross-cultural differences in robot attributions based on Hofstede’s cultural dimensions. It was carried out as an international survey study across three different continents (n = 610) and successfully linked robot attributions of empathy, expertise, reliability and trust to the four cultural dimensions of individualism, masculinity, power distance and uncertainty avoidance. Further, the results unveil limitations in the comparability of HRI results across different cultures. The second major goal is the analysis of customer responses to humanoid robots at the service encounter. To reach this goal, two experimental studies with face-to-face HRI were conducted in a natural experimental service setting. Study 3 relies on the computers-are-social-actors paradigm (CASA) and the uncanny valley paradigm to examine differences between human-human interactions and human-robot interactions. The effects of discrepancy between customer expectations and perceptions are conceptualized based on expectation disconfirmation paradigm. Effects on customer delight are analyzed through polynomial modeling with surface response method analysis. It includes two experiments in a hotel setting with either a service robot or a human frontline employee as a service representative. The first experiment (n = 132) compares a human-human interaction with a human-robot interaction in a failure-free service encounter, whereas the second experiment (n = 137) simulates a service failure. The service robot’s innovative service behavior increases customer delight, though confirming their expectations is more important than the perception of innovative service behavior itself. Last, study 4 is designed to examine the effects of service failures on customer satisfaction comparing the service robot with a frontline employee in an experimental study (n = 120) based on script theory and cognitive dissonance theory. The results show that service robots can keep up with human frontline employees in terms of customer satisfaction and after a service failure customer satisfaction declines far less for a service robot compared with a human frontline employee. Overall, this thesis contributes to scientific research and managerial practice in several respects. First, it transfers knowledge from human-human interactions to human-robot interactions for the frontline encounter, showing that although robotic behaviors are artificial, they still lead to positive customer responses, introducing innovative service behavior to human-robot interactions and comparing human-robot interactions with human-human interactions in terms of service failures. Second, this thesis demonstrates benefits of integrating various theoretical perspectives, such as script theory, dissonance theory, expectation disconfirmation paradigm, CASA paradigm, uncanny valley paradigm, and attribution theory.
Summary
ix
Further, it applies polynomial modeling and surface response analysis to the expectation disconfirmation paradigm for human-robot interactions, and provides specific knowledge about customer reactions and desirable robotic behaviors at the service encounter.
Zusammenfassung
Heute verlassen sich Dienstleistungsunternehmen in vielen Bereichen wie dem Einzelhandel, dem Gesundheitswesen, dem Tourismus und dem Gastgewerbe zunehmend auf Serviceroboter im Kundenkontakt. Serviceroboter sind auf dem Vormarsch und die aktuellen Entwicklungen im Bereich der künstlichen Intelligenz unterstützen diesen Trend zusätzlich. Diese Dissertation befasst sich mit den Erwartungen und den Reaktionen der Kunden auf humanoide Serviceroboter, die in Forschung und Praxis bemerkenswerte Aufmerksamkeit erlangt haben. Einerseits leistet sie einen Beitrag zur neuesten wissenschaftlichen Forschung, indem sie experimentelle Studien zur Mensch-Roboter-Interaktion von Angesicht zu Angesicht in Dienstleistungsumgebungen durchführt, Wissen aus dem Dienstleistungsmarketing auf Mensch-Roboter-Interaktionen überträgt und die Vorteile der Integration verschiedener theoretischer Perspektiven aufzeigt. Andererseits ist sie von betriebswirtschaftlicher Relevanz, indem sie Wissen über Kundenreaktionen liefert und bevorzugte Roboterverhaltensweisen bei der Dienstleistungsbegegnung identifiziert. Dabei verfolgt diese Dissertation einen kundenzentrierten Ansatz, indem sie sich auf die Attributionen und Wahrnehmungen der Kunden zu robotergestützten Dienstleistungen konzentriert. Daher bezieht sich das erste Hauptziel auf die Analyse von Kundenattributionen und Kundenerwartungen gegenüber Servicerobotern. Um dieses Ziel zu erreichen, wurden zwei Studien entworfen, die auf etablierten Theorien und Konzepten basieren. Die erste Studie entwickelt und testet theoretisch ein Serviceroboter-Akzeptanzmodell (SRAM) und stützt sich dabei auf das Erwartungs-Diskonfirmations-Paradigma, die Rollentheorie und eine qualitative Studie mit n = 63 Kunden. Das Modell geht davon aus, dass Kunden intuitiv auf alternative Referenzen zurückgreifen, um die Interaktion mit Servicerobotern zu bewerten. Dies wiederum beeinflusst die Akzeptanz der Serviceroboter durch die
xi
xii
Zusammenfassung
Kunden. Die experimentelle Laborstudie mit n = 90 Teilnehmern und dem humanoiden Roboter vom Typ Nao zeigt, dass die Akzeptanz von Servicerobotern von der Referenzkategorie abhängt, auf die sich der Kunde vor der Interaktion mit einem Serviceroboter verlässt. Darüber hinaus beeinflusst die Angst vor Roboten die Relevanz der verschiedenen Referenzkategorien für die Akzeptanz von Servicerobotern. Studie 2 wurde mit internationaler Ausrichtung konzipiert, um die Erwartungen der Kunden an Serviceroboter zu ermitteln und kulturübergreifende Unterschiede in den Roboterattributionen auf Grundlage der kulturellen Dimensionen nach Hofstede zu identifizieren. Sie wurde als internationale Befragungsstudie über drei verschiedene Kontinente hinweg (n = 610) durchgeführt und verknüpft erfolgreich Roboterattributionen von Empathie, Expertise, Zuverlässigkeit und Vertrauen mit den vier kulturellen Dimensionen Individualismus, Maskulinität, Machtdistanz und Unsicherheitsvermeidung. Darüber hinaus legen die Ergebnisse dar, dass es Einschränkungen bei der Vergleichbarkeit von Ergebnissen bei Mensch-Roboter-Interaktionen über verschiedene Kulturen hinweg gibt. Das zweite Hauptziel ist die Analyse der Kundenreaktionen auf humanoide Roboter bei der Servicebegegnung. Um dieses Ziel zu erreichen, wurden zwei experimentelle Studien mit Mensch-Roboter-Interaktionen von Angesicht zu Angesicht in einer natürlichen experimentellen Serviceumgebung durchgeführt. Studie 3 stützt sich auf das „Computer-sind-soziale-Akteure“ Paradigma (CASA) und auf das “Uncanny Valley” Paradigma, um die Unterschiede zwischen Mensch-Mensch-Interaktionen und Mensch-Roboter-Interaktionen zu untersuchen. Die Auswirkungen der Diskrepanz zwischen Kundenerwartungen und Kundenwahrnehmungen werden auf der Grundlage des ErwartungsDiskonfirmations-Paradigmas konzeptualisiert. Die Auswirkungen auf die Kundenbegeisterung werden durch Polynom-Modellierung mit Analyse der ResponseSurface-Methodologie analysiert. Dazu gehören zwei Experimente in einer Hotelsituation mit entweder einem Serviceroboter oder einem menschlichen Mitarbeiter als Servicevertreter. Das erste Experiment (n = 132) vergleicht eine MenschMensch-Interaktion mit einer Mensch-Roboter-Interaktion in einer störungsfreien Servicebegegnung, während das zweite Experiment (n = 137) einen Servicefehler simuliert. Das innovative Serviceverhalten des Serviceroboters erhöht die Begeisterung der Kunden, obwohl die Bestätigung ihrer Erwartungen wichtiger ist als die Wahrnehmung innovativen Serviceverhaltens. Zuletzt werden in Studie 4 die Auswirkungen von Servicefehlern auf die Kundenzufriedenheit untersucht, indem der Serviceroboter in einer experimentellen Studie (n = 120), die auf der Skript-Theorie und der kognitiven DissonanzTheorie basiert, mit einem Menschlichen Kundenkontaktmitarbeiter verglichen
Zusammenfassung
xiii
wird. Die Ergebnisse zeigen, dass Serviceroboter in Bezug auf die Kundenzufriedenheit mit menschlichen Kundenkontaktmitarbeitern mithalten können und dass die Kundenzufriedenheit nach einem Servicefehler bei einem Serviceroboter im Vergleich zu einem menschlichen Kundenkontaktmitarbeiter weitaus weniger abnimmt. Insgesamt trägt diese Dissertation in mehrfacher Hinsicht zur wissenschaftlichen Forschung und zur Unternehmenspraxis bei. Erstens überträgt sie das Wissen aus dem Kundenkontakt von Mensch-Mensch-Interaktionen auf MenschRoboter-Interaktionen. Sie zeigt, dass robotisches Verhalten zwar künstlich ist, aber dennoch positive Kundenreaktionen hervorruft, führt innovatives Serviceverhalten in Mensch-Roboter-Interaktionen ein und vergleicht Mensch-RoboterInteraktionen mit Mensch-Mensch-Interaktionen im Hinblick auf Servicefehler. Zweitens zeigt diese Arbeit die Vorteile der Integration verschiedener theoretischer Perspektiven auf, wie z. B. der Skripttheorie, der Dissonanztheorie, dem Erwartungs-Diskonfirmations-Paradigma, dem CASA-Paradigma, dem Uncanny Valley Paradigma und der Attributionstheorie. Darüber hinaus wendet sie Polynom-Modellierung und Response-Surface-Methodologie auf das ErwartungsDiskonfirmations-Paradigma für Mensch-Roboter-Interaktionen an und liefert spezifisches Wissen über Kundenreaktionen und erwünschtes Roboterverhalten bei der Servicebegegnung.
Contents
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.1 Managerial Relevance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2 Scientific Relevance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.3 Goals and Research Questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.4 Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1 1 5 8 10
2 Conceptual Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1 Basic Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2 Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3 Theoretical Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
15 16 28 56
3 Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1 Construct Measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2 Data Analysis Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
77 77 99
4 Study 1: A Service Robot Acceptance Model: Customer Acceptance of Humanoid Robots During Service Encounters . . . . . . 4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2 Theoretical Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3 Qualitative Study on Social Robot Acceptance (Study I) . . . . . . . 4.4 Testing the Social Robot Acceptance Model (Study II) . . . . . . . . 4.5 Experimental Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.6 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.7 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
113 113 116 117 119 121 125 128
5 Study 2: A Cross-Country Comparison of Attitudes toward Humanoid Service Robots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
131 131
xv
xvi
Contents
5.2 5.3 5.4 5.5 5.6
Theoretical Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Hypotheses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Empirical Basis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6 Study 3: Beyond the Call of Duty: The Impact of Innovative Service Behavior by Robots on Customer Delight . . . . . . . . . . . . . . . . 6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2 Theoretical Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.3 Framework and Hypotheses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.4 Empirical Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.5 Innovative Service Behavior in a Failure-Free Service Encounter (Study I) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.6 Innovative Service Behavior after a Service Failure (Study II) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.7 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 Study 4: Customer Responses to Service Robots: Comparing Human-Robot Interaction with Human-Human Interaction . . . . . . . 7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.2 Literature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.3 Conceptual Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.4 Hypotheses Development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.5 Data Collection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.6 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.7 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
134 136 138 139 142 145 145 148 150 155 162 167 167 173 173 175 176 179 181 183 186
8 Overall Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8.1 Overall Scientific Contribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8.2 Overall Managerial Contribution . . . . . . . . . . . . . . . . . . . . . . . . . . . 8.3 Overall Limitations and Recommended Areas for Future Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
191 192 196
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
201
198
Abbreviations
AI ANCOVA ANOVA AVE BMS CAGR CASA CR DRA df EDT EMS FLE FSR HHI HRI HSD ICC IFR IR IS ISB LSD M MANCOVA MANOVA
Artificial Intelligence Analysis of Covariances Analysis of Variance Average Variance Extracted Between-Targets Mean Compound Annual Growth Rate Computers-Are-Social-Actors Paradigm Composite Reliability Degree of Robotics Adoption Degrees of Freedom Expectation Disconfirmation Theory Expected Mean Squares Frontline Employee Frontline Service Robot Human-Human Interaction Human-Robot Interaction Honestly Significant Difference Intraclass Correlation Coefficient International Federation of Robotics Indicator Reliability Information Systems Innovative Service Behavior Least Significant Difference Mean Multivariate Analysis of Covariance Multivariate Analysis of Variance
xvii
xviii
MASA N n.s. p RASA SD SE SRAM SRM SS UTAUT WASA α
Abbreviations
Media are Social Actors Paradigm Sample Size Non-significant Significance Level Robots are Social Actors Paradigm Standard Deviation Standard Error Service Robot Acceptance Model Surface Response Method Sum of Squares Unified Theory of Acceptance and Use of Technology Websites are Social Actors Paradigm Cronbach’s Alpha
List of Figures
Figure 1.1 Figure Figure Figure Figure Figure Figure
1.2 1.3 1.4 1.5 1.6 2.1
Figure 2.2 Figure Figure Figure Figure Figure Figure Figure Figure Figure
2.3 2.4 2.5 2.6 2.7 3.1 3.2 3.3 4.1
Figure 4.2 Figure 4.3 Figure 5.1 Figure 6.1
Exemplary Applications of Professional Service Robots in the Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Professional Service Robots—Global Revenue . . . . . . . . . . . Number of Research Publications on Service Robots . . . . . . Major Goals of the Thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . Overarching Research Framework of the Thesis . . . . . . . . . . Organization of the Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . Categorization of Robot Types According to Application Purpose . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Categorization of Robot Types According to Appearance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Overview of Applied Theories . . . . . . . . . . . . . . . . . . . . . . . . . Framework of the Technology Acceptance Model . . . . . . . . . Framework of the Expectation Disconfirmation Theory . . . . Representation of the Uncanny Valley Paradigm . . . . . . . . . . Framework of the Cognitive Dissonance Theory . . . . . . . . . . Sequential Process of the Analysis of Variance . . . . . . . . . . . Sequential Process of the Regression Analysis . . . . . . . . . . . . Sequential Process of the Surface Response Method . . . . . . . Service Robot Acceptance Model (SRAM) During Service Encounter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Experimental Setting of the Study . . . . . . . . . . . . . . . . . . . . . . Check-in Experiences Prior to Human-Robot Interaction . . . Hofstede’s Cultural Dimensions for India, Germany, and the US . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Framework of the Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2 4 6 9 10 12 20 22 57 58 60 63 72 100 104 108 120 122 123 135 151
xix
xx
List of Figures
Figure 6.2 Figure 6.3 Figure 6.4 Figure Figure Figure Figure
7.1 7.2 7.3 7.4
Service Representative and Experimental Setting, Studies I and II . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sample Gestures by Pepper to Express Innovative Service Behaviors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Effects of Innovative Service Behavior on Customer Delight . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Uncanny Valley Paradigm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Study Framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Pepper Robot as Mechanical Basis . . . . . . . . . . . . . . . . . . . . . Experimental Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
158 159 166 179 180 182 183
List of Tables
Table Table Table Table Table Table
1.1 2.1 2.2 2.3 2.4 2.5
Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table
2.6 2.7 2.8 2.9 2.10 2.11 2.12 2.13 3.1 3.2 3.3 3.4 3.5 3.6 3.7
Table 3.8 Table 3.9
Study Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Exemplary Definitions of Service Robots . . . . . . . . . . . . . . . . Overview about the Reviewed Literature . . . . . . . . . . . . . . . . Overview on Conceptual Studies in Service Marketing . . . . Overview on Empirical Studies in Service Marketing . . . . . . Overview on Experimental Studies in Human-Robot Interaction Literature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Summary of Technology Acceptance Model . . . . . . . . . . . . . Summary of Expectation Disconfirmation Theory . . . . . . . . . Summary of Uncanny Valley Paradigm . . . . . . . . . . . . . . . . . Summary of Computers-Are-Social-Actors Paradigm . . . . . . Summary of Script Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . Summary of Role Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Summary of Cognitive Dissonance Theory . . . . . . . . . . . . . . Summary of Cultural Dimensions Theory . . . . . . . . . . . . . . . Selected Criteria of Reliability and Validity . . . . . . . . . . . . . . Selected Quality Measures with Threshold Values . . . . . . . . Items and Quality Measures for “Attributed Empathy” . . . . . Items and Quality Measures for “Attributed Reliability” . . . Items and Quality Measures for “Attributed Expertise” . . . . Items and Quality Measures for “Attributed Trust” . . . . . . . . Items and Quality Measures for “Expected Innovative Service Behavior” . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Items and Quality Measures for “Robot Anxiety” . . . . . . . . . Items and Quality Measures for “Technological Affinity” . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
11 18 31 33 44 51 59 62 65 67 69 71 74 76 82 87 89 90 91 92 93 94 95
xxi
xxii
List of Tables
Table 3.10 Table 3.11 Table 3.12 Table 3.13 Table 3.14 Table 3.15 Table 4.1 Table 4.2 Table 4.3 Table 5.1 Table 5.2 Table 5.3 Table Table Table Table
5.4 5.5 6.1 6.2
Table 6.3 Table 6.4 Table 6.5 Table 6.6 Table 6.7 Table 7.1 Table 7.2 Table 7.3
Items and Quality Measures for “Ease of Use” . . . . . . . . . . . Items and Quality Measures for “Usefulness” . . . . . . . . . . . . Items and Quality Measures for “Perceived Innovative Service Behavior” . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Items and Quality Measures for “Customer Delight” . . . . . . Items and Quality Measures for “Customer Satisfaction” . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Interpretation of Surface Test Values . . . . . . . . . . . . . . . . . . . . Manipulation Check: t-Test for Mean Differences . . . . . . . . . Service Robot Acceptance: t-Test for Mean Differences in User Perception . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Moderating Effects: Linear Regression with Robot Anxiety as a Moderator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Literature Review about Robots in Different Countries . . . . Mean Differences Among the Robot Attributes in the US . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Mean Differences Among the Robot Attributes in Germany . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Mean Differences Among the Robot Attributes in India . . . . Comparison Between the Countries . . . . . . . . . . . . . . . . . . . . . Four Experimental Conditions . . . . . . . . . . . . . . . . . . . . . . . . . Descriptive Statistics, Reliabilities, and Intercorrelations for HRI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Bonferroni Post Hoc Test on Delight, Depending on Service Behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Bonferroni Post Hoc Test on Delight, FLE versus FSR . . . . Polynomial Regression and Response Surface Analysis . . . . Descriptive Statistics, Reliabilities, and Intercorrelations (Service Failure) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Bonferroni Post Hoc Test on Delight, FLE versus FSR (Service Failure) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Literature Review: Customer Responses to Self-Service Technology Failures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Differences in Customer Satisfaction . . . . . . . . . . . . . . . . . . . Scheffé’s Post Hoc Test for Mean Differences in Customer Satisfaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
95 96 97 98 98 111 125 126 127 133 139 140 140 141 160 163 164 164 165 168 168 177 184 185
1
Introduction
1.1
Managerial Relevance
Since more than half a century ago, we are relying on robots. It started with the robot Unimate that General Motors applied at their automotive production line in 1961 (Hockstein et al. 2007). Soon, many other industrial robots of the first generation followed in the automotive industry. Industrial robots developed fast and were soon applied for missions humans could not fulfill, such as at the bottom of the sea, out in space, in disaster areas and mine seeking in war zones. Later on, a second type of robot emerged. Unlike an industrial robot that is “an automatically controlled, reprogrammable, multipurpose manipulator, programmable in three or more axes which may be either fixed in place or mobile for use in industrial automation applications” (ISO 8373), service robots “are systembased autonomous and adaptable interfaces that interact, communicate and deliver service to an organization’s customers” (Wirtz et al. 2018, p. 909). Many service industries such as retail (Kanda et al. 2010; Sabelli and Kanda 2016; Shiomi et al. 2013), tourism and hospitality (Gockley et al. 2005; Ivanov et al. 2017; Pan et al. 2015; Pinillos et al. 2016), healthcare (Lee et al. 2017; Mirheydar and Parsons 2013; Piezzo and Suzuki 2017) and even education (Barakova et al. 2015; Conti et al. 2017; Kanda et al. 2007) are increasingly relying on these professional service robots. Figure 1.1 shows exemplary applications of professional service robots in the field. Currently these robots focus on giving information on products, such as Softbank’s robot Pepper that assists frontline employees to help customers with the selection of the best suitable coffee maker in Japan (Sanborn 2015). In its own stores in Japan Softbank relies on Pepper robots to advise customers about
© The Author(s), under exclusive license to Springer Fachmedien Wiesbaden GmbH, part of Springer Nature 2021 M. Merkle, Humanoid Service Robots, Neue Perspektiven der marktorientierten Unternehmensführung, https://doi.org/10.1007/978-3-658-34440-5_1
1
2
1
Introduction
available mobile devices and cell phone contracts (Alpeyev and Amano 2015). Beyond retailing, many service robots are already applied in tourism and hospitality, such as Pepper robots entertaining and informing passengers on AIDA cruises (Linnhoff-Popien and Hofbauer 2018) and checking in customers at the Henn-na Hotel in Japan (Northfield 2015). Even in healthcare, first projects started relying on robots such as therapy robots for children with autism (Richardson et al. 2018) and exploring robots for elderly care (Sgorbissa et al. 2018).
Retail
Tourism / Hospitality
Healthcare
Pepper robot as sales promotor at a Nescafé shop in Japan1.
Service robots welcoming guests at the Henn-na hotel in Japan2.
Nao robots helping children with autism in the DREAM project, Portsmouth3.
Pepper robots advising customers at a Softbank shop in Tokyo4.
Pepper robot working on AIDA cruise ships to guide tourists5.
Pepper caring for elderly in cooperation with Advinia Health Care in UK6.
Notes: 1) https://japan.cnet.com/article/35057230/ 2) https://www.nytimes.com/2015/08/13/travel/can-a-hotel-robot-read-a-do-not-disturb-sign.html 3) https://phys.org/news/2017-06-robots-children-autism.html 4) https://www.dailymail.co.uk/sciencetech/article-3507541/The-phone-store-run-ROBOTS-Tokyo-firm-replaces-staff-10-versionsPepper-emotional-humanoid.html 5) https://www.com-magazin.de/news/forschung/service-roboter-heuert-aida-cruises-an-1065413.html 6) https://www.technocracy.news/elderly-to-be-cared-for-by-robots-to-solve-staff-shortage/
Figure 1.1 Exemplary Applications of Professional Service Robots in the Field
However, most of these service robots were just applied for simple support and routine tasks (Jacobs and Graf 2012) and might primarily serve as a marketing activity to attract customers (Pandey and Gelin 2018) at first. In the meanwhile, researchers developed much more sophisticated AI algorithms going far beyond the simple realization of simple standard tasks, such as creating automated news reports (Brooks 2014) or outperforming human Go (Lee et al. 2016) and Jeopardy
1.1 Managerial Relevance
3
players (Ferrucci et al. 2013). Leading media such as The New York Times, Forbes and Associated Press are already publishing automated news. AI is useful to create news articles based on structured data and comes along with lower costs and an increased number of news articles (Graefe 2016). In future, AI will be able to go far beyond that level, as the algorithms from the London company DeepMind showed (Wang et al. 2016). In 2016, the AI AlphaGo was able to win against the current world champion in four out of five Go matches. That was surprising, as the game was known to be really complex and impossible to handle by an algorithm due to its high number of possible moves and around 10700 theoretically possible games. (Wang et al. 2016). However, AlphaGo was able to learn from a large dataset of recent human matches. Even more impressive is the algorithm of AlphaGo Zero that did not require any dataset of recent human matches. It was just provided with the rules of the game and learned autonomously by playing against itself (Silver et al. 2018). Although this example seems not to be directly related to a significant business case, it provides a striking impression of future AI capabilities. As research on AI and robotics advances, possible areas of application for robots increase. Based on expert assessments and job structures, Frey and Osborne (2013) found that over the next two decades robots will be qualified to take over 47% of the current US jobs and the World Bank expects 57% of jobs in OECD countries could be automated within the next 20 years (World Development Report 2016). Service robots are on the rise and sales volumes are increasing significantly. Figure 1.2 shows the global revenue with professional service robots in their main applications (International Federation of Robotics 2019). From 2017 to 2018 revenues with professional service robots raised by 30.8% and reached a total of 8.5 billion U.S. dollars in 2018. Further, the International Federation of Robotics (2019) predicts a market volume of 34.7 billion U. S. dollars for professional service robots in 2020 coming along with a compound annual growth rate (CAGR) of 42.1% for the next four years. The service robot market is growing with high speed (Han et al. 2013) and service robots are becoming increasingly important during service encounters (Park et al. 2010). As depicted in Figure 1.1, many companies are already applying service robots at the customer encounter as service representatives. Service representatives shape customer experiences through their behavior, as they are the primary representative of companies (Grewal et al. 2009; Stock and Bednarek 2014). However, knowledge how service robots are accepted and how they shape
4
1
Introduction
Billions of USD 40 34.7
35 30 23.6
25 20 16.4 15 11.7 8.5
10 6.5 5 0 2017*
2018
2019**
2020**
2021**
2022**
Notes: *revised; **forecast
Figure 1.2 Professional Service Robots—Global Revenue. (see International Federation of Robotics 2019)
customer experiences with their behaviors is still scarce. Although research focuses on the creation of human-like service interactions with robots, there is not much knowledge about customer responses to service robots, yet. If firms apply service robots at the customer encounter without knowing which responses they create at the boundary to their customers, they face the risk of negative customer perceptions, even risking damage to their reputation and in worst case even losing loyal customers, due to disappointing service experiences (Duffy 2003). Recently, this might be especially risky as the press pays a lot of attention to topics around robotics and poorly programmed robots that do not match customer demands might cause ripples, such as for example the SEMMI robot at the Deutschen Bahn (Götz 2019) or the Pepper robot at the shopping mall Das Gerber in Stuttgart (Mayr and Kläsgen 2017). The SEMMI robot was developed by the German companies Fraport AG and Deutsche Bahn AG to support employees by providing information to customers at train stations and airport terminals (Götz 2019). However, the introduction of the robot in 2019 with press attending developed into a disaster: at first, the robot did not understand customers and came up with random answers (Götz 2019). At the shopping mall Das Gerber, researchers implemented Softbank’s Pepper
1.2 Scientific Relevance
5
robot to entertain customers and provide information. However, the robot had problems to understand many customers and was perceived as clumsy with a lack of intelligence (Mayr and Kläsgen 2017). These examples show that it is essential to prepare service robots properly before applying them at the customer encounter. To introduce service robots successfully, firms need to know more about the service robot acceptance of their customers. For example, it is interesting to know which artificial behaviors their customers would appreciate coming from a service robot and how the disconfirmation of customer expectations affects customer responses. Managers should also know more about how to handle service failures within human-robot interactions and how customer reactions differ compared to humanhuman interactions as good service recovery is essential for firm to keep their customers satisfied (Bitner et al. 1990). Companies will benefit from a better understanding of service robots at the frontline encounter and will be able to exploit the full potential of service robots. Therefore, it is of high importance investigating customer responses to robots at the service encounter.
1.2
Scientific Relevance
In accordance with the rapid diffusion of robots and service robots (International Federation of Robotics 2019), the number of studies in the field is rising quickly, as Figure 1.3 shows. It is based on search results with Google Scholar for the searching term “service robot” to get a rough quantitative overview about the research in the field. First research on service robots dates back to 1980 (Kelly and Huston 1980), although the first humanoid service robots were built later, such as the so-called Help Mate in 1998 (Evans et al. 1989) and Honda’s P-series starting in 1993 (Hu and Gu 2000). The extant research on service robots originates in three different research streams with focus on different research questions and varying extents. Although all of the research streams have a comparable understanding of service robots, the perspectives are based on significantly different viewpoints: – Research on robotics and IT provide the basis for any service robot. Their knowledge in the fields of robotic mechanics as well as controlling and programming of robots enable other research streams to deal with robots in general. There are many studies indicating a good understanding of the technical facets of robotic mechanics (Ceccarelli 2004; Mason 2001) as well as software programming (Biggs and MacDonald 2003; Brooks 1986; de Wit
6
1
Introduction
Number of Publications
3500 3000 2500 2000 1500 1000 500 0 1990
1995
2000
2005
2010
2015
2020
Notes: Search results for the term “service robots” at Google Scholar for each year.
Figure 1.3 Number of Research Publications on Service Robots. (According to results on Google Scholar)
et al. 2012) and even the early stages of artificial intelligence (Perez et al. 2018; Russel et al. 1995), but largely neglecting the user perspective in service settings. Human expectations and responses to robots are out of focus in this field. Although this research stream is not the focus of this thesis, it draws insights from this literature stream as it is the underlying basis for the other two research streams and any robot related research. – Human-robot interaction literature addresses general acceptance of robots (Breazeal 2003; Broadbent et al. 2010; Brooks 2002) and interaction quality (Bartneck 2002; Rehm and André 2005; Sidner et al. 2004) in experimental studies and field studies throughout fields as retail (Kanda et al. 2010; Shiomi et al. 2013), hospitality (Gockley et al. 2005; Pinillos et al. 2016), education (Conti et al. 2017; Kanda et al. 2007) and healthcare (Lee et al. 2017; Piezzo and Suzuki 2017). Although these studies do not explicitly refer to service robots and do not focus on service-specific outcomes, they reveal insights about human reactions on service robots (Ogata and Sugano 2000; Zhang et al. 2008). However, these studies predict that interactions between service robots and customers might soon be part of daily service experiences (Mende et al.
1.2 Scientific Relevance
7
2019) and therefore confirm the need for service marketing research on service robots. – Service marketing just started to consider service robots as a promising trend in the service domain (Bitner 2017) and has not yet widely examined service robots (Xiao and Kumar 2019). Most studies are conceptual work (Hoffmann and Nowak 2018; Huang and Rust 2018; Shankar 2018; Xiao and Kumar 2019) proposing to integrate service robots at the customer interaction. Mende et al. (2019) offer the first empirical work with embodied service robots relying on video recordings. However, there is no study in service marketing yet examining face-to-face interactions with physically embodied service robots yet, although physical embodiment has been shown to increase perceived social presence (Fasola and Mataric 2011). As this research is based on a multidisciplinary research approach, all research streams will be considered, aiming to close the research gap between human-robot interaction research and service marketing. On the one hand, service marketing has not yet much examined service robots; on the other hand, human-robot interaction literature already examined many details of human-robot interaction having the potential to be transferred to service marketing. Therefore, this thesis does not focus on (the basics known from IT, Robotics, and AI research), but instead on customer responses to the application of physical service robots at the frontline service encounter. This thesis investigates customer responses and attitudes toward service robots from a holistic perspective and addresses the following shortcomings in extant research: – There are no studies examining face-to-face human-robot interactions in a natural setting at the customer frontline encounter yet. Studies such as Mende et al. (2019) examine interactions with robots; however, they rely on video recordings instead of face-to-face interactions. This thesis directly compares human-robot interactions to human-human interactions in various service settings and introduces a rarely researched topic to service marketing research. – Extant research on human-robot interaction mainly focused on basic customer responses but did not derive specific conclusions for service marketing and did not apply natural service settings. Recently, first marketing research just started to examine robots at the service encounter (Mende et al. 2019). – Despite conceptual research (van Doorn et al. 2017; Wirtz et al. 2018; Xiao and Kumar 2019) and first video-based experiments (Mende et al. 2019), service marketing has not yet examined face-to-face interactions of customers with physical robots.
8
1
Introduction
– This thesis includes a theoretical foundation to explain customer reactions on service robots, relying psychological mechanisms and theoretical concepts with roots in social psychology. This includes script theory (Tomkins 1978), dissonance theory (Festinger 1957), expectation disconfirmation paradigm (Oliver and DeSarbo 1988) as well as computers-are-social-actors theory (Reeves and Nass 1996) and the uncanny valley paradigm (Mori 1970).
1.3
Goals and Research Questions
As presented in the previous chapters, customer reactions on humanoid service robots at the frontline service encounter are highly relevant for managerial practice as well as for scientific research. The main goal of this thesis is to answer the question how the application of humanoid service robots at the service encounter affects customers and how customer reactions can be shaped. To address the encountered research gap, this thesis pursues two major goals that are depicted in Figure 1.4 with their corresponding research questions. The first major goal relates to the analysis of customer attributions and expectations towards service robots. To achieve that goal, two studies were designed based on established theories and concepts. The first study proposes a conceptual model on robot acceptance based on the well-established technology acceptance model (TAM) and qualitative interviews. It aims at identifying relevant reference categories and comparing their effect on customer acceptance of service robots in an experimental setting. An additional study with an international focus was designed to identify customers’ expectations toward service robots and to identify cross-cultural differences in robot attributions based on Hofstede’s cultural dimensions. It was carried out as an international survey study. The first major goal can be achieved by answering the corresponding research questions depicted on the left-hand side of Figure 1.4. The second major goal is the analysis of customer responses to humanoid robots at the service encounter. To reach this goal, two experimental studies with face-to-face HRI were conducted in a natural experimental service setting. The third study relies on the computers-are-social-actors paradigm (CASA) and on the uncanny valley paradigm to examine differences between HHI and HRI. The effects of discrepancy between customer expectations and perceptions are conceptualized based on expectation disconfirmation paradigm. Results are analyzed through polynomial modelling with surface response method analysis, whose methodological foundations will be discussed beforehand. Additionally, a fourth
1.3 Goals and Research Questions
9
Major Goal 1:
Major Goal 2:
Analysis of customer attributions and expectations towards service robots
Analysis of customer responses to humanoid robots at the service encounter
• What reference categories are respon-
• Can a service robot delight customers
sible for customers’ acceptance of a service robot?
• How do various reference categories affect a customer acceptance of service robots?
• What do customers expect from humanoid robots at the service encounter?
• How do cultural dimensions affect attributions towards service robots?
with innovative service behavior?
• How does customer delight differ in response to innovative service behavior in HRI as opposed to HRI?
• How does the discrepancy of customer expectations regarding innovative service behavior affect customer delight?
• Do service robots create lower customer satisfaction than frontline employees?
• How does a service failure impact customer satisfaction with a service robot compared to a frontline employee?
Figure 1.4 Major Goals of the Thesis
experimental study is designed to examine the effects of service failures comparing the service robot with a frontline employee based on script theory and cognitive dissonance theory. The results are analyzed using analysis of variance with subsequent post-hoc tests. These two studies will provide answers to the research questions on the right-hand side of Figure 1.4 to accomplish the second major goal. Figure 1.5 illustrates an overall framework of this thesis, including the four studies and their conceptual association. It depicts the main relations between the variable categories and reveals the association between customer expectations and customer reactions towards humanoid service robots at the frontline service encounter. As described above, four independent studies were set up to address all research questions and empirically assess the research framework of this thesis. Table 1.1 gives an overview about the individual studies and their publication.
10
1
Customer Expectations Towards Service Robots
Introduction
Customer Responses Towards Service Robots Study 1
Expected Roles
General
Robot Acceptance Service Outcomes
Cultural Dimensions
Expected Characteristics
Customer Satisfaction Study 3
Study 2 Expected Service Behavior
Innovative Service Behavior
Customer Delight
Type of Service Representative Service Setting
Service Appropriateness Study 4
Figure 1.5 Overarching Research Framework of the Thesis
1.4
Structure
The structure reflects the preceding goals and concept of the thesis. The organization of the studies is illustrated in Figure 1.6. This first introduction chapter outlines the managerial and scientific relevance of service robots and introduces the major goals and research questions of the thesis. Chapter 2 provides a conceptual background as basis for the subsequent chapters that focus on the major goals of the thesis by analyzing the research questions. It defines and differentiates the term service robot and introduces the major theoretical backgrounds, contributing to a profound postulation of coherences in the following chapters based on a strong theoretical foundation. Ensuing this chapter provides an extensive overview about extant research in the field of service robots. The two research streams of service marketing and human-robot interaction are framing this thesis and will be analyzed. Chapter 3 focuses on construct measurements giving an overview on the basics and operationalizing the main construct applied within this study. Further, it introduces the major data analysis methods that are applied in the result section of the subsequent studies.
1.4 Structure
11
Table 1.1 Study Overview Study 1
Study 2
Short Title
A Service Robot Acceptance Model
Service Robots: The Impact of Customer A Cross Country Innovative Responses to Comparison Service Behavior Service Robots by Robots on Customer Delight
Study 3
Study 4
Authors
Stock, Ruth Merkle, Moritz
Homburg, Nadine Merkle, Moritz
Stock, Ruth Merkle, Moritz
Merkle, Moritz
Publication
IEEE International Conference on Pervasive Computing and Communications (PERCOM) 2017
52nd Hawaii International Conference on System Sciences (HICSS) 2019
International Conference on Information Systems (ICIS) 2018 Submitted to Journal of Product Innovation Management (JPIM)
52nd Hawaii International Conference on System Sciences (HICSS) 2019
Own 50% Contribution
30%
50%
100%
Robot Type
Nao Robot (Softbank)
N/A
Pepper Robot (Softbank)
Pepper Robot (Softbank)
Study Type
Experimental Study N = 90
Survey Study N = 610
Experimental Study N = 269
Experimental Study N = 120
The following two chapters describe the two studies related to the first major goal and focus on the analysis of customer attributions and expectations towards service robots. Chapter 4 postulates and empirically assesses a Service Robot Acceptance Model and focuses on Customer Acceptance of Humanoid Robots During Service Encounters, whereas chapter 5 contributes a Cross-Country Comparison of Attitudes toward Humanoid Robots. Chapter 6 and 7 comprise two studies with focus on customer responses to service robots at the frontline service encounter, comparing human-human interactions with human-robot interactions regarding the service outcomes of customer
12
1
Chapter 2 2.1 2.2 2.3 Chapter 3 2.1 2.2 Introduction to Study
Major Goal 1
Chapter 4 Section 4.1 Chapter 5 Section 5.1
Major Goal 2
Chapter 6 Section 6.1 Chapter 7 Section 7.1 Chapter 8 8.1 8.2 8.3 8.4
Introduction
Conceptual Background Basic Definitions Theoretical Background Literature Review Method Construct Measurement Data Analysis Methods Conceptual Background
Methodology
Results
Discussion
Study 1: A Service Robot Acceptance Model Section 4.2
Section 4.3 4.4 and 4.5
Section 4.6
Section 4.7
Study 2: Cross Country Comparison of Service Robots Section 5.2 and 5.3
Section 5.4
Section 5.5
Section 5.6
Study 3: The Impact of Innovative Service Behavior Section 6.2 and 6.3
Section 6.4
Section 6.5 and 6.6
Section 6.7
Study 4: Customer Responses to Service Robots Section 7.2 7.3 and 7.4
Section 7.5
Section 7.6
Section 7.7
Overall Discussion Overall Scientific Contribution Overall Managerial Contribution Overall Limitations and Future Areas of Research Conclusion
Figure 1.6 Organization of the Studies
1.4 Structure
13
satisfaction and customer delight. Chapter 6 sheds light on The Impact of Innovative Service Behavior by Robots on Customer Delight and examines the effects of discrepancy between customer expectations and perceptions on customer responses. The following chapter 7 extends the service setting by adding a service failure and the corresponding Customer Responses to Service Robots—Comparing Human-Robot Interaction with Human-Human Interaction. Chapter 8 presents the overall discussion of this thesis. The overall contributions are discussed regarding scientific contribution in section 8.1 and managerial contribution in section 8.2. Further, section 8.3 provides future areas of research and outlines the overall limitations of the studies conducted. Finally, section 8.4 draws an overall conclusion including the major findings of this thesis.
2
Conceptual Background
The second chapter builds the theoretical and conceptual bases for the examination of the research questions dealing with customer expectations and customer responses on humanoid service robots at the frontline service encounter (see section 1.3). Therefore, the key objective of this chapter is the development of a consistent understanding of relevant terms, providing an overview on previous findings in extant research and forming a theoretical foundation for the further examination of the research questions. First, section 2.1 defines the basic terms of the thesis, differentiates these terms from related definitions and provides a general categorization of robots with special focus on service robots. The subsequent section provides a profound analysis of extant literature, revealing the latest state of research. As the topic of the thesis is located at the intersection of two research domains, both will be analyzed. This concerns human-robot interaction literature on the one hand and research in the field of service marketing regarding robots on the other hand (see section 2.2). Section 2.3 introduces the theoretical framework of the thesis by proximately introducing the relevant primary theories. Thus, this section lays the foundation to examine the specific customer expectations and reactions on service robots and supports the postulated coherences based on a comprehensive description of the underlying theories.
© The Author(s), under exclusive license to Springer Fachmedien Wiesbaden GmbH, part of Springer Nature 2021 M. Merkle, Humanoid Service Robots, Neue Perspektiven der marktorientierten Unternehmensführung, https://doi.org/10.1007/978-3-658-34440-5_2
15
16
2
2.1
Basic Definitions
2.1.1
Robots
Conceptual Background
Although the term robot is derived from the Czech “robota”, meaning “worˇ ker” (Capek 2001), nowadays, a robot refers to an “automatically controlled, reprogrammable multipurpose manipulator programmable in three or more axes” (ISO 8373:2012). This definition applies wide terms, as it just requires a certain mechanical component with a certain level of intelligence. As this thesis focuses on humanoid service robots, this section will further narrow and specify the underlying definition. This section begins with an overview about definitions of the term service robots in literature followed by a categorization of robots according to fields of application and according to the outer appearance.
2.1.1.1 Definition of Service Robots in Literature The examination of service robots with corresponding customer expectations and responses requires a distinct definition of the term service robot. In a first step, the inconsistent application of the term service robot in current research is elaborated, that hampers the interpretation and the comparability of previous findings. In the next step, the term service robot is specified for the further use in this thesis based on extant definitions. Due to the outstanding relevance of service robots (see section 1.1), current research already addressed the topic in different contexts (see section 1.2). However, on closer inspection of current literature we find a manifold understanding of the term service robot as shown in Table 2.1. According to the relevant research streams in the field of robotics and IT, human-robot interaction and service marketing, every field came up with different definitions for service robots with diverse perspectives. Especially in the field of robotics and IT, many definitions are based on the distinction from industrial robots, defining service robots as any type of robot that is not considered as industrial robot (Asami 1994; ISO 8373:2012), whereas other more refined definitions across all research fields are based on a more detailed description of the system and its corresponding skills (Engelhardt and Edwards 1992; Schraft et al. 1993; Wirtz et al. 2018). In contrast, service marketing does not consider service robots primarily from a technical perspective, but rather from a service provider perspective or a customer perspective. Therefore, first definitions in service marketing focus on the purpose and the application of service robots (Ivanov et al.
2.1 Basic Definitions
17
2017; Wirtz et al. 2018; Xiao and Kumar 2019) as well as their outer appearance (Mende et al. 2019). As this thesis is also based in the field of service marketing and focuses on the application of service robots by service providers and corresponding customer expectations and responses, it relies on a more specific definition than just the distinction from industrial robots and renounces the overly technical definition of robotic systems. This thesis is rather based on a definition focusing on a general description of the robotic system with its basic skills and field of application (see Wirtz et al. 2018), as well as its appearance, defining service robots as “systembased autonomous and adaptable interfaces that interact, communicate and deliver service to an organization’s customers” (Wirtz et al. 2018, p. 909) and focusing on service robots with a humanoid appearance.
2.1.1.2 Categorization of Robot Types After defining the term service robot for this thesis adopting a narrow understanding of service robots (Wirtz et al. 2018) in professional applications, this section differentiates the term from related types of robots and provides a general categorization of robots. In line with the provided definition that focuses on the purpose of application as well as on the appearance, the following section (2.1.1.2.1) categorizes robots based on the application purpose and the section after next (2.1.1.2.2) categorizes robots based on the outer appearance.
2.1.1.2.1 Categorization Based on Application Purpose The categorization of robots according to the purpose of application is based on ISO standard 8373 and classifies robots in industrial robots and service robots (ISO 8373:2012). Service robots in turn are classified in personal service robots and professional service robots (ISO 8373:2012) as illustrated in Figure 2.1. An industrial robot is defined as “automatically controlled, reprogrammable, multipurpose manipulator, programmable in three or more axes, which can be either fixed in place or mobile for use in industrial automation applications” (ISO 8373:2012, 2.9) and includes the manipulator (including actors) and the controller (including hardware and software communication interfaces) (ISO 8373:2012). The main features of these robots are on the one hand the opportunity to apply many diverse effectors allowing manifold applications. On the other hand, industrial robots are able to reach highest positioning accuracy and repeatability at high precision leading to a reduction of costs, increasing productivity and quality as well as reducing employees’ workloads (Siciliano et al. 2009). In 2018, more than 422,000 industrial robots were installed worldwide mostly in countries as China,
18
2
Conceptual Background
Table 2.1 Exemplary Definitions of Service Robots Reference
Definition
Asami 1994, p. 22
“Consider the service Robotics & IT robot to be a generic term covering all robots that are not for industrial use, i.e. that are not intended for rationalizing production at a manufacturing site.”
Research Field
Perspective Distinction from Industrial Robots
Engelhardt and Edwards 1992, p. 316
Service robots are HRI “systems that function as smart, programmable tools that can sense, think, and act to benefit or enable humans or extend/enhance human productivity”.
System, Skills
ISO 8373:2012, 2.10
In contrast to Robotics & IT industrial robots, a service robot is a robot “that performs useful tasks for humans or equipment excluding industrial automation applications”.
Distinction from Industrial Robots
Ivanov et al. 2017, p. 1503
“Service robots are designed to support and service humans through physical and social interactions.”
Service Marketing
Purpose, Application
Mende et al. 2019, p. 535
“Humanoid service Service Marketing robots […][are] robots with a human-like morphology such as a face, arms, and legs”.
Appearance
(continued)
2.1 Basic Definitions
19
Table 2.1 (continued) Reference
Definition
Schraft et al. 1993, p. 163
“A service robot is a Robotics & IT freely programmable kinematic device which performs services semi or fully automatically.”
Research Field
Perspective System, Skills
Wirtz et al. 2018, p. 909
“Service robots are system-based autonomous and adaptable interfaces that interact, communicate and deliver service to an organization’s customers.”
Service Marketing
System, Skills, Application
Xiao and Kumar 2019, p. 2
“Service robots refer Service Marketing to those robots that are used for service applications, especially those that perform customer-oriented service tasks in many business applications.”
Application
Japan, the United States, Republic of Korea and Germany with the automotive industry as the largest adopter of industrial robots (IFR 2019). In contrast, service robots (ISO 8373:2012, 2.10) are applied to perform useful tasks for humans besides manufacturing, inspection, packaging and other industrial automation applications and can be further segmented in personal service robots and professional service robots (ISO 8373:2012). While personal service robots (ISO 8373:2012, 2.11) are designed for personal use in non-commercial applications for laypersons, professional service robots (ISO 8373:2012, 2.11) are designed for professional use in commercial applications, usually operated by a trained operator. Personal service robots are not pursuing commercial aims but are entering daily lives of ordinary people playing a beneficial role (Breazeal 2005) in many fields
20
2
Conceptual Background
Categorization of Robots Based on Application Purpose ISO 8373:2012 (2.9)
ISO 8373:2012 (2.10)
Industrial Robots
Service Robots ISO 8373:2012 (2.11)
Personal Service Robots
ISO 8373:2012 (2.12)
Professional Service Robots
Figure 2.1 Categorization of Robot Types According to Application Purpose
such as household robots, entertainment robots, care robots and many more. Household robots as the robotic vacuum Roomba, the robotic pool cleaner Aquabot or the robotic lawn mower Automower Solar Hybrid have in common that they support private households but do not yet work completely autonomous (Syridon and Eleftheria 2012). Sustainable prices combined with advanced technology made these robots ready for mass market (ibid). Other personal service robots focus rather on entertainment such as the robotic dog Aibo that shows emotions and can be trained or soccer robots playing matches against each other (Veloso 2002). Even in the field of elderly care, first personal service robots have already been applied: the robotic seal Paro improves the well-being as companion for elderly with positive effects on communication and activity of users (Argall and Billard 2010) and the Care-O-bot supports elderly in daily life with homecare and housekeeping (Graf et al. 2009). Professional service robots are generally designed for professional use in commercial applications and this thesis further specified the term with a strong focus on service marketing as “system-based autonomous and adaptable interfaces that interact, communicate and deliver service to an organization’s customers” (Wirtz et al. 2018, p. 909). Other robots that might fall under a broader definition of professional service robots such as field robots, defense robots or medical robots (Fernando et al. 2017) are not considered as professional service robots within the narrow definition of this thesis. These professional service robots are already applied in many industries such as retail, hospitality, tourism and many more as already mentioned in section 1.1 with a variety of exemplary applications in various fields. As this thesis focuses on professional service robots rather than
2.1 Basic Definitions
21
industrial robots and personal service robots, section 2.2 provides a detailed analysis of the current state of research on professional service robots according to the definition of this thesis.
2.1.1.2.2 Categorization Based on Appearance Besides the previous categorization based on the purpose of application, this section categorizes robots according to their outer appearance and focuses especially on the different morphological approaches. As this thesis focuses on customer expectations and reactions, it is essential to take a closer look on robotic appearances: the robotic appearance influences customer expectations (Lohse 2007) and affects customer reactions on robots (Mori et al. 2012). A general distinction can be made between robots with an all-functional design and robots with a design that also focuses the morphologic appearance. Especially industrial robots are mostly designed with a strong focus on function (Mori et al. 2012), as many industrial robots just have little to none interaction with humans and no interaction with customers. For service robots that are applied at the frontline service encounter with the customer, it is beneficial to design the robot according to a morphologic design, i.e. focusing on the form and structure of robots, as it helps to establish expectations and biases interactions (Fong et al. 2003). Regarding the morphologic design (see Fong et al. 2003) of robots we can distinguish between caricatual robots (Sebastian et al. 2015), zoomorphic robots (Bernstein et al. 2008) and anthropomorphic robots (Złotowski et al. 2015). Anthropomorphic robots can further be distinguished between humanoid robots (Gupta et al. 2006) and android robots (Nishio et al. 2007). Figure 2.2 gives an overview about the categorization of robots according to their appearance. Caricatural robots have an appearance that is not—in contrast to zoomorphic or anthropomorphic robots—based on a real living being. The main focus of caricatural robots is the ability to display certain motions in an exaggerating way (Sebastian et al. 2015). Therefore, these robots are defined as “a non-humanoid robot that can show simplified humanoid motions in exaggerated ways” (Sebastian et al. 2015, p. 324). Further, these stereotypical appearances might be applied to create intended interaction biases and to distract attention from robotic characteristics (Fong et al. 2003). This type of robot is mainly applied for entertainment purposes, offers various designing options, reduced manufacturing costs and easy accessibility for elderly and people with less technological affinity (Sebastian et al. 2015). Zoomorphic robots are designed to resemble animals such as the robotic dog Aibo from Sony (Veloso 2002). These robots rely on a zoomorphic design to establish human-animal relationships. Zoomorphism is the “attribution of animal traits
22
2
Conceptual Background
Categorization of Robots Based on Appearance
Functional Design
Morphologic Design
Caricatural
Anthropomorphic
Zoomorphic
Humanoid
Android
Figure 2.2 Categorization of Robot Types According to Appearance
to human beings, deities, or inanimate objects” (Bernstein et al. 2008). Compared to android robots, the avoidance of the uncanny valley (see section 2.3) might be easier as human-animal relationships are simpler than human-human relationships (Fong et al. 2003), leading to higher user familiarity with the zoomorphic robot. The vast majority of professional service robots with a morphologic design nowadays are rather designed anthropomorphic (Pan et al. 2015; Pinillos et al. 2016; Sabelli and Kanda 2016) than caricatural or zoomorphic. As “anthropomorphism is the act of attributing humanlike qualities to non-human organisms or objects” (DiSalvo and Gemperle 2003, p. 68), the goal is to design anthropomorphic robots resembling humans that users behave as naturally as possible with robots treating them like humans (Fong et al. 2003). Depending on the degree of human-likeness, anthropomorphic robots can further be distinguished in humanoid robots and android robots. There are many definitions on humanoid robots as sub-category of anthropomorphic robots. Generally, “humanoid robotics includes a rich diversity of projects where perception, processing and action are embodied in a recognizably anthropomorphic form in order to emulate some subset of physical, cognitive and social dimensions of the human body” (Gupta et al. 2006). In comparison with android robots, humanoid robots just roughly resemble human characteristics but are not designed to look exactly like humans. Therefore, humanoid robots do not have human details such as skin, hair or eyelashes (Haring et al. 2013). The
2.1 Basic Definitions
23
most widespread service robot types in this category are Softbank’s Nao and Pepper robot (Pandey and Gelin 2018). Both of them have a roughly human-like shape, arms, eyes, a face and a mouth. Both robot types will be applied in the experimental studies of this thesis. Android robots, however, are designed to strongly resemble the appearance of a human (Nishio et al. 2007). Besides the high technological hurdles to produce android robots that really strongly resemble humans, it is a challenge to design android robots in a way that is accepted by humans (Nishio et al. 2007) and avoids the negative effects of the uncanny valley (see section 2.3) (Mori et al. 2012). This category of robots is associated with high costs and therefore not widely spread in service settings yet.
2.1.2
Services
This thesis focuses merely on service robots providing services to customers rather than industrial robots for the production of goods. Thus, it is essential for this thesis to distinguish between services and goods (Lovelock and Gummesson 2004), although this distinction faces some challenges (Homburg and Garbe 1999). Service research determined a distinct definition of services as “application of specialized competences (…) through deeds, processes, and performances for the benefit of another entity” (Vargo and Lusch 2004, p. 326). This definition of services can still be categorized in business services for companies and consumer services delivering service to individuals or groups of individuals (Homburg and Garbe 1999). Subsequently, the term services just refers to consumer services in this thesis and business services will not be further examined in the context of service robots. This section starts with an overview about the characteristics of services relying on the IHIP criteria considering the intangibility, heterogeneity, inseparability, and perishability (Fisk et al. 1993) of services. Further, it introduces innovative service behavior as a preferable behavioral cue for service representatives and concludes with the exceptional case that services fail occasionally and result in a service failure.
2.1.2.1 Service Characteristics Distinguishing services from goods, literature predominantly specifies four constituent characteristics of services that are referred to as IHIP criteria: intangibility, heterogeneity, inseparability and perishability (Lovelock and Gummesson 2004).
24
2
Conceptual Background
Intangibility means that services are not comprised of touchable physical products and it is closely linked to the immaterial nature of a service (Lovelock and Gummesson 2004). Intangibility challenges customers as it makes it difficult to verify service quality in advance of the service provision (Homburg 2016). Therefore, service firms strive to establish a positive image, causing customers to trust in the service quality and building up positive expectations toward future services and service representatives (Setiawan and Sayuti 2017). It is hardly possible to standardize services leading to a wide variety and heterogeneity of services due to variations in the situation and in employee characteristics (Lovelock and Gummesson 2004). Service firms attempt to keep services as constant as possible by setting strict process guidelines and providing detailed service scripts to their service representatives (Stewart 2003). Inseparability between the service provision and the service consumption connotes that the provision of a service is not possible without including customers into the service process (Vargo and Lusch 2004). Inevitable interaction with the customer in turn means that in most cases a service representative is involved in the service provision who is considered as the face of the organization in the eyes of the customer (Homburg et al. 2009). Thus, it is essential for service firms to understand customer responses toward service representatives (Bettencourt and Brown 2003). Perishability means that services cannot be produced ahead of time and then stored to satisfy later demands (Moeller 2010). That makes it difficult to handle fluctuations in demand over time, as lacking service capacities might lead to unrealized sales potential (Wyckham et al. 1993). Considering these service characteristics and the associated challenges for service firms, it is worth considering service robots as service representatives to meet these challenges. Service robots may help to counter high service heterogeneity, as they can be tied more strictly to service scripts and have a more constant behavior as human FLEs. Further, service robots could be applied to extend service capacities and to cope with high fluctuations in demand of services. However, it is essential to be aware of customer responses to service robots as they are inseparable from service consumption and represent the service firm at the frontline service encounter.
2.1.2.2 Innovative Service Behavior Nowadays, products are increasingly interchangeable (Payne et al. 2008) so that companies find it hard to differentiate from their competitors merely through their products and experience an increasing need to set itself apart from competitors
2.1 Basic Definitions
25
through soft factors such as good service. Therefore, companies started to encourage their employees’ innovative service behavior (Chang et al. 2011). Innovative service behavior “defines as the extent to which FLEs generate new problemsolving ideas and transform these into uses during the service encounter” (Stock 2015, p. 574). Service representatives can express innovative service behavior by quickly thinking up new ideas, loving to think up new ways of doing things (Hogan and Hogan 1995), being full of ideas, coming up with something new (Lee and Ashton 2004), finding something of interest in any situation and being able to come up with new and different ideas (Peterson and Seligman 2004). This inspires customers, helps them to solve specific problems, creates pleasurable experiences for them (Stock et al. 2017) and leads to successful long-term relationships with customers (Coelho et al. 2011), customer delight (Oliver et al. 1997) and customer loyalty (Stock et al. 2017). Innovative service behavior depends on job characteristics (De Jong and Kemp 2003), employee engagement (Slåtten and Mehmetoglu 2011) and employees’ affective states (Stock 2015). It is hard to implement innovative service behavior straightforward among employees, as their lack of resources and energy may severely hamper innovative service behavior at the frontline encounter (Stock et al. 2015). Service robots, however, show a constant behavior, rely constantly on their script and do not depend on personal daily conditions such as rather erratic human frontline employees. Integrating recent creative AI in service robots (see section 1.1), enables them to apply innovative service behavior cues that might lead to delighted customers (Oliver et al. 1997) and increase their loyalty with the service provider (Stock et al. 2017). Study 3 will take this into consideration examining customer reactions to innovative service behavior in human-human and human-robot interactions (see chapter 6).
2.1.2.3 Service Failures Although companies and their service representatives try their best in providing satisfying services, service failures occur regularly due to unavailable or unreasonably slow services or unsolicited behaviors of frontline employees (Stock 2018) and are inevitable (Berry and Parasuraman 1991). Therefore, a holistic investigation of customer responses in service settings should not ignore this incident. Service failures are “activities that occur as a result of customer perceptions of initial service delivery behaviors falling below the customer’s expectations”
26
2
Conceptual Background
(Holloway and Beatty 2003, p. 93) and cause negative consequences for the service firm. Service failures lead to negative word of mouth, a decrease in customer loyalty and may cumulate in diminished firm performance (Bitner et al. 1994; Tax and Brown 1998). As this thesis examines customer reactions at the service encounter, Study 4 takes service failures at the service encounter into consideration and compares customer responses in human-human settings with human-robot settings (see chapter 7).
2.1.3
Customer Expectations and Responses
Customer expectations and customer reactions are closely linked because customer expectations have a direct influence on corresponding customer responses. Customer expectations are defined as each customers’ “probability density function that describes the relative likelihood that a particular quality outcome will be experienced” (Rust et al. 1999, p. 77). Usually these expectations are based on prior experiences in comparable situations and customers rely on their experiences as prediction for future situations (Rust et al. 1999). As frontline service robots are a new trend in the service domain (Bitner 2017), customers do not yet have many experiences with this new technology. Therefore, it remains unclear how customers form their expectations towards frontline service behaviors. Study 1 identifies the major categories and links the resulting expectations to respective customer responses (see chapter 4), whereas study 2 reveals that cultural differences affect customer expectations towards frontline service robots (see chapter 5). Further, these expectations serve as reference to judge service performance (Zeithaml et al. 1993), as the interplay of customer expectations and actual perceptions affects customer responses according to expectation disconfirmation theory (Oliver 1980; Oliver and DeSarbo 1988). Customer responses comprise a variety of customer reactions to services including satisfaction responses, emotional reactions and behavioral responses (Sparks and Fredline 2007). For a long time, customer satisfaction was regarded as the most relevant customer response as performance goal for service firms (Howard and Sheth 1969). Later the focus of service firms evolved and they strived for customer delight besides customer satisfaction (Keiningham and Vavra 2001). Therefore, this thesis will consider both customer responses, customer delight (see study 3) and customer satisfaction (see study 4), as the performance of a service robot cannot be measured with the robot itself, but has to be measured
2.1 Basic Definitions
27
via corresponding customer responses (Bartneck at al. 2009). Further, this thesis compares customer responses in human-robot interactions at the service encounter comparing it with customer responses in human-human interactions.
2.1.3.1 Customer Satisfaction Customer satisfaction is an important measure for companies as it is linked with positive effects on repeated purchases (Szymanski and Henard 2001), retention (Bolton 1998), loyalty (Anderson and Sullivan 1993), and company profitability (Bernhardt et al. 2000). Customer satisfaction with a frontline service representative is defined as “the customer’s evaluation of her or his interaction with a frontline [service representative]” (Stock and Bednarek 2014, p. 402). Regarding customer responses, it may be noted that customer satisfaction is one of the most widely studied constructs in service marketing (Homburg and Stock 2004). However, the effect of customer satisfaction on positive effects as loyalty (Fullerton and Taylor 2002) and repurchase intentions (Mittal and Kamakura 2001) is non-linear and much stronger for high levels of satisfaction (Finn 2005). Therefore, service science conceptualized a further customer response with a much stronger effects on service companies (Schneider and Bowen 1999), including an emotional response component: customer delight (Finn 2005).
2.1.3.2 Customer Delight Although customer delight and customer satisfaction are interrelated they are different constructs (Wang 2011). While customer satisfaction just relates to a customer’s evaluation of the interaction with the service representative (Stock and Bednarek 2014), customer delight also refers to a customer’s excitement and pleasure during the interaction with the service representative (Barnes et al. 2003). Thus, customer delight is “a profoundly positive emotional state generally resulting from having one’s expectations exceeded to a surprising degree” (Rust and Oliver 2000, p. 86) and is directly linked to positive effects on service companies such as an increase in customer loyalty (Coelho et al. 2011). According to this definition, customer delight consists of two dimensions (Finn 2005): a cognitive dimension (Howard and Sheth 1969) and an emotional dimension (Westbrook and Reilly 1983). This implies that although a service representative that meets the customer expectations and leads to customer satisfaction, this does not necessarily delight the customer if there is no emotional response to a surprise (Finn 2005). Study 3 examines the effects of surprising discrepancies to customer expectations on customer delight with a service representative comparing human-robot interactions with human-human interactions.
28
2.2
2
Conceptual Background
Literature Review
Subsequent to the definitional basis in the previous section 2.1, this section provides an overview on the current state of research regarding professional service robots at the customer encounter. This provides a sound basis for identifying the limitations of current research (section 2.2.2.3) and for the following studies within the framework of this thesis to further develop extant research and to gain further insights into customer expectations and responses toward service robots (see section 1.3). There are two major research streams in literature—service marketing literature and human-robot interaction literature—dealing with the deployment of professional service robots at the customer service encounter. Of course, there is also a wide field of research regarding robotics and IT, but there the focus is not on the customer within a service interaction, so this thesis further focuses on the first two research streams. Since this thesis focuses customer expectations and customer responses to service robots, it is appropriate to first cover research in the field of services marketing that focuses on the interaction between customers and companies (section 2.2.1). However, since this research stream has only recently discovered service robots as a promising trend (Bitner 2017), this thesis further deals with the HRI literature selectively focusing on research regarding FSRs in the service context (section 2.2.2).
2.2.1
Studies in the Field of Service Marketing
This section provides a thorough analysis of the current service marketing literature with regard to findings on service robots at the service encounter. It aims to identify extant research on potential customer expectations and responses toward service robots in service marketing literature. As this research stream just started to consider service robots as a promising trend in the service domain (Bitner 2017) and has not yet widely examined service robots (Xiao and Kumar 2019), this section provides a broad literature review in the field. First, an overview is given on the criteria and the corresponding focus of the literature review (2.2.1.1). Subsequently, the conceptual studies (2.1.1.3) and the empirical studies (2.1.1.3) in the field are analyzed and their relation to this thesis is pointed out.
2.2 Literature Review
29
2.2.1.1 Criteria and Focus of the Literature Review Besides the focus on service robots, further factors were defined for the appropriate selection of studies for this literature review. This systematic approach is necessary in order to focus on the most relevant studies and to optimize the knowledge gained as result of this review. Therefore, the following criteria are considered when reviewing service marketing literature: – – – –
Research focus on robots in the service context. Publication in top service marketing journals. Topical observation period. Method: conceptual and empirical research.
The research focus of this literature review was kept relatively broad mainly focusing on robotic search terms as service marketing literature is already focused on the service context itself and the literature focusing on robots is already sparse in this field (Xiao and Kumar 2019). To identify relevant studies according to the defined research focus, the search term ‘robo*’ was applied, including all associated terms including this fragment, such as ‘robots’, ‘robotics’, ‘service robot’, ‘human-robot-interactions’, ‘roboadvisor’ and many more. Literature was mainly accessed via the online research platform EBSCOhost including the largest business research database Business Source Premier and supplemented through research via Google Scholar, as one of the identified target journals (Journal of Service Research) is not fully accessible in the Business Source Premier database. Additional search terms such as ‘bot’, ‘agent’ and ‘automation’ had an even broader focus and corresponding results had to be filtered thoroughly as there was not necessarily a connection to service robots. The initial list of all articles was then screened manually based on the information provided in the abstracts and most of the studies had to be excluded, as they were not relevant for this thesis. This process reduced the number of studies from service literature down to 32, that were then screened full-text, finally leaving 27 relevant studies for further detailed analysis. The publication in top service marketing journals is an essential criterion for the selection of relevant studies. This literature review just considers studies published in top service marketing journals, although this rather strict focus might neclect some of the most recent research presented on conferences before subsequent journal publication. For example, most recent research examines HRI from a psychological perspective (Stock and Nguyen 2019), introduces robotic experiments to service marketing (Homburg 2018), examines HRI at the service encounter (Merkle 2019; Stock and Merkle 2918; Stock 2018) and at the workplace (Stock
30
2
Conceptual Background
et al. 2019). This thesis further relies on the ranking from ‘Verband der Hochschullehrer für Betriebswirtschaft e. V.’ and considers the relevant rankings in three sub disciplines. The required ranking in ‘VHB-JOURQUAL 3’ for consideration in the review is based on the relevance of the individual sub-disciplines. ‘Service Management’ as most relevant sub discipline for this thesis was considered for rankings of at least ‘B’ including four journals, ‘Marketing’ sub discipline was considered ranking ‘A+’ or at least ‘A’ leading to a total of eleven relevant journals and the least relevant sub discipline ‘Organization’ was just considered for one journal ranking ‘A+’. Table 2.2 provides an overview about the relevant journals and their ranking and shows how many studies were included in the further analysis from each journal. Further, this thesis focuses on a topical observation period including relevant publications since the year 2000, as the research field of service marketing just recently started to consider service robots (Bitner 2017). Research regarding service robots started to increase significantly mid of the 2000s and gained further attention in the mid of 2010s (see section 1.2, Figure 1.3). The further analysis is categorized in conceptual and empirical research methods excluding mere review articles such as Ivanov et al. (2019), Honig and Oron-Gilad (2018), and Kaartemo and Helkkula (2018) as they have a different focus of review. Conceptual studies (section 2.2.1.2) propose assumptions based on plausibility considerations leading to conceptual models that are argumentatively supported but not tested empirically (Leeflang and Koerts 1973). Empirical studies (section 2.2.1.3) include qualitative-empirical and quantitativeempirical studies. While quantitative studies analyze data with statistical methods to test proposed hypotheses, qualitative studies rely on observations and descriptive analysis without verification of causal relationships (Hanson and Grimmer 2007).
2.2.1.2 Conceptual Studies in Service Marketing The first studies in service marketing considering service robots were of rather conceptual than experimental nature (Molinari 1964). As of today, most studies in service marketing on service robots are conceptional studies (16 out of 27), proposing models based on plausibility considerations and theories (Leeflang and Koerts 1973). Although various researchers consider service robots as one of the most relevant trends for the next years in service marketing (Bitner 2017; Wirtz and Zeithaml 2018), corresponding research in service marketing is still scarce. This literature review screened the 14 top journals in the field of service marketing
2.2 Literature Review
31
Table 2.2 Overview about the Reviewed Literature Sub Discipline: Service Management Name of Journal
Acronym
VHB-JOURQUAL 3
Number of Reviewed Studies
Journal of Retailing
JR
A
1
Journal of Service Research
JSR
A
7
Manufacturing & Service Operations Management
M&SOM
A
0
Journal of Service Management
JSM
B
3
Name of Journal
Acronym
VHB-JOURQUAL 3
Number of Reviewed Studies
Journal of Marketing Research
JMR
A+
1
Journal of Marketing
JM
A+
2
Journal of Consumer Research
JCR
A+
4
Marketing Science
MS
A+
1
Journal of Applied Psychology
JAP
A
0
International Journal of Research in Marketing
IJRM
A
1
Journal of the Academy of Marketing Science
JAMS
A
5
Journal of Retailing
JR
A
1
Sub Discipline: Marketing
Journal of Service Research
JSR
A
7
Journal of Product and Innovation Management
JPIM
A
0
Journal of Consumer Psychology
JCP
A
0
Name of Journal
Acronym
VHB-JOURQUAL 3
Number of Reviewed Studies
Organization Science
OS
Sub Discipline: Organization
A+
2
Total
27
32
2
Conceptual Background
(see Table 2.2) extensively and listed every study with an even remotely relation to service robots that was published in the last two decades. Besides the scarcity of relevant studies in service marketing, this review identified 16 studies on a conceptual basis relating to robotic technologies in service settings with customers. However, these studies are quite widely scattered and most of the studies do not refer to each other either, which makes it even more obvious how heterogeneous the relevant research in the field has been so far. Table 2.3 gives an overview on the variety of theoretical foundations and the different conceptions of robots in the individual studies. The key content published in each study is summarized in the structured literature table for further analysis and contains information on the underlying conception of robots, the theoretical basis if applicable and a brief description of the proposed conceptual model. As the few extant studies are rather fragmented in their propositions, this thesis applies a narrative summary approach for the literature review (Paré et al. 2015). Reviewed studies and their conceptual frameworks relied on TAM (Bolton et al. 2018; Wirtz et al. 2018; Xiao and Kumar 2019), mostly in combination with role theory (Bolton et al. 2018; Wirtz et al. 2018) and assemblage theory (Bolton et al. 2018; Hoffmann and Novak 2018). Studies further applied automated social presence (van Doorn et al. 2017), uncanny valley paradigm (Xiao and Kumar 2019), identity theories (Reed et al. 2012) and further theories from adjacent fields. Other studies introduce new conceptual models such as the development of a job replacement theory (Huan and Rust 2018), the calculation of pricing strategies (Andreassen et al. 2018), and the development of technology-driven service strategies (Huang and Rust 2017). The proposed models in these studies can be organized in four main categories: studies (a) developing comprehensive conceptual frameworks (van Doorn 2017; Wirtz et al. 2018; Xiao and Kumar 2019), (b) focusing just on how service robots affect customers (Belk 2013; Bolton et al. 2018; Hoffmann and Novak 2018; Reed et al. 2012; Smith 2002), (c) analyzing the opportunities for process improvement in service firms (Fleming 2019; Huang and Rust 2018; Marinova et al. 2017), and (d) considering the effects of robot characteristics and environmental conditions (Huang and Rust 2017; Rijsdijk et al. 2007). There are three comprehensive conceptual frameworks in service marketing: the automated social presence model (van Doorn 2017), the service robot acceptance model (Wirtz et al. 2018) and the concept of the degree of robotics adoption (Xiao and Kumar 2019). First, the automated social presence model (ASP) is based on “the extent to which technology [such as a service robot] makes customers feel
2.2 Literature Review
33
Table 2.3 Overview on Conceptual Studies in Service Marketing Author/s Conception of (Year), Journal Robots
Theoretical Foundation
Proposed Model
Andreassen, van General automated Calculation of Oest, and service interactions Pricing Lervik-Olsen Mechanisms (2018), JSR
• Firms taking a short-term perspective do not automate their services although it would pay off in the long term. • Lower prices can be used to compensate customers for inconveniences caused by automation. • Automated services do not necessarily guarantee higher profits even if automation is more cost-efficient. • Completely human service is not always optimal, as the market becomes more sensitive to service quality. • Automated services should not always be cheaper for customers even if automated services are cheaper for the service firm.
Belk (2013), JCR
Robots as re-embodiment of the extended self
Concept of the Extended Self
Conceptual update for the concept of extended self due to digitization trends: • The basic concept of the extended self remains vital. • At first, users started with re-embodiment in the digital world via avatars. • In the future robots may become part of our extended self just as avatars are today.
Bolton et al. (2018), JSM
Service robots provide new technology-enabled services with automated social presence
TAM, Role Theory, • Service robots will replace Assemblage FLEs and start a new era in Theory, Affective which digital, physical and Events Theory social realms will become intertwined and blend into a holistic customer experience. (continued)
34
2
Conceptual Background
Table 2.3 (continued) Author/s Conception of (Year), Journal Robots
Theoretical Foundation
Proposed Model • A sense of social presence will arise between customers and service robots and increasingly characterize service settings. • This will change customer service experiences. Depending on the goals and preferences of the individual customer, the responses will differ.
Fleming (2019), Robots as Bounded OS technology with Automation the ability to Concept perform jobs of physical, cognitive and emotional kind
Demonstration how organizational forces mold the application of service robots: • Service robots provide the chance to optimize service processes by taking over common and repetitive tasks and replacing human service employees. • Automation of services will still take time due to employee resistance.
Hoffmann and Novak (2018), JCR
New technologies such as robots have the potential to revolutionize consumer experiences: • As consumers actively interact with new technologies such as robots, the traditional human-centric conceptualization of consumer experiences may not be sufficient to conceptualize consumer experiences in the future. • The emergence of consumer-robot assemblages strongly implies that robots play a role in consumption-related processes.
Autonomous robots Assemblage as smart objects Theory that can work in collaboration with humans
(continued)
2.2 Literature Review
35
Table 2.3 (continued) Author/s Conception of (Year), Journal Robots
Theoretical Foundation
Huang and Rust Service robots (2017), JAMS performing physical tasks at the service encounter without involving human FLEs
Development of • Services are categorized by Technology-Driven customer attributes (relational Service Strategies vs. transactional) and service attributes (standardized vs. personalized) requiring four different service strategies. Service robots suit best for standardized transactional services. • Advances in technologies such as service robots empower firms to leverage the full benefit of technology based on the strategic position. • Service robots can enhance productivity when used for standardization and enhance customer satisfaction when used for personalization. • Robots can replace or augment frontline service employees and facilitate thinking and feeling.
Huang and Rust Robots as Development of a (2018), JSR technology that can Job Replacement perform physical Theory tasks and operate autonomously
Proposed Model
• Automated service delivery technologies such as service robots will streamline service processes and will be applied widely starting with mechanical tasks and proceeding with tasks requiring higher intelligence. • FLEs should rather focus on soft skills such as intuitive and empathic skills. (continued)
36
2
Conceptual Background
Table 2.3 (continued) Author/s Conception of (Year), Journal Robots
Theoretical Foundation
Proposed Model
Marinova et al. (2017), JSR
Service robots as technology to empower frontline interactions
Pragmatic Learning Theory
Smart technologies such as service robots are rapidly transforming frontline customer interactions: • Smart technologies such as service robots can substitute or complement FLE’s efforts to deliver customized service over time. • Service robots can enable learning from and across customers and interactions and therefore resolve the tension between service efficiency and effectiveness. • Robots may elevate service effectiveness and efficiency empowering frontline interactions.
Olazabal, Cava, and Sacasas (2005), JAMS
Search robots for online shopping customers
Judicial Decisions
• Search robots are not allowed to hijack customers from the intended shopping destination by confusing customers with unlabeled banner ads. • Search robots have to be fair and transparent to customers, labelling banner ads properly in a way that they can be identified with a particular company by the customer. (continued)
2.2 Literature Review
37
Table 2.3 (continued) Author/s Conception of (Year), Journal Robots
Theoretical Foundation
Proposed Model
Reed et al. (2012), IJRM
Service robots as anthropomorphic social robots
Identity Theories
Identity theories argue that an activated identity will influence judgement and behavior to the extent that the identity is relevant to the domain: • Robots will increase their identity relevance by becoming ever more useful tools by increasing autonomy and effectiveness. • However, robots might also undermine key aspects of what it means to perform a certain identity and anthropomorphism triggers disadvantageous social comparisons threatening users if robots happen to be stronger and better.
Rijsdijk, Hultink, and Diamantopoulos (2007), JAMS
Robots as Adoption autonomous Theory intelligent technology product
• Robot intelligence is formed by six dimensions: autonomy, ability to learn, reactivity, cooperation, humanlike interaction and personality. • Robot intelligence positively affects the mediators ‘relative advantage’, ‘compatibility’ and ‘complexity’ that in turn positively affect customer satisfaction with it.
Shankar (2018), Robots as one out JR of many options to apply AI in retailing
Providing a Framework to Leverage AI in Retailing
AI might influence retailing in several areas: • Customers: better service experience, efficiency and satisfaction, getting help and guidance. • Retailers: analyze customer behaviors, identify new sources of revenues, increase efficiency. (continued)
38
2
Conceptual Background
Table 2.3 (continued) Author/s Conception of (Year), Journal Robots
Theoretical Foundation
Proposed Model
Smith (2002), JAMS
Based on a Literature Review
• Shopbots will radically reduce customer search costs and shift the balance of power in online markets toward customers. • Shopbots will reduce retailer opportunities to differentiate their products and therefore place significant pressure on retailer margins. • Retailers retain a variety of strategic options to mitigate this pressure by strategic pricing, price discrimination, bait-and-switch, as well as search obfuscation.
Automated Social Presence, Social Cognition, Psychological Ownership Theory
• Robotic technologies have the potential to facilitate the interactions between employees and customers. • Service robots can deliver customer service collaborating with FLEs or replacing FLEs. • Customer responses to service robots (satisfaction, loyalty/repatronage, engagement and well-being) depend on customer attributes (relationship orientation, anthropomorphization of the robot, technological readiness) and robot characteristics (warmth, competence, receptiveness, attractiveness, manipulability).
Service robots as internet shopbots applied in online shopping
Van Doorn et al. Service providing (2017), JSR humanoid robots
(continued)
2.2 Literature Review
39
Table 2.3 (continued) Author/s Conception of (Year), Journal Robots
Theoretical Foundation
Wirtz et al. (2018), JSM
Technology • Service robots should be Acceptance Model, applied for tasks with simple Role Theory emotional-social complexity. In combination with a human, robots should be applied for services with high emotional-social complexity and high cognitive-analytical complexity. • Customer acceptance of service robots depends on its functional elements (perceived ease of use, perceived usefulness, subjective social norms), relational elements (trust, rapport) and socio-emotional elements (perceived humanness, perceived social interactivity, perceived social presence). • Service robots can optimize service processes with simple emotional-social complexity. They provide accurate service delivery, reliability, efficiency and cost-effectiveness, being convenient and fast.
Service robots are autonomous and adaptable interfaces delivering service to customers
Xiao and Kumar Service robots as (2019), JSR mechanical machines or intangible computer programs
Proposed Model
Uncanny Valley Antecedents of robots at the Theory, Innovation service encounter: Diffusion Theory, • Robot characteristics TAM (anthropomorphism, autonomy, relative advantages, and comparability) • Customer characteristics (customer readiness, self-efficacy, role clarity, motivation, and demographics) (continued)
40
2
Conceptual Background
Table 2.3 (continued) Author/s Conception of (Year), Journal Robots
Theoretical Foundation
Proposed Model Consequences of robots at the service encounter: • Customer satisfaction resulting in direct customer engagement such as repeated purchases leading to firm profits. • Customer emotions leading to indirect customer engagement such as referral, influencing and feedback.
Notes: IJRM = International Journal of Research in Marketing; JAMS = Journal of the Academy of Marketing Science; JCR = Journal of Consumer Research; JM = Journal of Marketing; JMR = Journal of Marketing Research; JR = Journal of Retailing; JSM = Journal of Service Management; JSR = Journal of Service Research; MS = Marketing Science; OS = Organization Science.
the presence of another social entity” (van Doorn et al. 2017, p. 43) and the associated effects on service outcomes such as satisfaction, loyalty, engagement and well-being. Van Doorn et al. (2017) further take into account that the basic effect is moderated through customer attributes (relationship orientation, anthropomorphization, and technological readiness) and mediated by social cognition (warmth and competence) and psychological ownership (receptiveness, attractiveness, and manipulability). Second, the service robot acceptance model (SRAM) largely relies on the technology acceptance model (TAM, see section 2.3) in combination with role theory (see section 2.3) extending the TAM by applying it to service robots. Therefore, Wirtz et al. (2018) further include relational elements such as trust and rapport as well as social-emotional elements including perceived humanness, perceived social interactivity and perceived social presence. These factors affect customer acceptance of service robots and increase the actual use of service robots. Third, the concept of the degree of robotics adoption (DRA) examines antecedents and consequences related to “the extent of the robotics adoption that a firm employs to automate the tasks that were previously performed by human employees” (Xiao and Kumar 2019, p. 5). In their DRA framework, Xiao and Kumar (2019) specified robot characteristics (anthropomorphism, autonomy, relative advantages, and comparability) and customer characteristics (customer readiness, self-efficacy, role clarity, motivation, and demographics) as major antecedents of
2.2 Literature Review
41
the DRA. Further, they proposed that the DRA affects customer responses in terms of direct customer engagement (repeated purchases and firm profits) and indirect customer engagement (referral, influencing, and feedback). Further studies have a more narrow focus on how service robots affect customers. On the one hand, service robots are increasingly sophisticated and therefore becoming more useful to customers (Reed et al. 2012), bringing customers in better positions (Smith 2002) and revolutionizing customer experiences (Hoffmann and Novak 2018) by increasing social presence between customers and service robots (Bolton et al. 2018). However, depending on individual preferences (Bolton et al. 2018) and disadvantageous social comparisons (Reed et al. 2012), customers might also show negative responses toward service robots. Therefore, service firms should be aware of individual customer preferences (Bolton et al. 2018), that traditional human-centric conceptualizations of customer experiences may not be sufficient anymore (Hoffmann and Novak 2018), and that future robots may become part of the extended self (Belk 2013). Researchers have started to analyze the opportunities for process improvement in service firms resulting from the application of service robots. Although the automation of services with frontline service robots might not pay off in the short-term (Andreassen et al. 2018), service robots might elevate service effectiveness and efficiency by empowering frontline interactions (Marinova et al. 2017), as well as streamlining (Huang and Rust 2018) and optimizing (Fleming 2019) service processes. Service robots can learn from and across customers and therefore resolve the tension between efficiency and effectiveness of services (Marinova et al. 2017). They will be applied widely starting with common and repetitive tasks (Fleming 2019) proceeding with tasks requiring higher intelligence (Huang and Rust 2018), substituting or complementing human FLEs (Fleming 2019; Marinova et al. 2017) and require human FLEs to focus on their soft skills such as intuitive and empathic skills (Huang and Rust 2018). Regarding pricing strategies, service firms should decide selectively by compensating customers for robotic inconveniences on the one hand, but also charging customers for the convenience of automated services on the other hand (Andreassen et al. 2018). Considering the effects of robot characteristics and environmental conditions, this literature review identified robot characteristics (Rijsdijk et al. 2007), as well as service characteristics (Huang and Rust (2017) and their effect on customer responses to service robots on a conceptual basis. Rijsdijk et al. (2017) identified six factors driving the perception of the intelligence of a technology such as service robots (autonomy, ability to learn, reactivity, cooperation, humanlike interaction, and personality). This perceived intelligence is proposed to increase customer satisfaction mediated through complexity, compatibility and
42
2
Conceptual Background
relative advantage (Rijsdijk et al. 2017). Further, the technology-driven service strategy positioning map from Huang and Rust (2017) advises service firms in what scenarios to apply service robots. Characterizing services based on service attributes (standardized vs. personalized) and customer attributes (relational vs. transactional), the best scenario to start with is applying service robots to service encounters with a high level of standardization and for rather transactional customer interactions (Huang and Rust 2017). One further study in service marketing had a rather specific focus that does not match with the previous categories. Olazabal et al. (2001) focus on judicial court decisions regarding the behavior of service robots that support customers in searching products, showing that the law requires service robots to be fair and transparent to customers and are not allowed to hijack customers by confusing them in their shopping process. All of these conceptional studies proposed enlightening models with a profound basis in extant theories and literature. However, none of these conceptual frameworks has been tested and validated empirically.
2.2.1.3 Empirical Studies in Service Marketing Besides the conceptual studies proposing conceptual frameworks and models, service marketing research recently also conducted first empirical studies, including ˇ c et al. 2018; qualitative (Barrett et al. 2012; Beane and Orlikowski 2015; Cai´ Green et al. 2016) and quantitative work (Herrmann et al. 2015; Holzwarth et al. 2006; Jörling et al. 2019; Kim and McGill 2011; Kim et al. 2016; Mende et al. 2019; Touré-Tillery and McGill 2015). Qualitative work focuses on observations and descriptive analysis such as observational field studies (Barrett et al. ˇ c et al. 2018; 2012; Beane and Orlikowski 2015) and exploratory interviews (Cai´ Green et al. 2016), whereas quantitative work focuses on empirical studies such as experiments for the verification of causal relationships (Holzwarth et al. 2006; Jörling et al. 2019; Kim and McGill 2011; Kim et al. 2016; Mende et al. 2019; Touré-Tillery and McGill 2015). In addition to the few conceptual studies (see section 2.2.1.2), this literature review identified even fewer empirical studies (11 out of 27) that were conducted in top service marketing publications. Table 2.4 gives an overview about the individual studies and the examined effects with the corresponding data basis. As empirical research in service marketing on the topic is scarce, this thesis applied broad selection criteria including studies with direct and indirect relation to the application of robots at the service encounter. Therefore, the conception of service robots is inhomogeneous among the studies as shown in the
2.2 Literature Review
43
second column of Table 2.4. For the analysis of the scattered research in service marketing, this thesis relies on the narrative summary approach (Paré et al., 2015). Among the diverse identified empirical studies in service marketing research, there are significant differences in terms of the conception of service robots, the field of application and the perspective on the application of service robots. Overall, the identified studies provide a broad overview about potential opportunities and risks regarding service robots at the customer encounter, by applying various types of interactions with participants in the different experimental studies. ˇ c The conception of robots varies from physical robots (Barrett et al. 2012; Cai´ et al. 2018; Mende et al. 2019) over telepresence technologies (Beane and Orlikowski 2015; Green et al. 2016) and anthropomorphized online agents (Holzwarth et al. 2006; Kim et al. 2016; Touré-Tillery and McGill 2015) to digital softwarebots (Herrmann et al. 2015) and physical entities (Jörling et al. 2019; Kim and McGill 2011) that are not service robots in terms of the underlying definition of this thesis. These differences occur due to the different definitions of service robots (see section 2.1.1), varying between narrow definitions including a ˇ c mechanical component (Barrett et al. 2012; Beane and Orlikowski 2015; Cai´ et al. 2018; Mende et al. 2019) and a broader definition of service robots without requiring a mechanical component. These non-mechanical robots focus more on intelligence, therefore also including software systems (Herrmann et al. 2015) and avatar-like online representations (Holzwarth et al. 2006; Kim et al. 2016; Touré-Tillery and McGill 2015). Service robots were examined across many fields of application in service marketing, ranging from retail (Mende et al. 2019) and online shopping (Holzwarth et al. 2006; Touré-Tillery and McGill 2015), over health (Barrett et al. 2012; Beane and Orlikowski 2015; Mende et al. 2019), e-health (Green et al. 2006) and ˇ c et al. 2018) to entertainment (Kim and McGill 2011; Kim et al. elderly care (Cai´ 2016) and the application in private environments (Jörling et al. 2019). Further, researchers applied service robots in hospitality and restaurant settings (Mende et al. 2019) as well as supporters in financial decisions in auctions (Herrmann et al. 2015). Extant empirical research provides three major perspectives on the application of service robots. While most studies focus on customer-robot interactions ˇ c et al. 2018; Green et al. 2016; Holzwarth et al. 2006; Jörling et al. 2019; (Cai´ Mende et al. 2019; Touré-Tillery and McGill 2015) and consumer responses (Kim and McGill 2011; Kim et al. 2016), a few studies address the performance of robots (Herrmann et al. 2015) and the effects on service firms and their employees (Barrett et al. 2012; Beane and Orlikowski 2015; Green et al. 2006).
44
2
Conceptual Background
Table 2.4 Overview on Empirical Studies in Service Marketing Author/s (Year), Journal
Conception of Robots
Data Basis
Examined Effects
Barrett et al. (2012), OS
Physical robot located in a professional service setting, dispensing pharmaceuticals
Qualitative observational field study with N = 20 observation days N = 41 interviews
Analyzing implications from the application of a service robot dispensing pharmaceuticals in a hospital pharmacy on organizational practice: • On the one hand, the robot takes off workload from the employees, giving them more time for specific customer-centered work and facilitating team collaboration. • On the other hand, the introduction of the service robot disrupted employees’ routine requiring them to take care of the robot and feeling a loss of autonomy and control.
Beane and Orlikowski (2015), OS
Physical, remote telepresence robot RP-7 to assist employees
Qualitative observational field study with N = 34 observation days and a total of 424 ‘customers’ N = 80 interviews
Effects of robotic telepresence on the coordination among employees: • The application of service robots allows better connections among teams and intensifies coordination. • When employees have different understandings of the work and colleague participation, the application of telepresence robots decreases coordination effectiveness.
ˇ c, Cai´ Odekerken-Schröder, and Mahr (2018), JSM
Service robots as socially assistive robots for elderly care
Qualitative interview study with N = 20 elderly people
Identification of six potential roles (co-creation /co-destruction) for service robots in elderly care networks along three dimensions: • Robot’s cognitive support dimension: extended self / deactivator • Robot’s safeguarding dimension: enabler / intruder • Robot’s social support dimension: ally / replacement (continued)
2.2 Literature Review
45
Table 2.4 (continued) Author/s (Year), Journal
Conception of Robots
Data Basis
Examined Effects
Green, Hartley, and Gillespie (2016), JSR
Technology-infused service delivery by human service providers
Exploratory qualitative interviews with N = 33 technology-infused service providers
Revealing potential negative consequences of technology-infused service delivery: • Depersonalization of the service • Potential threat of lacking customer privacy • Eeriness or unfamiliarity in the interaction
Herrmann, Kundisch, and Rahman (2015), MS
Service robots as digital, IT-enabled automated bidding robots
Empirical analysis of data from N = 7,000 pay-per-bid auctions
Digital IT-enabled bidding robots reduce customers’ likelihood to fall for the sunk cost effect by more than 50%. Robotic agents help customers to act more rational due to lower behavioral investments when delegating decisions to the bidding robot.
Holzwarth, Digital service Janiszewski, and representative Neumann (2006), JM interacting with online shopping customers
Empirical online experiments with N = 400 and N = 596 participants as customers
• Digital service representatives can help to make online shopping less impersonal and lead to more customer satisfaction and a greater purchase intention. • For moderate levels of product involvement, the attractiveness of the digital agent increases its effectiveness. • For high levels of product involvement, digital agents perceived as experts are more effective. (continued)
46
2
Conceptual Background
Table 2.4 (continued) Author/s (Year), Journal
Conception of Robots
Data Basis
Examined Effects
Jörling, Böhm, and Paluch (2019), JSR
Service robots with a physical embodiment
Exploratory, Investigation on attribution of qualitative responsibility for service interviews with N outcomes: = 10 and N = 18 • The autonomy of a service robot participants decreases perceived behavioral Quantitative control and perceived scenario-based responsibility for positive online experiments outcomes. • Perceived ownership of the with N = 321 | 236 service robot increases | 223 participants responsibility for negative outcomes. The potential to interrupt the service robot’s autonomy increases perceived control and responsibility for positive outcomes.
Kim and McGill (2011), JCR
Customers interacting with anthropomorphized entities
Empirical analyses of three experimental studies with N = 61 | 84 | 79 participants
Anthropomorphism affects customers perceived risk and this effect is moderated by perceptions of social power: • Customers with low power perceive higher risk for high levels of anthropomorphism. • Customers with high power perceive higher risk for low levels of anthropomorphism. • Customers with high (low) power perceive a greater (lesser) degree of control over the anthropomorphized entity. (continued)
2.2 Literature Review
47
Table 2.4 (continued) Author/s (Year), Journal
Conception of Robots
Data Basis
Examined Effects
Kim, Chen, and Zhang (2016), JCR
Service robots as digital assistants
Empirical analyses of six computer-based experimental studies with a total of N = 784 participants
Assessing the effects of anthropomorphic representations of computerized helpers on enjoyment: • The presence of anthropomorphized helpers can reduce consumers’ perceived autonomy. • Consumers feel less joy when receiving assistance from an anthropomorphized helper than from a mindless entity. • Moderator: consumer autonomy importance. • Although this study focuses on game enjoyment, it can be extended to other outcomes.
Mende et al. (2019), JMR
Humanoid service robots with a human-like morphology
Empirical analyses of seven experimental studies with a total of N = 1,281 participants
Research on the influence of service robots on service experiences and compensatory consumer responses: • Consumers show stronger compensatory responses when interacting with a service robot rather than with a human employee. • Service robots elicit greater consumer discomfort leading to compensatory responses. • Boundary conditions: social belongingness and anthropomorphization of the robot.
Touré-Tillery and McGill (2015), JM
Anthropomorphized Empirical analyses messengers of three experimental studies with N = 57 | 79 | 186 participants
Investigation on persuasiveness of anthropomorphized messengers compared with human spokespeople: • Customers with low (high) interpersonal trust are more persuaded by anthropomorphized messengers (human spokespeople). • Under conditions with low attentiveness, all customers are unaffected by the nature of the agent.
48
2
Conceptual Background
Through their heterogeneity, studies cover a broad field of opportunities and risks regarding the application of service robots at the service encounter. On the one hand, robots were identified to provide a wide range of opportunities, such as reducing the workload for employees, giving them more time for specific customer-centered work (Barrett et al. 2012). Regarding service employees, service robots further facilitate team collaboration (Barrett et al. 2012), intensify coordination and allow better connections between teams (Beane and Orlikowski 2015). Customers experience more satisfaction and a higher repurchase intention due to the application of service robots (Holzwarth et al. 2006) and are more persuaded by anthropomorphic design (Touré-Tillery and Mc Gill 2015). Further, the design of service robots has implications on effectiveness (Holzwarth et al. 2015), perceived risk (Kim and McGill 2011), perceived responsibility for outcomes (Jörling et al. 2019) and perceived customer autonomy (Kim et al. 2016). Digital service robots can help to make online shopping less impersonal (Holzwarth et al. 2006) and help customers to act more rational in their decision process (Herrmann et al. 2015) and in an assistive role service robots might help supporting as an ˇ c et al. 2018). ally, safeguarding as enabler and supporting as extended self (Cai´ On the other hand, service robots can disrupt the routine of service employees, giving them a feeling of loss in autonomy and control (Barrett et al. 2012) and decreasing coordination effectiveness when employees have different understandings of the work and colleague participation (Beane and Orlikowski 2015). Regarding service employees, service robots also might decrease the perceived level of autonomy and decrease consumers joy (Kim et al. 2016). Compared to human-human interactions, service robots elicit greater customer discomfort resulting in compensatory responses (Mende et al. 2019) and might lead to feelings of eeriness or unfamiliarity in the interaction (Green et al. 2016). Further concerns address the depersonalization of the service and a potential lack of customer privacy (Green et al. 2016). In an assistive role, service robots comprise the ˇ c et al. 2018). risk of co-destruction in the role of a deactivator or intruder (Cai´ However, the type of interaction varied broadly across the studies. Some empirical studies in service marketing relied on pictures (Jörling et al. 2019; Kim and ˇ c et al. 2018; Jörling et al. 2019) of robots McGill 2011) or photographs (Cai´ ˇ c that were shown to participants to elicit responses in qualitative interviews (Cai´ et al. 2018; Jörling et al. 2019) or online-based experiments (Jörling et al. 2019; Kim and McGill 2011). Other studies applied a two-dimensional representation of a service robot on a screen as kind of avatar without a physical representation (Holzwarth et al. 2006; Kim et al. 2016; Touré-Tillery and McGill 2015), conducting sets of online-based studies with various experiments. Barrett et al. (2012), as well as Beane and Orlikowski (2015) relied on observational studies with a
2.2 Literature Review
49
physical representation in a live-interaction, although the interaction between the customer and the robot itself was rather short and not in the focus of the interaction. Only recently, the first study relied on physical robots in a service setting to examine the effects of service robots on customers at the service encounter (Mende et al. 2019). However, even in the studies from Mende et al. (2019), participants in the experimental studies just were shown video recordings of service robots on a screen and the authors did not rely on a live face-to-face interaction between the participants and the service robots.
2.2.1.4 Limitations in Service Marketing Research Research in service marketing on robots at the service encounter is still scarce. Across 14 top journals in the field, this thesis could just identify 27 studies over the last two decades concerning service robots directly or indirectly. Most of the work was conceptional and rather fragmented, based on different definitions of service robots. Within the few 11 empirical studies on service robots, there was only one (Mende et al. 2019) examining customer responses at the service encounter with physical service robots, rather than pictures of service robots or virtual two-dimensional representations on a screen. However, there is no research on customer expectations and customer responses on frontline service robots at the service encounter relying on live face-to-face interactions between customers and service robots in service marketing research. These insights from literature are not comprehensive enough as basis for the examination of customer expectations and customer responses on service robots at the service encounter. Therefore, the following section takes a closer look on what has been examined in the field of human-robot interaction literature regarding service robots in service settings and how this affects customers.
2.2.2
Studies in the Field of Human-Robot Interaction Literature
In contrast to service marketing literature, human-robot interaction literature has largely studied interactions between humans and robots, especially focusing on physical robots and real face-to-face interactions. Many different types of robots have been applied in various fields, interacting with patients (Lee et al. 2017), elderly (Hudson et al. 2017; Jayawardena et al. 2010; Piezzo and Suzuki 2017) and children (Looije et al. 2016; Pulido et al. 2017) in the field of healthcare, in
50
2
Conceptual Background
households (Ferrús and Somonte 2016; Vaussard et al. 2014) and in the field of education (Conti et al. 2017; Fernández-Llamas et al. 2017; Kanda et al. 2007).
2.2.2.1 Criteria and Focus of the Literature Review As this thesis focuses on professional service settings with humanoid robots as service representatives and corresponding customer expectations and responses, the review of HRI literature applies a strict service focus. Further, this review just relies on most recent work (published in the year 2000 or more recently), published in HRI journals (excluding conference publications), and focusing on studies including customer interactions with physical service robots, excluding studies without such physical interactions (e.g., Reich and Eyssel 2013).
2.2.2.2 Experimental Studies in Human-Robot Interaction Literature Table 2.5 shows the most relevant studies in human-robot interaction literature for this thesis. It provides information about the applied type of robot such as Nao (Pan et al. 2015) or Sacarino (Pinillos et al. 2016; Rodriguez-Lizundia et al. 2015) and the examined effects. Further, it contains information on the underlying data basis such as lab studies (Trovato et al. 2017) or field studies (Shiomi et al. 2013; Rodriguez-Lizundia et al. 2015), short-term (Trovato et al. 2017) or long-term studies (Sabelli and Kanda 2016; Kirby et al. 2010) and experimental (Yamazaki et al. 2010; Pan et al. 2015) or observational studies (Kanda et al. 2010; Pinillos et al. 2010). HRI studies in service settings involve physically embodied robots with different appearances and varying degrees of human-likeness, including the robots Valerie (Kirby et al. 2010), KOBIAN (Trovato et al. 2017), different versions of Robovie (Kanda et al. 2010; Sabelli and Kanda 2016; Shiomi et al. 2013; Yamazaki et al. 2010), Sacarino (Pinillos et al. 2016; Rodriguez-Lizundia et al. 2015) and NAO (Pan et al. 2015). According to the classification of robots in section 2.1.1, these robots are rated as professional service robots regarding the purpose of application and as humanoid robots in term of their outer appearance. The service robot Valerie was built on basis of the physical RWI B21r body with a virtual head displayed on a screen (Kirby et al. 2010). It is applied in the first long-term interaction field studies, conducted on a university campus (Gockley et al. 2005) and a mobile conference (Sabanovic et al. 2006). Most customers responded with reluctance and only a few customers started to interact with it after a certain period (Gockley 2005). Subsequently, Kirby et al. (2010) taught Valerie to display different mood expressions that could be identified by customers and shown to influence customer interaction.
2.2 Literature Review
51
KOBIAN was also applied in the role of a receptionist to analyze customer reactions based on the physical appearance and the sound of voice of service robots by Trovato et al. (2017). It was compared to a far more anthropomorphized virtual agent that was preferred by the customers. In terms of the service robots’ voice, there was a similar effect showing that customers preferred the service robot with a more human-like voice rather than a mechanical voice (Trovato et al. 2017). In the identified studies Robovie was the most popular robot as many studies relied on different versions of this robot type in shopping malls (Kanda et al. 2010, Sabelli and Kanda 2016, Shiomi et al. 2013), as well as in museum and Table 2.5 Overview on Experimental Studies in Human-Robot Interaction Literature Author/s (Year), Journal
Applied Robot
Data Basis
Examined Effects
Kanda et al. (2010), IEEE T ROBOT
Humanoid physical service robot Robovie-IIF in a shopping mall
Observational • Customers’ acceptability field study over was promising and five weeks with customers responded N = 2642 positively to the service interactions in robot in the shopping mall. a retail setting • Compared with an information display, the service robot was considered as more useful and elicited more shopping. • Comparing repeat and single-time visitors, the repeating visitors had better impressions and perceived higher familiarity with the service robot.
Kirby, Forlizzi, and Simmons (2010), ROBOT AUTON SYST
Humanoid roboreceptionist Valerie with a physical RWI B21r body and a virtual head
Long-term field • Service robots are able to experiment display different mood over nine expressions with different weeks with N levels of intensity and = 2679 customers can identify it. • Customers’ interaction with interactions a service robot is based on its mood and tends to change when it displays different moods. (continued)
52
2
Conceptual Background
Table 2.5 (continued) Author/s (Year), Journal
Applied Robot
Data Basis
Examined Effects
Pan et al. (2015), Humanoid INT J SOC ROBOT physical service robot Nao in a hotel setting
Field experiment with N = 3*50 customers in a hotel setting
Examining the effects of indirect and direct speech from the service robot on customers: • Compared to traditional information broadcast, indirect speech by robots triggers a human-like social interaction that is easier for customers to join in. • Direct verbal robotic interaction (facing and greeting, asking if customer needs help) works even better than just reciting information.
Pinillos et al. (2016), ROBOT AUTON SYST
Observational • Customers most demanded field study on topics during the interaction 23 days with N with the service robots were = 349 information about hotel interactions facilities, news service and with hotel accompanying guests to customers different areas of the hotel. • Customers were reluctant talking loudly to the robot in the presence of other guests. Therefore, the communication via touchscreen was the most used way. • The robot was continuously improved through successive refinements based on the analysis of acquired interaction information. (continued)
Humanoid physical service robot Sacarino in a hotel environment
2.2 Literature Review
53
Table 2.5 (continued) Author/s (Year), Journal
Applied Robot
Data Basis
Examined Effects
Rodriguez-Lizundia Humanoid et al. (2015), INT J physical bellboy HUM-COMPUT ST service robot Sacarino in a hotel environment
Field experiments with N = 95 valid interactions in a hotel setting
Analyzing the effects of service robot behavior on customer engagement and comfort: • Embodiment of service robots engages customers in maintaining longer interactions. • Active-looking service robots are more attractive for customers than passive-looking service robots.
Sabelli and Kanda (2016), INT J SOC ROBOT
Qualitative • The service robot triggered long-term study positive overall customer over three years responses indicating based on acceptance. No one observations expressed outright rejection and N = 67 of the robot. • The service robot was able to qualitative attract customer attention by interviews with its appearance (shape and customers gestures) • Customers perceived the service robot as autonomous even after learning that it was operated. • The vast majority of customers identified a connection between the robot and the shopping mall.
Humanoid physical service robot Robovie in a shopping mall
Shiomi et al. (2013), Two humanoid INT J SOC ROBOT physical service robots (Robovie-II and robovie-mini r2) in a shopping mall
Two field experiments with N = 5000 customers passing by
Examining the effect of different-sized service robots in the context of a shopping mall: • Attracting customers in advertising tasks: Service robots increase the number of attracted customers and the smaller robot outperformed the regular-sized one. (continued)
54
2
Conceptual Background
Table 2.5 (continued) Author/s (Year), Journal
Applied Robot
Data Basis
Examined Effects
Trovato et al. (2017), J BEHAV ROBOT
Humanoid physical service robot KOBIAN as receptionist robot
Experimental lab study with N = 60 participants
Analysis of customer reactions on the physical appearance and the sound of the voice of a service receptionist robot: • Customers preferred the virtual agent rather than the mechanical robot, as the virtual agent is far more anthropomorphized in its design. • Customers preferred the service robot with a human voice rather than a mechanical one.
Yamazaki et al. (2010), J PRAGMAT
Humanoid physical robot Robovie ver.2 as guide in a museum
Three field • The communication between experiments service robots and customers with may be enhanced by N = 16 | 12 | 46 programming robots with a in a museum coordination of verbal and and exhibition non-verbal actions (e.g. head setting turn, gaze, pointing). • The coordination of robotic verbal and non-verbal actions is an important means for displaying recipiency toward customers.
Notes: 1) IEEE T ROBOT = IEEE Transaction on Robotics; 2) ROBOT AUTON SYST = Robotics and Autonomous Systems; 3) INT J SOC ROBOT = International Journal of Social Robotics; 4) INT J HUM-COMP ST = International Journal of Human-Computer Studies; 5) J BEHAV ROBOT = Journal of Behavioral Robotics; 6) J PRAGMAT = Journal of Pragmatics
exhibition settings (Yamazaki et al. 2010). Long-term observational studies in shopping malls showed that customers responded positively (Kanda et al. 2010) and that the overall reactions indicated acceptance (Sabelli and Kanda 2016). Service robots such as Robovie are able to attract customer attention due to its shape and gestures (Sabelli and Kanda 2016) and are able to elicit more shopping (Kanda et al. 2010). Shiomi et al. (2013) further showed that the size of the service
2.2 Literature Review
55
robot had an effect on the interaction with customers in a way, that the Roboviemini r2 outperformed the regular-sized Robovie. A field experiment with Robovie conducted by Yamazaki et al. (2010) revealed a significant effect of coordination between verbal and non-verbal robotic actions to enhance communication and display recipiency toward customers. The interactive bellhop robot Sacarino was applied in a hotel setting in two studies (Pinillos et al. 2016; Rodriguez-Lizundia et al. 2015), walking alongside hotel guests and providing information about hotel facilities and news service (Pinillos et al. 2016). On the one hand, it revealed customer reluctance talking loudly to the robot in the presence of other guests (Pinillos et al. 2016), but on the other hand, embodiment of service robots and active-looking behavior could be shown to engage customers (Rodriguez-Lizundia et al. 2015). Installing Softbank’s NAO robot in another field experiment in a hotel setting provided further insights into customer responses to different types of robot speech (Pan et al. 2015); direct speech gained customers’ attention and offered potential for replacing traditional information channels.
2.2.2.3 Limitations in the Field of Human-Robot Interaction Literature These studies in human-robot interaction literature provide valuable insights on customer responses to service robots in the field, as all presented studies involved face-to-face interactions with physical robots that are far more realistic that comparable studies in service marketing research relying on pictures, photographs and video recordings. However, many of these studies derived implications on observations of customer-robot interactions. In doing so, the studies are essentially non-random, as customers’ self-selection already depends on their characteristics (Austin 2011). For example, Kanda et al. (2010) derived the implication that interacting with the service robot repeatedly leads to a better impression and higher familiarity with the service robot, based on customer decisions when and how to interact with the robot. It might be the case that customers with a less positive attitude regarding the robot would most probably decide not to interact with the robot again, leading to problems with causality of the observations. Although that was not the main focus of their study, further experimental studies with different manipulation groups are necessary to identify causal relationships, as observational studies are not suitable to test hypotheses (Austin 2011). Further, human-robot interaction literature widely examined human reactions toward robots but did not specifically compare human frontline employees with
56
2
Conceptual Background
frontline service robots. HRI literature is not focused on service-related outcomes and customer responses in commercial service settings. Instead, HRI studies imply a rather technical focus. For example, Kanda et al. (2010) included a strong focus on the development of a robotic system with radio-frequency detection and Pinillos et al. (2016) focused on the hardware configuration, software architecture, the navigating subsystem, motion metrics and the gyroscope, rather than service outcomes.
2.3
Theoretical Background
This section lays the theoretical foundation of the thesis by introducing the most relevant theories to support the postulated coherences in each of the comprising studies. The analysis of the current state of research (see section 2.2) clearly shows that a wide range of theories from across various research fields are applied to explain customer expectations and responses towards service robots. The most relevant referenced theories are widely applied in the fields of IS and HRI (section 2.3.1), as well as in the field of psychology and social psychology (section 2.3.2). Technology acceptance model, expectation disconfirmation paradigm, as well as uncanny valley paradigm and the computers-are-social-actors paradigm are widely applied in the field of IS and HRI and are specifically tailored to explain effects during the interaction with a technology such as professional service robots. Apart from that, cognitive dissonance theory, script theory, as well as role theory and cultural dimensions theory are borrowed from the field of psychology and social psychology and applied in this thesis to gain further insights in customer expectations and responses towards service robots. Figure 2.3 provides an overview about the applied theories that will be introduced and discussed in terms of their relevance for this thesis in the following sections.
2.3.1
Theories from Information Systems and HRI Research
In a first step, the two theories that are mainly applied to predict user reactions on general technical systems are introduced: technology acceptance model (section 2.3.1.1) and expectation disconfirmation theory (section 2.3.1.2). The next step focuses on two rather specific paradigms, that are applied especially regarding anthropomorphized technical devices, as the computers-are social actors
2.3 Theoretical Background
57
Main Area of Application
Information Systems & Human-Robot Interaction
General Technical Systems
Anthropomorphized Technical Devices
Technology Acceptance Model (TAM)
Uncanny Valley Paradigm
Psychology & Social Psychology
Psychology
Role Theory
Script Theory Expectation Disconfirmation Theory
Computers/ RobotsAre-Social-Actors Paradigm (CASA/ RASA)
Social Psychology
Cognitive Dissonance Theory
Cultural Dimensions Theory
Figure 2.3 Overview of Applied Theories
paradigm (section 2.3.1.4) and regarding humanoid robots as the uncanny valley paradigm (section 2.3.1.3).
2.3.1.1 Technology Acceptance Model The technology acceptance model (TAM) is widely applied in the field of information systems to examine technology use and acceptance (King and He 2006). Davis (1989) first introduced the TAM to explain the effects of a technology on corresponding user behavior, relying on the theory of reasoned action (Fishbein and Ajzen 1975). While the theory of reasoned action describes behavior based on intentions focusing on human behavior in general (Fishbein and Ajzen 1975), the TAM is strictly referring to technology acceptance (Davis et al. 1989). It is a model to explain the reasons why people use or avoid certain technologies (Davis et al. 1989). According to TAM, the attitude toward using a technology is mainly based on perceived usefulness and perceived ease of use (Davis et al. 1989). Perceived ease of use is the “extent to which a person believes that using a particular system will be free of effort” (Sun et al. 2009, p. 52) and perceived usefulness is the “extent to
58
2
Conceptual Background
which a person believes that using a particular system will enhance his or her […] performance” (ibid., p. 52). The behavioral intention to use a technology depends on perceived usefulness and the attitude toward using as shown in Figure 2.4. One of the most important findings is the relatively strong effect of usefulness on usage compared to the effect of ease of use. In many studies, the effect of usefulness on usage was significantly stronger than the effect of ease of use (Davis 1989). This can be explained as users’ technology adoption is mainly based on functionality of the technology and not primarily on how easy it is to use the system to get the desired functionality (Davis 1989). Therefore, users are ready to accept some shortcomings in terms of ease of use, if the technology provides useful functionalities, as ease of use cannot compensate critically needed useful functions (Davis 1989).
Perceived Usefulness
External Variables
Attitude Toward Using
Behavioral Intention to Use
Actual System Use
Perceived Ease of Use
Figure 2.4 Framework of the Technology Acceptance Model. (see Davis et al. 1989)
Venkatesh and Davis (2000) expanded the TAM by including further independent variables regarding social influence (subjective norm, voluntariness, and image) and the cognitive instrumental processes (job relevance, output quality, and result demonstrability) with experience as moderating effect and designed the TAM2. As this extension of the TAM has a strong intra-organizational focus, applied and validated on employees within organizations (Venkatesh and Davis 2000), this extension does not contribute much to the application on customer responses on service robots. Some studies came up with further TAM developments, such as the unified theory of acceptance and use of technology (UTAUT) including a multitude of independent variables leading to increased unnecessary complexity (Bagozzi 2007). The TAM contributes properly to this thesis as it explains technology acceptance and technology use due to users’ perception of a technology. Applied to service robots, this model helps to predict customer responses according to their
2.3 Theoretical Background
59
perception of the service robot. TAM has been criticized for its limited explanatory power, lack of practical implications (Chuttur 2009), for its triviality with just two presumed predicting factors and for its lack of novelty (Bagozzi 2007; Benbasat and Barki 2007). Nevertheless, it provides a general approach to assess customer reactions on service robots. Further, the TAM was mainly focused on computer systems validated with computer programs (Davis et al. 1989). Study 1 extends the TAM and adapts it to the application on service robots, conceptualizing a service robot acceptance model based on the TAM and empirically evaluating it in an experimental study.
Table 2.6 Summary of Technology Acceptance Model Name
Technology Acceptance Model (TAM)
Field of Application
Information Systems
Central Sources
Davis 1989; Davis et al. 1989; Venkatesh and Davis 2000
Basic Assumptions
1. People’s technology use can be predicted from their intentions. 2. Usefulness and ease of use mediate the influence of external variables on actual technology use.
Core Message
People’s technology use can be predicted from their intentions: Perceived usefulness and ease of use are major determinants of people’s intention to use a technology.
Main Contribution to Thesis 1. Predicts customer responses / acceptance of service robots based on the functional perception (usefulness, ease of use) 2. Serves as basis for the development of an extended Service Robot Acceptance Model Criticism
– Limited explanatory power due to its triviality with just two presumed predictors of user behavior neglecting further factors. – Lack of novelty and lack of practical implications
2.3.1.2 Expectation Disconfirmation Theory Although expectation disconfirmation theory (EDT) is rooted in psychology and marketing research literature (Oliver 1977; Oliver 1980), it was soon adopted (Ginzberg 1981) and widely applied (Venkatesh and Goyal 2010) in the field of information systems.
60
2
Conceptual Background
EDT and attempts to explain how customers respond to a product or a service provided by a firm, according to the interplay of the customer’s prior expectations at t1 and actual perceptions at t2 (Oliver and DeSarbo 1988; Tse and Wilton 1988; Yi 1990). Expectations (see section 2.1.3) refer to the anticipated performance (Churchill and Suprenant 1982), so the “comparison of expectations and perceptions will result in either confirmation or disconfirmation” (Bloemer and Odekerken-Schroder 2002, p. 70). Confirmation occurs when the customer’s service expectations are met; disconfirmation implies a discrepancy between the customer’s expectations and the perceived performance. Disconfirmation can take two forms (Churchill and Suprenant 1982): The perceived performance is higher than expected (positive disconfirmation), or the expectations exceed the perceived performance (negative disconfirmation). Confirmation is associated with positive customer responses, such as customer satisfaction that leads in turn to positive outcomes for companies, such as increased repurchase intention (Bhattacherjee 2001a), as shown in Figure 2.5.
Expectation (t1)
Confirmation (t2)
Satisfaction (t2)
Repurchase Intention (t2)
Perceived Performance (t2)
Figure 2.5 Framework of the Expectation Disconfirmation Theory. (based on Oliver 1977)
IS has already widely relied on EDT (see Venkatesh and Goyal 2010). For example, McKinney et al. (2002) developed an approach to measure web consumer satisfaction; Bhattacherjee (2001b) examined users’ motives to continue using a system; and Bhattacherjee and Premkumar (2004) explained how beliefs and attitudes about the use of information technologies change over time. Various outcome variables derive from the expectation disconfirmation process, such as technology adoption (Brown et al. 2012; Venkatesh and Goyal 2010), online shopping behavior (Hsu et al. 2006), and web customer satisfaction (McKinney at al.
2.3 Theoretical Background
61
2002). In turn, study 3 applies EDT to explore the relationship between customer expectations and perceptions of the robot on customer responses to service robots at the service encounter. The theory predicts that consumers form different standards to which they compare a current performance of a company. This in turn will result in a confirmation (i.e., met comparison standard) or disconfirmation (i.e., a discrepancy between the comparison standard and the individual’s perceptions). Important implications of the confirmation disconfirmation theory relate to the comparison standard. Based on this paradigm, study 1 predicts that users are likely to compare their perceived performance with three comparison standards, i.e., an ideal performance of the service robot, their expectations toward a service robot’s performance, and prior experiences during similar situations, i.e., service encounters. The outcome of the comparison process is likely to affect the user’s acceptance of the service robot and is evaluated with expectation disconfirmation theory. Analyses of the EDT have been widely conducted adopting traditional difference scores as a direct measure of disconfirmation that suffers from weak explanatory potential as it cannot assess effects according to varying degrees of discrepancy and does not account for nonlinear effects (Shanock et al. 2010). As Brown et al. (2014, p. 749) assert that „the influence of confirmation on the outcomes is dependent on the absolute levels of confirmation”, study 3 examines the EDT with polynomial models and surface response analysis (see section 3.2.3). Further, there is no research on the assumption that customers just rely on their expectations for the service evaluation. There might be other relevant comparison levels as for example a minimal tolerable level whose achievement still might lead to satisfaction for some customers (Yüksel and Yüksel 2001) (Table 2.7).
2.3.1.3 Uncanny Valley Paradigm The uncanny valley paradigm (Mori 1970) presumes a relationship between the appearance of a robot and corresponding human responses. It suggests that a robot’s degree of human-like appearance relates positively to a shinwakan feeling with it (Mori et al. 2012). Mori (1970) related the degree of a robot’s humanlikeness to the Japanese term shinwakan that might be translated as affinity, familiarity or likeability (Ho and MacDorman 2010). Therefore, extant literature refers to a variety of terms for the vertical axis in the uncanny valley chart, such as pleasantness (Looser and Wheatley 2010; Seyama and Nagayama 2007), familiarity (Cheetham et al. 2014; Thompson et al. 2011), and likeability (Carter at al. 2013; Yamada et al. 2013). However, all of them are referring to a positive human reaction indicating a comfort level that this thesis will further cite as premise of acceptance (see study 3).
62
2
Conceptual Background
Table 2.7 Summary of Expectation Disconfirmation Theory Name
Expectation Disconfirmation Theory (EDT)
Field of Application
Widely adopted in information systems literature, although its origins in psychology and marketing research.
Central Sources
Oliver 1977; Oliver 1980; Venkatesh and Goyal 2010
Basic Assumptions
1. Actual performance itself cannot explain user satisfaction. 2. compare perceived performance with their prior expectations. 3. User confirmation is one of the key antecedents of user satisfaction.
Core Message
Confirmation between user expectations and perceived performance leads to positive user responses.
Main Contribution to Thesis 1. Provides a theoretical basis for the comparison between customer expectations and customer perceptions of service robots. 2. Predicts positive customer responses for frontline service robots that meet prior customer expectations. Criticism
– Analyses of the EDT are widely conducted with traditional difference scores as measure of disconfirmation. It has weak explanatory potential as it omits varying degrees of discrepancy and does not account for non-linear effects. – The unilateral focus on prior expectations as benchmark is not all-encompassing.
The horizontal axis refers to the degree of human-likeness of the robot (see section 2.1.1). Although many researchers criticized that human-likeness is difficult to operationalize consistently (Bartneck et al. 2009; Kätsyri et al. 2015), Mori (1970) already provided some examples of different degrees of human-likeness such as industrial robots, humanoid robots and healthy humans (see Figure 2.6). A detailed description and comparison of the different degrees of human-likeness can be found in section 2.1.1.2. However, there is a drop in this positive relationship, as there is an increased sensitivity for defects, as the robots almost resemble humans as shown in Figure 7.1 (Mori et al. 2012). Overly human-like robots might resemble corpses, zombies, prosthetic hands or bunraku puppets (Mori et al. 2012). Mori (1970) described the decline of this positive relationship for robots with an overly humanlike appearance as uncanny valley. Potential premises for this negative effect
2.3 Theoretical Background
63
might be based on pathogen avoidance (Moosa and Ud-Dean 2010), mortality salience (MacDorman 2005) and evolutionary aesthetics (MacDorman et al. 2009). According to pathogen avoidance, overly human-likeness increases sensitivity for defects, as they indicate diseases and leads to an avoidance of defective—potentially infective—individuals (MacDorman et al. 2009; Moosa and Ud-Dean 2010). Further, the interaction with overly human-like robots might remind users to the inevitability of death and lead to a feeling of eeriness (MacDorman 2005). From an evolutionary point of view, physical attractiveness regarding symmetry and averageness is considered to be appealing (Thornhill and Gangestad 1993) and android robot’s aesthetics quality is just not yet high enough (Hanson 2006; MacDorman et al. 2009). Figure 2.6 depicts the curve of this negative effect as uncanny valley.
shinwakan (comfort level)
uncanny valley
still moving
+
healthy person humanoid robot (e.g. Pepper) stuffed animal
industrial robot
human likeness
50 %
100 % corpse
prosthetic hand
_ zombie
Figure 2.6 Representation of the Uncanny Valley Paradigm. (see Ho and MacDorman 2010)
The uncanny valley paradigm provides insights to more deeply understand customers’ responses to the two different service representatives (service robot versus FLE), as the degree of human-likeness is an important robot perception dimension (Belk 2016; Broadbent 2017; Mori et al. 2012). Specifically, that means that customers in the totally human-like FLE interaction experience higher values
64
2
Conceptual Background
of familiarity, whereas customers interacting with the less human-like service robot experience lower values of familiarity with the service representative. This thesis relied on Softbank’s Nao robot (study 1) and Pepper robot (study 3 and study 4), that are both humanoid robots and are located clearly on the left side of the uncanny valley (Mori 1970). Further, the following studies compare FLEs as healthy humans with humanoid robots, assuming that according to uncanny valley paradigm, human service representatives lead to higher levels of customer comfort that robotic ones. This effect is even enhanced, as the following studies rely on moving service representatives (dashed line in Figure 2.6). The uncanny valley paradigm has been criticized as too simple to explain the heterogeneous phenomena with just two factors (Bartneck et al. 2009; MacDorman et al. 2009), a general lack of empirical evidence for the design of the graph (Wang et al. 2015) and for the denial of good design being able to avoid the uncanny (Hanson et al. 2005). Although Mori (1970) proposed rising familiarity with robots that cannot be distinguished from real humans, more recent research doubts that curve and assumes an uncanny cliff instead, where overly human-like robots are not able to reach rising levels of familiarity (Bartneck et al. 2007) (Table 2.8). However, the approach of this thesis to apply the uncanny valley explaining different reactions to FSRs as compared to human FLEs is hardly affected by the valid criticism on the paradigm.
2.3.1.4 Computers-Are-Social-Actors Paradigm The computers-are-social-actors paradigm (CASA) postulates a social relationship between humans and machines (Nass et al. 1994). It assumes that humans mindlessly react with a set of social scripts from human-human interaction to simple social cues even if they are provided by non-human systems (Nass and Moon 2000; Reeves & Nass 1996). In the simplest form, the CASA paradigm can be described as the so-called media equation stating ‘media equals life’ (Reeves and Nass 1996) and assuming that social dynamics in HHIs are also existing in HRIs (Johnson et al. 2004). The CASA paradigm is based on two major presuming assumptions regarding the type of technology, as well as the sourcing of information (Gambino et al. 2020). While the technology is supposed to be an “object that has enough cues to lead the person to categorize it as worthy of social responses” (Nass and Moon 2000, p. 83), CASA tests if a person “can be induced to make attributions toward computers as if the computers were autonomous sources” (Nass and Steuer 1993, p. 511). This restricts the applicability of the paradigm on technologies that are
2.3 Theoretical Background
65
Table 2.8 Summary of Uncanny Valley Paradigm Name
Uncanny Valley Paradigm
Field of Application
Information Systems
Central Sources
Mori 1970; MacDorman and Ishiguro 2006; Mori et al. 2012
Basic Assumptions
1. The appearance of a robot affects human responses. 2. Human-likeness of robots influences human’s affinity with it. 3. Human-likeness and affinity are related through a non-linear function.
Core Message
Robotic human-likeness relates positively to feeling familiar with it. When robots look overly human-like this positive relationship declines.
Main Contribution to Thesis 1. Customers feel familiar with humanoid robots to a certain degree. 2. Customer responses differ between service robots and human employees. 3. According to the outer appearance, customers respond more positively to human employees as compared to service robots. Criticism
– Difficulties with the operationalization of human-likeness. – With just two factors (human-likeness and familiarity), the uncanny valley paradigm is too simple to explain the heterogeneous phenomena. – General lack of empirical evidence for the design of the graph.
perceived as source of communication itself, rather than just serving as channel for human-human communication (Gambino et al. 2020). In contrast to anthropomorphism that involves the thoughtful belief that objects have human characteristics, the CASA paradigm describes people mindlessly applying human social categories to computers (Nass and Moon 2000). Therefore, the CASA paradigm helps to predict unconscious human behaviors in human-computer interactions (Moshkina et al. 2014). Various studies confirmed the CASA paradigm of humans treating computers (Fogg and Nass 1997; Johnson et al. 2004) and robots (Kim et al. 2013; Moshkina et al. 2014) as social actors, including social behaviors such as gender stereotyping (Nass et al. 1997), personality preferences (Nass and Lee 2001) and feeling the obligation for the principle of reciprocity (Fogg and Nass 1997). Overall, CASA
66
2
Conceptual Background
paradigm has been verified to apply for over 100 social rules (Katagiri et al. 2001). Researchers adapted the general CASA paradigm to the applications on many technologies, such as robots (RASA: Dautenhahn 1999), Media (MASA: Xu and Lombard 2016) and websites (WASA: Karr-Wisniewski and Prietula 2010), showcasing the general applicability of the paradigm to all kind of technologies. However, there raised some criticism (Gambino et al. 2020), as recent dissonant evidence showed that computers are perceived differently from how humans are perceived (Blascovich et al. 2002; Fox et al. 2015; Krämer et al. 2012). CASA paradigm (Nass et al. 1994) strongly focuses on how people perceive certain technologies. However, over the last two decades people, technologies and the way how people interact with technologies have changed (Gambino et al. 2020). As technologies have become more complex and much more integrated into our daily lives, people got used to the interactions with technologies and may have developed specific scripts for the interaction with such technologies in the meanwhile (Gambino et al. 2020). The theoretical limitations of the CASA paradigm are becoming increasingly visible, as studies showed that humans have different expectations for interactions with technologies as for interactions with humans (Edwards 2018; Edwards et al. 2020; Spence et al. 2014), indicating that people have different scripts for the interactions with technologies than for human interactions. Further, the underlying assumptions of the CASA paradigm should be specified more precisely as the term ‘computer’ suggests applicability to a broad variety of technologies, contradicting the original definition of Nass and Steuer (1993) and the phrasing of “enough cues” (Nass and Moon 2000, p. 83) remains rather subjective. Therefore, Gambino et al. (2020) have suggested to extend the original CASA paradigm. Based on the CASA paradigm, this thesis argues that findings from HHI in service literature are transferable to HRI. Customers respond to service robots as social actors and tend to apply comparable social responses as in the interaction with the human frontline employee (in line with Nass et al. 1994), as they unconsciously attribute agency, personality and intentionality to a technology (Carpenter et al. 2009) such as the service robot (Table 2.9). The CASA paradigm applies especially when humans perceive social cues expressed by the computer (Nass and Moon 2000) or assess social presence (Lee and Nass 2003). In this thesis, the computer is represented in form of a humanoid robot with a human-like appearance and human-like communication via voice and gestures leading to the assumption that customers will show comparable responses to the FSR as to the FLE at the service encounter.
2.3 Theoretical Background
67
Table 2.9 Summary of Computers-Are-Social-Actors Paradigm Name
Computers-Are-Social-Actors Paradigm (CASA)
Field of Application
Human-Computer Interaction
Central Sources
Reeves and Nass 1996; Nass and Moon 2000
Basic Assumptions
1. Computers have social attributes. 2. Social attributes of computers trigger scripts from human-human interactions. 3. Humans unconsciously apply social rules and expectations to computers.
Core Message
Humans mindlessly apply social norms used for interacting with humans during the interaction with computers.
Main Contribution to Thesis 1. Customers mindlessly treat service robots like human frontline employees. 2. Results from service research (HHI) can be transferred to human-robot interactions. Criticism
2.3.2
– In the meanwhile, technologies, people and the way they interact with technologies have changed: people might have developed specific scripts for the interaction with technologies that differ from their scripts for HHIs. – Presuming assumptions of the CASA paradigm are unclear, making it difficult to establish objective analysis of the proposed effects.
Theories from Psychology and Social Psychology
Following the previous theories from the field of IS and HRI, this section introduces relevant theories from psychology and social psychology that are going to be applied to explain customer responses and expectations on a profound theoretical basis. In a first step, the theory originating from psychology is introduced: script theory (section 2.3.2.1). Subsequently the three theories rooted between psychology and sociology are presented and discussed: role theory (section 2.3.2.2), cognitive dissonance theory (section 2.3.2.3) and from cross-cultural social psychology Hofstede’s cultural dimensions theory (section 2.3.2.4).
2.3.2.1 Script Theory Script theory was first introduced by Tomkins (1978) and is well established in cognitive psychology. It predicts that people structure their behaviors based on scenes and scripts. A scene is a representation of a (real or imagined) event (Carlson and Carlson 1984). A set of scenes compose a script, “internalized through
68
2
Conceptual Background
actions or interactions and they require a situation to externalize the concepts in the mind” (Meng 2008, p. 132). In turn, scripts can provide behavioral guidance for specific situations (Searleman and Herrmann 1994). In particular, customers’ scripts serve as standards for evaluating their satisfaction with a service provider and its performance (Bitner et al. 1994; Mohr and Bitner 1991), as well as their buying behavior (Erasmus et al. 2002; Taylor et al. 1991). Falces et al. (2002) show that satisfaction is influenced by the customer’s script, and Taylor et al. (1991) propose using customers’ script schemata to understand their decision making. One disadvantage of assumptions based on script theory is, that cognitive scripts are not steadily constant over time (Sénécal et al. 2012) and may be associated with considerable differences from person to person (Erasmus et al. 2002). Sénécal et al. (2012) showed that exposure during a certain amount of time or interactions quickly changes cognitive scripts, leading to different perceptions of the same service interaction. Therefore, the application of script theory might be helpful to estimate customer responses during the first interactions with service robots, but also predicting changes over time. As cognitive scripts are formed based on prior environmental interactions and individual persons experienced different prior interactions over their lifetime (Nottenburg and Shoben 1980), the script formation is an individual process leading to different scripts from person to person for the same event (Erasmus et al. 2002). Further, script theory is a rather abstract concept that has mostly been applied in contexts with high levels of abstraction, although examination on a more specific level would be more productive and lead to richer results (Stoltman et al. 1989). So far, literature has done little to apply and extend script theory in practice (Erasmus et al. 2002). In line with that, empirical use of script theory in service marketing has not been reported for a long time (Erasmus et al. 2012). This thesis anticipates that script theory has important implications for understanding customer responses to HRI during service encounters, such that customers may transfer scripts from well-known service encounters to both human–human interactions and to HRIs. As cognitive scripts have been shown to be contextually determined (Read 1987) and even seen as a function of context leading to different scripts in different situations (Stoltman et al. 1989), it is important to specify the situation in corresponding experimental research to trigger the same scripts across all participants. Therefore, the experimental settings within the studies of this theses were designed as realistic as possible and validated through a special pre-test (Table 2.10).
2.3 Theoretical Background
69
Table 2.10 Summary of Script Theory Name
Script Theory
Field of Application
Psychology
Central Sources
Tomkins 1978
Basic Assumptions
1. People structure their behaviors based on scenes and scripts. 2. Scripts provide guidance in specific situations.
Core Message
People structure their behavior based on scripts based on their experiences for specific scenes and rely on these scripts when back in a comparable scene.
Main Contribution to Thesis 1. Customers’ scripts serve as standards for evaluating their satisfaction with a service provider and its performance. 2. Script theory helps to explain customer responses to human-robot interactions during service encounters. 3. Understanding customers’ script schemata helps to understand their decision making. Criticism
– Cognitive scripts of the same event might be considerably different from person to person and might not be constant over time for one person. This might limit assumptions based on script theory. – Script theory is a rather abstract concept with little implementation in practice and little empirical evidence.
2.3.2.2 Role Theory Role theory is a concept in social psychology that examines propositions regarding the emergence and acquisition of roles and the associated role expectations (Biddle 1979). It refers to social roles as “cluster of socially defined expectations that individuals in a given situation are expected to fulfill” (Franzoi 1996, p. 97) that are associated with socially defined positions (Solomon et al. 1985) and role expectations as standards for role behavior (Biddle 1986). Role taking is essential for interactions in a well-functioning society. Taking the role of others, people are able to imagine how the other might behave (Mead 1934). For most routine service interactions, employees roles and customers roles are well defined and the interacting individuals know how the other is going to behave and what can be expected (Bitner et al. 1994). Role theory proposes that individuals can act according to a socially defined position leading to role congruence or acting in contrast to this position leading to a role conflict (Biddle 1986).
70
2
Conceptual Background
Individuals expect that role congruence is rewarded and role conflicts are punished. Therefore, individuals strive for role congruence and make efforts to keep congruent with their own role and to sanction role violations from other individuals (Biddle 1986). In marketing research, role theory has been applied in personal selling, advertising (Wilson and Bozinoff 1980), and to explain customer expectations (Sheth 1967). Each situational role has to be learned in order to be able to behave consistently with the role at a high role performance. Professional service employees learn their role in trainings and apprenticeships to be able performing it at a high role performance at the interaction with the customer. In addition, customers have to get used to specific roles through education and experience as they grow up (Solomon et al. 1985) (Table 2.11). Although role theory has been applied from many different perspectives (Biddle 1986), it is still associated with some conceptual issues (Jackson 1998). As the concept of roles is defined too broadly and interpreted differently from different perspectives (Biddle 1986), role theory is a rather meaningless concept (Brissett and Edgley 2005). Further, role theory is accused to reinforce prejudices by dictating people how to behave and having difficulties to explain deviant behaviors such as creative individual behaviors (Jackson 1998). Role theory may also be over-simplistic in assuming that people act according to certain roles based on internalized behaviors, neglecting other factors such as power structures (e.g. work places requiring certain behaviors in return for a job) that induce certain role congruent behaviors (Jackson 1998). Regarding service robots, many people have no prior experiences with robots in general and many customers have no experience with a frontline service robot at the customer encounter, yet. It seems likely that in such a new role, most people are going to rely on a comparable role that is structured in a similar way. Alternatively, people may rely on idealized scripts that are internalized through vicarious socialization (Solomon et al. 1985). This thesis relies on role theory in study 1 to examine customer expectations towards a service robot at the frontline service encounter. As most customers had no prior experiences with frontline service robots, role theory was helpful to assess what type of similar script customers employed or if they even came up with idealized scripts for the interaction with the robot and how the choice of role affected customer responses.
2.3 Theoretical Background
71
Table 2.11 Summary of Role Theory Name
Role Theory
Field of Application
Social Psychology
Central Sources
Mead 1934; Moreno 1934; Biddle 1979; Solomon et al. 1985
Basic Assumptions
1. 2. 3. 4.
Core Message
Humans act according to socially defined roles and build role expectations towards the behaviors of other individuals in specific situations.
Humans act according to role requirements. Roles have to be learned. Humans strive for role congruence. Humans expect role congruence to be rewarded and role violation to be punished.
Main Contribution to Thesis 1. Customers have to build new roles for the new interaction with the FSR. 2. Differences in the choice of roles (comparable situation or idealized role) affect customer expectations towards the FSR. 3. Different role expectations for the FSR lead to different customer responses. Criticism
– The role concept is defined too broadly to draw really meaningful implications based on the application of this theory. – Role theory reinforces prejudices by dictating people how to behave and comprises difficulties to explain deviant behaviors. – Role theory neglects other factors such as power structures that also induce certain role congruent behaviors without being based on internalized roles.
2.3.2.3 Cognitive Dissonance Theory Festinger’s (1957) cognitive dissonance theory is based on the assumption that humans have a need for cognitive consistency. When humans have two or more cognitions that are relevant to each other, they can be either consonant or dissonant. Cognitions that match each other consistently are consonant, whereas opposing cognitions are dissonant. Cognitive structures are formed based on beliefs or actions and in case of dissonance, they are associated with psychological discomfort. The level of dissonance and associated psychological discomfort caused by the inconsistency between a human’s cognitions depends on the importance of the cognition and the degree of inconsistency (Festinger 1957; Szanja and Scamell
72
2
Conceptual Background
1993). High importance of the cognition and high levels of inconsistence are associated with an increased magnitude of dissonance. Therefore, humans attempt to attain a state of consonance by reducing the dissonance. It can be reduced by changing beliefs, changing actions or changing the perception of actions in a way that new consonant cognitions are added or the importance of dissonant cognition is reduced (Festinger 1962). Figure 2.7 provides an overview on the framework of cognitive dissonance theory: starting on the left it shows the emergence of cognitive dissonance based on inconsistent actions and beliefs. On the right, it shows the three options for humans to reduce the dissonance by changing beliefs, changing actions or changing perceptions of the action (see Festinger 1957).
Action
Inconsistency
Belief
Change Belief
Dissonance…
Change Action
Dissonance…
Change Action Perception
Figure 2.7 Framework of the Cognitive Dissonance Theory. (adapted from Festinger 1957)
Marketing research has relied on this psychological theory to predict customer shopping behavior (Gbadamosi 2009) and customer responses to different pricing strategies (Mullikin 2003). Further, it is widely applied to examine post-purchase dissonance in comparison to advertising expectations (Engel 1963) and the effects on loyalty (Ehrlich et al. 1957) and quality perception (O’Neill and Palmer 2004). At first, marketing researchers experienced difficulties to measure dissonance with respect to marketing constructs and limitations of laboratory experiments (Telci et al. 2011), but then came up with suitable scales (Oliver 2014; Sweeney et al. 2000). Although cognitive dissonance theory has been incorporated within many frameworks and generated many new and intriguing hypotheses (Cummings and Venkatesan 1976), criticism raised from psychology (Harmon-Jones and Mills 1999) and marketing perspective (Oshikawa 1969). Earlier studies (Heine and
2.3 Theoretical Background
73
Lehman 1997) showed that the occurrence of this theory depends on the cultural context and might be much weaker if not even non-significant in collectivistic cultures such as many Asian countries. Further, the underlying mechanism of the results described by dissonance theory, might be caused and explained with other paradigms such as the ‘free-choice paradigm’ (Harmon-Jones and Mills 1999) or the ‘forced-compliance paradigm’ (Cummings and Venkatesan 1976), as the cognitive dissonance theory does not describe the mechanism itself sufficiently. From a marketing perspective, researchers experienced difficulties measuring dissonance regarding marketing constructs and identified limitations in marketing lab experiments (Telci et al. 2011). It is unclear what sorts of services are sufficiently important to arouse dissonance effects (Cummings and Venkatesan 1976), as not all conditions provoke dissonance that is strong enough to trigger dissonance reducing behaviors (Oshikawa 1969). Further, cognitive dissonance theory provides a large variety of possible dissonance-reducing behaviors and a priori estimation of the dissonance reduction mode that customers apply is difficult (Cummings and Venkatesan 1976). In study 3 this thesis relies on cognitive dissonance theory to explain customer responses to unexpected service robot behavior, arguing that discrepancies between expected and perceived robotic service behavior increase customer discomfort and linking it to the expectation disconfirmation theory (see section 2.3.1). Regarding service failures, study 3 assumes that robotic failures in the service interaction lead to cognitive dissonance for the customer (Table 2.12).
2.3.2.4 Cultural Dimensions Theory Culture can be defined as “the complex whole which includes knowledge, belief, art, morals, custom and any other capabilities and habit acquired by man as a member of society” (McCort and Malhotra 1993, p. 97). This definition shows that the general concept of culture is too global to be a meaningful explanatory variable in empirical research (van de Vijver and Leung 1997). Therefore, Hofstede (1980) introduced his cultural dimensions theory relying on just a limited number of dimensions for the comparison of cultures. Hofstede’s (1980; 2003) cultural dimensions are the most widely applied cultural framework in social psychology and marketing research (Steenkamp 2001). He relied on a survey with IBM employees comprising respondents from 67 countries and originally created four cultural dimensions: power distance, individualism, uncertainty avoidance, and masculinity (Hofstede 1980). Every country’s cultural dimension was rated on a scale of 1–100 for each cultural dimension and included further data on demographic and economic aspects of society (Kale and Barnes 1992):
74
2
Conceptual Background
Table 2.12 Summary of Cognitive Dissonance Theory Name
Cognitive Dissonance Theory
Field of Application
Psychology
Central Sources
Festinger 1957; Festinger 1962
Basic Assumptions
1. Cognitive dissonance causes a state of psychological discomfort. 2. Humans have a need for cognitive consistency, striving to reduce cognitive dissonance.
Core Message
Inconsistency between actions and beliefs lead to cognitive dissonance that is associated with discomfort. Therefore, humans strive to reduce the dissonance by reaching consistency.
Main Contribution to Thesis 1. Inconsistency between customer beliefs and interaction at the service encounter leads to cognitive dissonance for the customer. 2. Cognitive dissonance leads to customer discomfort. 3. The degree of customer discomfort varies based on the importance of the issue and the degree of inconsistency. Criticism
– The underlying mechanism is not described sufficiently and might be caused through other effects such as the ‘free-choice paradigm’. – The effects of the cognitive dissonance theory might vary significantly across different cultures and be less applicable for collectivistic cultures. – Situations that are not important enough to the customer, might not arouse dissonance that is strong enough to provoke dissonance-reducing behaviors.
– Power distance describes “the extent to which the less powerful members of organizations and institutions (such as the family) accept and expect that power is distributed unequally” (Hofstede and McCrae 2004, p. 62). – Individualism “refers to the degree to which individuals are integrated into groups. In individualist societies, the ties between individuals are loose: Everyone is expected to look after himself or herself” (Hofstede and McCrae 2004, p. 63). – Uncertainty avoidance describes “a society’s tolerance for ambiguity. It indicates to what extent a culture programs its members to feel either uncomfortable or comfortable in unstructured situations” (Hofstede and McCrae 2004, p. 62). – Masculinity deals with “the distribution of emotional roles between the sexes [in a way that …] women in feminine countries have the same modest, caring
2.3 Theoretical Background
75
values as the men; in masculine countries, they are somewhat assertive and competitive, but not as much as the men” (Hofstede and McCrae 2004, p. 63). Although originating in social psychology, marketing research intensely relied on Hofstede’s cultural dimensions in multiple studies (Sivakumar and Nakata 2001; Shamkarmahesh et al. 2003), showing effects of cultural dimensions on service performance, innovativeness and information exchange behavior (Soares et al. 2007). Service performance is influenced by collectivism, power distance and masculinity (van Birgelen et al. 2002); whereas innovativeness and information exchange behavior depend on collectivism (Steenkamp et al. 1999), uncertainty avoidance (Dawar et al. 1996; Yeniyurt and Townsend 2003), power distance (Dawar et al. 1996; van Everdingen and Waarts 2003), and masculinity (van Everdingen and Waarts 2003). As Hofstede’s (1980) cultural dimensions received increasing attention in research and were widely applied across several disciplines, critics on his work increased (Ailon 2008; Gerhart and Fang 2005; McSweeney 2002; Venaik and Brewer 2016; Witte 2012). Witte (2012) criticizes the choice of a nation-wide level and McSweeney (2002) argues in a comparable way, criticizing the conceptualization of culture and deriving implications on individual behaviors. In line with that criticism, Vernaik and Brewer (2016) show that Hofstedes correlations are only significant on an aggregated nation-wide level but not on individual levels, making it difficult to draw implications concerning individual people in a specific country. Further, questions were raised regarding the discriminant and convergent validity (see section 3.1.1) of the cultural dimensions (Vernaik and Brewer 2016), based on inconsistencies on a methodological and theoretical basis (Ailon 2004). Finally, it could be shown, that only less than 5 percent of the individual variance in the cultural dimensions can be explained by the nation-wide aggregated levels of cultural dimension from Hofstede (Fang and Brewer 2016). Besides criticism on the cultural dimensions theory as described above, Hofstede’s cultural dimensions are the most widely applied cross-cultural theory in psychology and management studies (Steenkamp 2001). This thesis relies on Hofstede’s cultural dimension theory in study 1 to explain intercultural differences regarding customer expectations toward professional service robots at the service encounter, as its reliable dimensions provide a basis for empirical research on cross-cultural effects (Table 2.13).
76
2
Conceptual Background
Table 2.13 Summary of Cultural Dimensions Theory Name
Cultural Dimensions Theory
Field of Application
Social Psychology
Central Sources
Hofstede 1980; Hofstede 2003
Basic Assumptions
1. Culture can be reduced to a limited number of dimensions for comparisons. 2. It makes sense to assess culture on a nation-wide level.
Core Message
Cross-national culture differences can be compared with the 4 cultural dimensions: power distance, individualism, uncertainty avoidance, and masculinity.
Main Contribution to Thesis 1. Cultural dimensions theory provides data on cultural dimensions. 2. Cultural dimensions can be compared to customer expectations toward service robots at the frontline encounter. Criticism
– Nation-wide aggregated cultural dimensions might only have little implications on individual behaviors. – Cultural dimensions showing problems regarding discriminant and convergent validity based on methodological and theoretical inconsistencies.
3
Method
Subsequent to the previous presentation of the conceptual background in chapter 2, consisting of basic definitions, a detailed literature review and theoretical basics, this chapter covers the methodical fundamentals for the further analysis of the research questions. Complying with the ambition of this thesis, it is the goal to meet the methodical requirements of the current state of marketing research. These methodical fundamentals are particularly relevant for the four empirical studies in the following chapters 4, 5, 6, and 7. While section 3.1 refers to the basic construct measurement and the operationalization of selected constructs, section 3.2 deals with data analysis methods.
3.1
Construct Measurement
In order to assess the postulated hypotheses regarding the research questions in the subsequent studies empirically, it is essential to measure the theoretical constructs in the experimental studies and surveys. Section 3.1.1 gives an overview about the basics of construct measurement and construct quality and section 3.1.2 applies these basics to selected relevant constructs of this thesis.
3.1.1
Basics of Construct Measurement
A construct is defined as „an abstract entity which represents the true, nonobservational state or nature of a phenomenon“ (Bagozzi and Fornell 1982, p. 24).
© The Author(s), under exclusive license to Springer Fachmedien Wiesbaden GmbH, part of Springer Nature 2021 M. Merkle, Humanoid Service Robots, Neue Perspektiven der marktorientierten Unternehmensführung, https://doi.org/10.1007/978-3-658-34440-5_3
77
78
3
Method
The following three sections will work out the basics of construct operationalization (3.1.1.1), different quality criteria for constructs (3.1.1.2) and introduce corresponding quality measures (3.1.1.2).
3.1.1.1 Basics of Construct Operationalization As the definition from Bagozzi and Fornell (1982) above already indicates, constructs are non-observable phenomena that are difficult to measure directly and designated as latent variables (Bagozzi and Phillips 1982; Homburg et al. 2008). In order to measure constructs correctly for empirical analyses, it is essential to specify observable indicators—so-called items—that allow to measure latent variables indirectly (Homburg and Giering 1996). This specification is referred to as operationalization and has to be conducted prior to the empirical study. There are several ways to operationalize constructs that can be characterized depending on the number of applied items and the direction of the specified relationship. Depending on the number of applied items, literature distinguished between single-item and multi-item measurement models (Diamantopoulos et al. 2012). Single-item measurement models just rely on one single item to operationalize a latent construct. Therefore, the manifestation of the single item resembles the latent construct in this specific case (Rossiter 2002). Compared to multi-item measurements, single-item measurements are associated with advantages, such as a higher practicability, keeping questionnaires short, reducing administrative work and costs of studies (Bergkvist and Rossiter 2007). However, a major drawback of single-item measurements occurs concerning the reliability, as it cannot be measured and leads to restrictions in terms of construct quality assessment (Bergkvist and Rossiter 2007). Moreover, the informative value of such measures is much weaker, as they are not measured through various indicators (MacKenzie et al. 2005). In comparison, multi-item measurement models allow a much higher reliability of the measurement (Churchill and Peter 1984) as there are various statistical measures to assess the measurement quality of the construct and to take measurement errors into account (see section 3.1.1.3). Further, the specification via multiple indicators provides a more comprehensive understanding of the construct (Rossiter 2002). Based on these advantages, this thesis widely relies on multi-item measures for all latent variables related to main effects and just includes single-item measures occasionally for control variables as suggested by Boyd et al. (2005). According to the direction of the specified relationship between the construct and its indicators, measurements can be classified in reflective measures and formative measures (Bollen and Lenox 1991). For reflective measures, indicators are caused by their corresponding latent construct (Homburg and Giering 1996)
3.1 Construct Measurement
79
and a change of the latent construct leads to a change within all associated indicators (Bollen and Lennox 1991). Therefore, indicators are just an exemplary and imperfect reflection of the latent construct and are generally exchangeable (Diamantopoulos and Winklhofer 2001; MacKenzie et al. 2005). The reflective measurement assumes that every indicator could work as a single representation of the corresponding construct (Bagozzi and Yi 2012), as the indicators are strongly correlated among each other (Jarvis et al. 2003). Constructs of formative measures, however, are a function of all underlying indicators (Fornell 1986) so that latent constructs are caused by the corresponding indicators (Diamantopoulos et al. 2008). As every single formative indicator contributes to the explanation of the construct, none of them is omittable without a loss of validity (Bollen and Lennox 1991). Overall, service marketing mainly relies on reflective measurement methods, although certain published measures are erroneously considered as reflective instead of formative constructs (Jarvis et al. 2003). Considering the theoretical argumentations as well as the causal relationships, this thesis applies reflective measurements for the latent variables. Accordingly, the subsequent sections regarding measurement quality bear on reflective measurements.
3.1.1.2 Basics on Quality Criteria In order to obtain solid findings and to generate valuable study results, it is essential to ensure high measurement quality. The quality of a measurement can be assessed based on the three following criteria: objectivity, reliability, and validity (Homburg and Krohmer 2006). Objectivity of a measurement means that the results of a measurement are independent from the researcher performing the measurement, striving to eliminate personal bias, emotions and false beliefs (Gaukroger 2012). Therefore, a measurement is objective if different researchers come to the same results when conducting the measurements independently. Depending on the stage of the measurement process, objectivity can be distinguished in objectivity of assessment, objectivity of analysis and objectivity of interpretation (Homburg and Krohmer 2006). For assessment objectivity it is essential that the examiner does not influence the behavior and the responses of the participants during data assessment (Homburg and Krohmer 2006). Subsequent to the collection of data, objectivity of analysis has to be ensured through a narrow scope of possible analyses. Objectivity of interpretation requires a limited leeway of interpretation and the reading of the measurements (Homburg and Krohmer 2006). According to this definition, objectivity has to be assured directly during the measurement.
80
3
Method
Subsequent to the data assessment, the quality of the measures has to be checked regarding the correct capturing and the operationalization of the applied constructs. Construct quality is usually examined according to criteria of reliability and validity (Homburg and Giering 1996). As reliability is a prerequisite for validity (Churchill 1979), it is the first quality requirement to analyze after data collection (Homburg et al. 2008). While reliability considers the internal consistency of a measurement, validity refers to the extent to which the indicators of a construct measure what they are intended to measure (Bagozzi and Yi 2012). Reliability of a measurement refers to “the degree to which measures are free from error and therefore yield consistent results” (Peter 1979, p. 6) and can generally be formulated (DeVellis 2016) as: reliabilit y =
tr ue scor e ; tr ue scor e = obser ved scor e − err or obser ved scor e
According to literature, there are four types of reliability: test-retest reliability, parallel-forms reliability, internal consistency reliability (Friedrichs 1990), and inter-rater reliability (Durand and Barlow 2012). Test-retest reliability is given when the results of one measurement at a certain time are strongly correlated with the results of a later measurement, based on the same indicators and the same participants (Peter 1979). Therefore, this reliability is calculated as correlation between similar measures at different points of time. However, this type of reliability is accompanied with the handicap that low values of test-retest reliability cannot clearly be attributed to an unreliable measurement, as other systematic changes between the points of measurement might also cause lower values of reliability (Rossiter 2002). Parallel-forms reliability compares the correlation of parallel data assessments relying on equivalent measurement instruments. Therefore, it requires the development of alternate measurements with equivalent content and method. The resulting correlation between the results of the equivalent measurements describes the parallel-forms reliability (Davidshofer and Murphy 2005). Relying on this type of reliability has to be viewed critically, as it might be difficult to create exactly equivalent forms of measurement instruments and the verification of equivalence is challenging (Davidshofer and Murphy 2005). Internal consistency reliability is much easier to investigate, as it just requires one data collection. It depends on the consistency of the single items of a construct. Homogeneous values across all items and high correlation among the items of a construct define a high internal consistency reliability (Peter 1979). Further, this type of reliability avoids interfering effects such as of changes over
3.1 Construct Measurement
81
time between to measurements, as well as contentual differences (Himme 2009). Therefore, this type of reliability is most widely used and well-recognized in empirical management literature (Hildebrandt 1998) and this thesis also mainly relies on this type of reliability. Inter-rater reliability describes the level of agreement between the assessments of one measure by multiple raters (Durand and Barlow 2012). Inter-rater reliability is applied when several persons conduct the same measurement or rating independently from each other and it tests whether measurement results are independent from the rater. Therefore, the correlation among the assessments of raters describes the inter-rater reliability (Durand and Barlow 2012). Besides self-reported data, this thesis also relies on external-rater assessments and applies inter-rater reliability for these measurements to evaluate measurement quality. Subsequent to ensured reliability as required precondition, validity is the next measurement quality criteria to examine (Churchill 1979). Validity describes “the degree to which instruments truly measure the constructs which they are intended to measure” (Peter 1979, p. 6). Homburg and Giering (1996) distinguish between four types of validity: content validity, nomological validity, convergent validity, and discriminant validity. Content validity refers to the degree how well the indicators of a construct represent its intended content (Rossiter 2002). A measurement is considered as contentually valid, if its indicators cover all essential aspects of the construct and do not cover other contentual aspects that are not associated with this construct (Homburg and Giering 1996). Content validity requires constructs that are operationalized based on clear and comprehensive definitions and is typically assessed on a qualitative basis (Parasuraman et al. 1988). Through the application of well-established measurements, this thesis ensures high content validity. Nomological validity is defined as the extent, to which predicted relationships between constructs are confirmed in the context of an overarching theory (Bagozzi 1979). This type of validity can be examined through consensus between empirically analyzed and theoretically postulated relationships (Homburg et al. 2008). This thesis relies on multiple theories to postulate hypotheses (see section 2.3) and therefore an assessment of the nomological validity is not applicable. Convergent validity relates to the degree of correlation among indicators of the same construct (Peter and Churchill 1986). High convergent validity means that the single indicators of one measurement show high alignment (Bagozzi and Phillips 1982). It can be determined quantitatively based on a confirmatory factor analysis (Homburg et al. 2008), whereby high levels of reliability positively affect convergent validity (Peter and Churchill 1986).
82
3
Method
Discriminant validity is a quality criterion for the extent to which measures of certain constructs differ. It describes how unique a construct is and not just reflects other variables (Peter and Churchill 1986). High discriminant validity means that the measurements of different constructs differ significantly from each other (Hildebrandt 1998). In line with convergent validity, discriminant validity can also be determined based on a confirmatory factor analysis (Homburg et al. 2008). Table 3.1 gives an overview about the most relevant criteria of reliability and validity for the measures applied in this thesis that will be quantitatively analyzed in the following sections. According to the preceding statements, section 3.1.1.3 will not include test-retest reliability and parallel-forms reliability as they would have required additional data collections. Further, the following section will not include content validity and nomological validity as it focuses on quantitative quality analyses rather than qualitative criteria.
Table 3.1 Selected Criteria of Reliability and Validity (see Hildebrandt 1998) Reliability Criteria Internal Consistency Reliability
Describes the consistency of a construct through the level of correlation among measurements of one construct.
Inter-Rater Reliability
Describes the level of agreement of multiple raters through correlation among their assessments of the same measure.
Validity Criteria Convergent Validity
Describes the extent to which measures of the same construct that should be related are indeed related in reality.
Discriminant Validity
Describes the extent to which measures of different constructs that should not be related are indeed not related in reality.
3.1.1.3 Basics on Quality Measures This section introduces several quality measures to assess the quality criteria of internal consistency reliability, inter-rater reliability, convergent validity, and discriminant validity as introduced in the previous section. It is important to mention that the presented quality measures only represent a selected extract of the wide
3.1 Construct Measurement
83
variety of quality measures that were developed in literature. The presented measures are not meant to exclude each other, but rather complement each other in terms of measurement quality regarding reliability and validity. It must be stated, that the following measures just apply for reflective constructs and are not generally suitable for the analysis of formative constructs (Jarvis et al. 2003). There is a large number of quality measures regarding reliability and validity of constructs. Basically, they can be classified in quality measures of the first generation and measures of the second generation (Homburg 1995). Quality measures of the first generation are rooted in classical test theory and have been developed in the 1950s (Homburg and Giering 1996). Hereafter, this thesis deals with the following quality measures of the first generation: the exploratory factor analysis, Cronbach’s alpha, the intraclass correlation coefficient, and the item-to-total correlation. The exploratory factor analysis is based on the idea to condense a large number of manifest indicator variables to a smaller number of latent factors (Hüttner and Schwarting 2008). This analysis is based on a correlation matrix indicating the mutual dependencies of the indicator variables (Hüttner and Schwarting 2008). Subsequent to the condensation of the indicator variables to factors, it is possible to assess convergent validity and discriminant validity on the basis of factor loadings (Homburg and Giering 1996). Sufficient convergent and discriminant validity can be assumed, if the indicator variables have a sufficiently high charge on one factor, while they have significantly lower charges on the other factors (Gerbing and Anderson 1988): as far as all indicators have a factor loading of >0.40 on one factor and all the loadings on other factors are significantly lower, the indicators can be condensed to one factor (Homburg and Giering 1996). Cronbach’s alpha is one of the most widely applied measures of the first generation (Homburg and Giering 1996) for the internal consistency of constructs (Peter 1979) and was first introduced by Cronbach (1951). It is defined as a “function of the extent to which items in a test have high communalities and thus low uniquenesses [and it can be considered as] a function of interrelatedness” (Cortina 1993, p. 100). Cronbach’s alpha is calculated as follows (Cronbach 1951): k 2 k i=1 σi ∝= 1− k−1 σt2 with k as the number of indicators, σi2 as the variance of the indicator i and σt2 as the total variance of all indicators summarized. This quality measure ranges between 0 and 1, whereas a high ∝ indicates strong internal consistency of the
84
3
Method
construct (Homburg and Giering 1996). In line with Nunnally (1978), this thesis considers values of ∝ ≥ 0.7 as reliable, although more recent studies even considered 0.6 as a moderate threshold value for acceptable reliability. The intraclass correlation coefficient (ICC) is a quality measure to assess the consensus among two or more raters regarding their rating of one measure (Shrout and Fleiss 1979). Judgements made by humans as independent raters are especially affected from the problem of measurement errors and thus, it is essential to assess the degree of this error by calculating a reliability index such as the ICC (Shrout and Fleiss 1979). The ICC measure was first introduced by Fisher (1954), who based the calculation on the Pearson correlation coefficient, but centering and scaling each variable based on a pooled mean and standard deviation (Bartko 1966). However, modern ICC quality measures are calculated by mean squares based on analyses of variance (Koo and Li 2016). There are various ICC versions that are corrected for specific types of data depending on the assignment of independent raters to the observed participants, the unit of reliability and the importance of the effects due to judges for the reliability index (Shrout and Fleiss 1979). In this thesis, each participant was rated by each of the same k raters, who were the only raters of interest. Therefore, the measures obtained through independent raters in this thesis will be rated with the ICC(3,1) based on a two-way mixed ANOVA, calculated as follows (Shrout and Fleiss 1979): I CC(3, 1) =
BMS − EMS B M S + (k − 1)E M S
where k is to the number of raters rating each participant, B M S refers to the between-targets mean square and E M S refers to the expected mean squares. ICC values above 0.60 are considered as good and values above 0.75 are considered as excellent quality indicators for inter-rater agreement measures (Cicchetti 1994). The item-to-total correlation is the last quality measure of the first generation that is mentioned in this section. It is a measurement for the convergent validity and ranges between 0 and 1 (Homburg and Giering 1996) with high values indicating a better validity, although there is no fix threshold value (Nunnally 1978). Item-to-total correlation is widely applied as criterion to eliminate specific indicators in order to improve the internal consistency of a measurement (Churchill 1979). Thus, the indicator with the lowest item-to-total correlation can be eliminated in case of an insufficient Cronbach’s alpha value (Churchill 1979), as all indicators are interchangeable anyway for reflective scales (Bollen and Lennox 1991).
3.1 Construct Measurement
85
Despite their widespread use, the quality measures of the first generation show some weaknesses (Fornell 1986; Hildebrandt 1998), as they are mainly based on heuristics (Bagozzi et al. 1991) with restrictive presumptions (Gerbing and Anderson 1988) and do not consider measurement errors (Homburg and Giering 1996). Therefore, more powerful quality measures of the second generation were developed based on confirmatory factor analyses, including the possibility to take measurement errors into account (Homburg and Giering 1996). The confirmatory factor analysis relies on a factor structure that has been postulated prior to the analysis that is verified by means of the available data (Homburg and Giering 1996). Quality measures of the second generation can be classified in global and local measures (Neil 2001). Global quality criteria are used to assess the overall quality of the entire model, whereas local quality criteria assess the quality of measurements on an indicator and factor level, focusing on reliability and validity (Homburg and Giering 1996), as intended in this section. Literature recommends local quality criteria focusing on indicator level such as indicator reliability, as well as on factor level criteria such as composite reliability, average variance extracted, and the Fornell and Larcker criterion (Homburg et al. 2008). On indicator level, the indicator reliability (IR) indicates how much variance of the considered indicator xi is explained by the corresponding construct (Bagozzi and Yi 2012). IR is calculated as follows with values ranging between 0 and 1, where high values indicate a strong indicator reliability (Homburg et al. 2008): I R(xi ) =
λi2j φ j j λi2j
φ j j + θii
where λi j is the factor loading of the indicator xi and φi j is the estimated variance of the corresponding factor ξ j . Further, θ j j is the estimated variance of the corresponding measurement error δi . Indicators can be considered as reliable for I R ≥ 0.40 according to Homburg and Giering (1996) or I R > 0.50 according to Bagozzi and Yi (2012), respectively. However, a strict orientation on these values for indicator reliability is accompanied by the danger of reducing content validity (Little et al. 1999). While IR assesses the reliability on an indicator level, the following quality measures focus on the factor level. Composite reliability (CR) describes how well a factor can be measured with its corresponding indicators (Homburg and Giering 1996). Therefore, the CR is a measure for convergent validity as well as for confirming reliability (Bagozzi and Yi 2012). CR is calculated as follows with values ranging between 0 and 1, where high values indicate a strong indicator
86
3
Method
reliability (Homburg et al. 2008): C R ξj = k
k
i=1 λi j
i=1 λi j
2
2
φjj
φjj +
k
i=1 θii
with ξ j as latent construct, k as number of indicators, λi j as estimated factor loading of the indicator, φ j j as estimated variance of the construct ξ j and θii as estimated variance of the measurement error. Values of C R ≥ 0.60 are considered to indicate high measurement quality with good composite reliability (Bagozzi and Yi 1988). The average variance extracted (AVE) also assesses how well a construct is measured by its corresponding indicators (Homburg and Giering 1996). In detail, the AVE describes the amount of the total variance of all indicators that can be explained by the corresponding factor (Homburg et al. 2008). Comparable to the CR, the AVE ranges between 0 and 1, with high values indicating a high measurement quality. The AVE is calculated as follows: AV E ξ j = k
k
2 i=1 λi j φ j j
2 i=1 λi j φ j j
+
k
i=1 θii
with ξ j as latent construct, k as number of indicators, λi j as estimated factor loading of the indicator, φ j j as estimated variance of the construct ξ j and θii as estimated variance of the measurement error. Literature suggests that values of AV E ≥ 0.50 indicate high measurement quality (Fornell and Larcker 1981). However, this thesis mainly relies on the AVE for the calculation of the Fornell and Larcker criterion. This criterion is considered as an extremely strict criterion for the assessment of discriminant validity (Anderson and Gerbing 1993). It requires that a construct is associated stronger with its indicators than with all other considered constructs (Homburg and Klarmann 2006). This means, that the AVE of a factor has to be higher than the squared correlation r 2 of this factor with all other factors considered in the same study (Fornell and Larcker 1981): AV E ξ j > r 2 ξi , ξ j ,
∀i = j
3.1 Construct Measurement
87
Table 3.2 gives an overview about the presented quality measures with corresponding quality criteria and associated requirement levels of the individual measures. The order of the listed measures aligns with the process of measurement quality assessment (Homburg and Giering 1996). Beginning with the first generation of quality measures, the process starts with the calculation of Cronbach’s alpha. In case of values below the required threshold, the indicator with the lowest itemto-total correlation is eliminated and this process is repeated until alpha meets the requirements. The next step comprises an exploratory factor analysis, checking if the extracted factors can explain at least 50 percent of the variance from the corresponding indicators. If that is not the case, indicators with a factor loading below 0.40 should be eliminated. Subsequently, the quality measures of the second generation can be calculated (Homburg and Giering 1996). Table 3.2 Selected Quality Measures with Threshold Values Quality Measure
Quality Criteria
Threshold Value
1. Generation Quality Measures - Cronbach’s alpha
reliability
≥0.70
- item-to-total correlation
convergent validity
ordinal scale
- exploratory factor analysis
convergent and discriminant validity
≥0.40
- intraclass correlation coefficient (ICC)
reliability
≥0.60
- indicator reliability (IR)
reliability
≥0.40
- composite reliability (CR)
reliability and convergent validity
≥0.60
- average variance extracted (AVE) reliability and convergent validity
≥0.50
- Fornell and Larcker criterion
AVE (ξj ) > r2 (ξi, ξj )
2. Generation Quality Measures
discriminant validity
Further, it is important to consider the variety of methods in survey-based research, as common method bias might distort the measurements. This type of
88
3
Method
potential error describes the possible biases of variables as well as the relation between two variables that are caused by the assessment of the same measurement method (Podsakoff et al. 2012). Common method bias can be reduced through adjustments in the study design and the measurement method (Podsakoff et al. 2003). If applicable, it is beneficial to gather input and output variables from different sources or at different points of time and always assure complete anonymity for participants (Podsakoff et al. 2003).
3.1.2
Operationalization of Constructs
Following to the general basic of construct operationalization, quality criteria and corresponding quality measures (in section 3.1.1), this section applies the basics of construct measurement to selected major constructs of this thesis. According to the research framework (in section 1.3), this thesis focuses on customer expectations and customer responses regarding humanoid robots at the service encounter. Customer expectations will be examined based on customer attributions of empathy, reliability, expertise, and trust of service robots (study 2), as well as on expected levels of innovative service behavior (study 3). Further, this thesis relies on constructs for robot anxiety (study 1) and technological affinity (study 3) to examine moderating effects and as control variables. As dependent variables, this thesis relies on customer responses such as ease of use and usefulness (study 1) inspired from TAM, as well as perceived innovative service behavior and customer delight (study 3). Further, it considers customer satisfaction (study 4) as another customer response that is well-established in service marketing. As these constructs are of special importance for the studies included in this thesis, they will be operationalized and tested for measurement quality in the following sections.
3.1.2.1 Construct Operationalization for Customer Expectations Service marketing literature identified service representative attributions of empathy, reliability, expertise, and trust as driving factors for customer responses (Homburg and Stock 2005). Therefore, these expectations are especially interesting to examine regarding service robots, as most of the customers have not yet experienced real human-robot interactions at the service encounter. As Homburg and Stock (2005) developed their measures for human-human interactions with FLEs, this thesis adapted the items of these constructs to apply for service interactions with FSRs.
3.1 Construct Measurement
89
First studies (see section 5.1) have shown, that there are significant intercultural differences regarding the perception of service robots. Therefore, this thesis took a closer look on intercultural differences, including participants of several countries across three continents (India, the US, and Germany) to identify cultural differences. This intercultural comparison required the method of online surveys to facilitate recruiting of participants across the countries mentioned. Participants were shown a picture of Softbank’s Pepper robot as potential service robot, to give an example on what type of service robots the study focuses on. The following constructs were adapted from Homburg and Stock (2005). Empathy is “the capacity to clearly project an interest in others and to obtain and reflect a reasonable complete and accurate sense of another’s thoughts, feelings, and experiences” (Bush et al. 2001, p. 394) and the construct is defined through the following five items: Table 3.3 Items and Quality Measures for “Attributed Empathy” Attributed Empathy Items
Item-to-Total Correlation
Factor Loading
In my opinion, a service robot is typically able to … … have a high level of empathy with respect to my need as a customer.
0.74
0.83
… have no difficulty to determine my needs.
0.77
0.86
… try to determine my needs by adopting my perspective.
0.84
0.90
… find it easy to adopt my perspective 0.82 as a customer.
0.89
… adapts its interactions to my needs in different situations.
0.72
0.81
Construct Quality Measures
Value
Threshold
Cronbach’s alpha (α)
0.91
≥0.70
Composite reliability (CR)
0.93
≥0.60
Average variance extracted (AVE)
0.74
≥0.50
Fornell and Larcker criterion
fulfilled
Table 3.3 gives an overview about the items and provides selected quality measures to assess the construct validity and reliability. The construct performs well on item-level (item-to-total correlation) that rank above required threshold values.
90
3
Method
Likewise, it performs well on the factor-level (Cronbach’s alpha, composite reliability, and average variance extracted), exceeding required threshold values and even the criterion of Fornell and Larcker is fulfilled as examined with a correlation table. Therefore, this construct for attributed empathy is considered as valid for the further examination of proposed effects.
Table 3.4 Items and Quality Measures for “Attributed Reliability” Attributed Reliability Items
Item-to-Total Correlation
Factor Loading
In my opinion, a service robot is typically able to … … be relied on.
0.57
0.81
… make sure that promised deadlines are met.
0.36
0.55
… be sure that my instructions are precisely followed.
0.63
0.86
… be very reliable.
0.62
0.87
Construct Quality Measures
Value
Threshold
Cronbach’s alpha (α)
0.71
≥0.70
Composite reliability (CR)
0.86
≥0.60
Average variance extracted (AVE)
0.61
≥0.50
Fornell and Larcker criterion
fulfilled
Reliability “is defined as the extent to which a salesperson assures that promises made to customers are met and that customer instructions are precisely followed” (Homburg and Stock 2005, p. 402) and the construct is defined through the four items shown in Table 3.4. The table further indicates, that all requirements of the first and second generation that were introduced in section 3.1.1.3 are met and therefore the construct of attributed reliability is considered as valid for further examinations. Expertise “is defined as the presence of knowledge and ability to fulfill a task” (Bush et al. 2001, p. 394) and in line with Homburg and Stock (2005) it is defined via eight items that were slightly adapted to fit the application of a service robot instead of a human frontline employee. Table 3.5 shows that all the quality criteria are met and although this construct contains as many as eight items, they still load all on the same factor with a high
3.1 Construct Measurement
91
Table 3.5 Items and Quality Measures for “Attributed Expertise” Attributed Expertise Items
Item-to-Total Correlation
Factor Loading
In my opinion, a service robot is typically able to … … find an adequate solution if I have individual requirements.
0.63
0.71
… offer me solutions which are very well thought through.
0.66
0.74
… have the expertise that is needed to understand the information provided by me as a customer.
0.69
0.77
… be very well organized.
0.72
0.81
… know its company’s product/service range very well.
0.67
0.78
… be very knowledgeable.
0.70
0.79
… hardly make mistakes.
0.58
0.67
… be knowledgeable about the newest developments (new products, new technologies, etc.).
0.63
0.73
Construct Quality Measures
Value
Threshold
Cronbach’s alpha (α)
0.88
≥0.70
Composite reliability (CR)
0.91
≥0.60
Average variance extracted (AVE)
0.56
≥0.50
Fornell and Larcker criterion
fulfilled
composite reliability. Therefore, the construct of attributed expertise is considered as valid and will be included in the further analysis of results in study 2. Trust can be defined as “a personal characteristic that refers to a willingness to rely on an exchange partner in whom one has confidence” (Moormann et al. 1993, p. 82) and is based on the following six items based on Homburg and Stock (2005). Table 3.6 shows that also the last attribution construct from Homburg and Stock (2005) meets all the measurement criteria and is considered as valid. This is not surprising, as these constructs were adapted from validated scales (Homburg and Stock 2005). While all four attribution constructs above have been examined in an international comparison, the next three constructs were assessed in lab studies, allowing real face-to-face interactions with service robots.
92
3
Method
Table 3.6 Items and Quality Measures for “Attributed Trust” Attributed Trust Items
Item-to-Total Correlation
Factor Loading
In my opinion, a service robot is typically able to … … trust this robot to a large extend.
0.73
0.82
… be convinced that this service robot would keep its promises made to me.
0.74
0.83
… believe that this service robot would 0.72 be fair and honest with me.
0.81
… believe that the information provided 0.73 by this service robot would be correct.
0.82
… be convinced that this service robot would deliver the products/services correctly.
0.72
0.82
… be convinced that this service robot would keep my best interests in mind.
0.67
0.77
Construct Quality Measures
Value
Threshold
Cronbach’s alpha (α)
0.89
≥0.70
Composite reliability (CR)
0.92
≥0.60
Average variance extracted (AVE)
0.66
≥0.50
Fornell and Larcker criterion
Fulfilled
As innovative service behavior has been identified as a relevant service behavior for service representatives at the customer encounter (Coelho et al. 2011), study 3 examines the effect of innovative service behavior comparing human-human interactions with human-robot interactions. Further, expectation disconfirmation theory suggests that prior expectations may play a major role for customer reactions to certain behaviors. Therefore, data were collected from the lab study participants prior to the experimental studies to assess expected innovative service behavior. The construct was assessed by a student sample with a digital survey in a lab associated with the university. It consists of four items that were adapted from Stock et al. (2017) and inspired by Janssen (2000) and Stock (2015). Table 3.7 gives an overview about the construct of expected innovative service behavior and shows that all quality measures of the first and second generation on item-level and on construct-level exceed the threshold values. In the later analyses of study 3 these results will be compared to the actually perceived innovative
3.1 Construct Measurement
93
service behavior and analyzed through polynomial modelling in combination with surface response analysis.
Table 3.7 Items and Quality Measures for “Expected Innovative Service Behavior” Expected Innovative Service Behavior Items
Item-to-Total Correlation
Factor Loading
I guess the service representative… … is innovative.
0.76
0.86
… comes up with creative solutions.
0.79
0.89
… develops new ideas.
0.80
0.89
… is creative.
0.77
0.88
Construct Quality Measures
Value
Threshold
Cronbach’s alpha (α)
0.91
≥0.70
Composite reliability (CR)
0.92
≥0.60
Average variance extracted (AVE)
0.54
≥0.50
Fornell and Larcker criterion
fulfilled
In addition to the constructs above that represent independent variables of the research framework, this section further operationalizes the constructs of robot anxiety and technological affinity as they will be considered as potential moderators and controls. Nomura et al. (2008) introduced a well-established robot anxiety scale and could already show first effects of customers’ robot anxiety on customer responses toward the robot. The original robot anxiety consists of three subscales, whereof the subscale focusing on anxiety regarding robotic behavior fits best to the considerations in study 1. Table 3.8 confirms the construct of robot behavior anxiety that was already validated by Nomura et al. (2008). Although one of the items just features a low item-to-total correlation of 0.49, it was not removed as the Cronbach’s alpha of the construct still exceeds the threshold value indicating a high reliability. Therefore,
94
3
Method
Table 3.8 Items and Quality Measures for “Robot Anxiety” Robot Behavior Anxiety Items
Item-to-Total Correlation
Factor Loading
What kind of movement the robot will make.
0.70
0.86
What the robot is going to do.
0.75
0.89
How strong the robot is.
0.49
0.68
How fast the robot will move.
0.62
0.79
Construct Quality Measures
Value
Threshold
Cronbach’s alpha (α)
0.81
≥0.70
Composite reliability (CR)
0.88
≥0.60
Average variance extracted (AVE)
0.65
≥0.50
Fornell and Larcker criterion
fulfilled
the construct of robot behavior anxiety can be examined as moderator within the social robot acceptance model in study 1. As participants for the experimental studies were recruited offline, there might be significant differences among participants in terms of technological affinity that might well affect their responses to service robots as technological product. Unfortunately, there is no established scale available in literature that fits for the desired kind of technological affinity. Therefore, the authors of study 3 (Stock and Merkle 2018) developed new scale consisting of three items based on expert interviews. The measurement quality could be confirmed, as all item-level quality measures as well as the construct-level based quality measures performed really well, ranging above the required threshold values. Table 3.9 presents the items of the construct with the corresponding quality measures.
3.1.2.2 Construct Operationalization for Customer Responses Besides the constructs from the previous section representing independent variables in the overall research framework of this thesis, this section operationalizes five constructs that the following studies are going to rely on as dependent variables. Study 1 introduces a service robot acceptance model relying on the wellestablished technology acceptance model (Davis 1989). Therefore, it also relies on the two corresponding constructs of ease of use as well as usefulness.
3.1 Construct Measurement
95
Table 3.9 Items and Quality Measures for “Technological Affinity” Technological Affinity Items
Item-to-Total Correlation
Factor Loading
I have good knowledge about technical 0.68 products.
0.86
It is important to me to have good knowledge about the functionality of technological products.
0.69
0.86
Technological products fascinate me.
0.67
0.85
Construct Quality Measures
Value
Threshold
Cronbach’s alpha (α)
0.79
≥0.70
Composite reliability (CR)
0.89
≥0.60
Average variance extracted (AVE)
0.73
≥0.50
Fornell and Larcker criterion
fulfilled
Table 3.10 Items and Quality Measures for “Ease of Use” Ease of Use Items
Item-to-Total Correlation
Factor Loading
I find it easy to make the robot do what I 0.61 want.
0.77
It is easy for me to learn how to interact with the robot.
0.66
0.82
The interaction with this robot is clearly understandable.
0.70
0.85
It is easy to interact with humanoid robots.
0.70
0.84
Construct Quality Measures
Value
Threshold
Cronbach’s alpha (α)
0.84
≥0.70
Composite reliability (CR)
0.89
≥0.60
Average variance extracted (AVE)
0.67
≥0.50
Fornell and Larcker criterion
fulfilled
Table 3.10 gives an overview about the four items of ease of use that were slightly adapted to fit the focus of study 3 with a robot as technology. As shown in the table, all quality measures of this validated scale were met, indicating a
96
3
Method
strong and valid construct for the implementation in the service robot acceptance model. The other construct adapted from the technology acceptance model is usefulness (Davis 1989). This three-item construct was slightly adapted to fit for the purpose of this research but it still strongly based on the validated construct from Davis (1989). All items show a decent loading on the same factor, that shows solid quality measures on the factor-level as well. Table 3.11 shows the items and the corresponding quality measures of the construct.
Table 3.11 Items and Quality Measures for “Usefulness” Usefulness Items
Item-to-Total Correlation Factor Loading
The capabilities this robot offers are useful.
0.72
0.89
This robot was useful for my needs.
0.80
0.92
This robot fulfilled all my personal needs 0.66 at a hotel check-in.
0.84
Construct Quality Measures
Value
Threshold
Cronbach’s alpha (α)
0.85
≥0.70
Composite reliability (CR)
0.91
≥0.60
Average variance extracted (AVE)
0.78
≥0.50
Fornell and Larcker criterion
Fulfilled
The previous section already introduced the construct expected innovative service behavior. The perceived innovative service behavior is based on the same items (Stock et al. 2017) with minimal adaptations, as this construct assesses participant ratings after the experimental interaction with the robot, whereas expected innovative service behavior was assessed prior to the experiment. Therefore, it is no surprise that Table 3.12 confirms the findings from above, indicating a valid construct that meets all the required quality criteria. In study 3, the two constructs will be compared to examine effects of disconfirmation between ex-ante expectations and ex-post perceptions of innovative service behaviors on corresponding customer responses. In particular, study 3 examines the effects of disconfirmation on customer delight. The construct is based on a three-item scale adapted from Finn (2005) and Riek et al. (2010) and is supposed to measure the “profoundly positive emotional state generally resulting from having one’s expectations exceeded to a surprising degree” (Rust and Oliver 2000, p. 86).
3.1 Construct Measurement
97
Table 3.12 Items and Quality Measures for “Perceived Innovative Service Behavior” Perceived Innovative Service Behavior Items
Item-to-Total Correlation
Factor Loading
The service representative… … was innovative.
0.75
0.87
… came up with creative solutions.
0.63
0.79
… developed new ideas.
0.62
0.78
… was creative.
0.75
0.87
Construct Quality Measures
Value
Threshold
Cronbach’s alpha (α)
0.83
≥0.70
Composite reliability (CR)
0.86
≥0.60
Average variance extracted (AVE)
0.60
≥0.50
Fornell and Larcker criterion
fulfilled
As section 3.1.1.3 already mentioned, the method of data collection should be designed in a way to reduce the risk of common method bias. It suggested that besides the measurement at different points of time, it might be beneficial to assess data from different sources or different participants, respectively. Therefore, study 3 relied on independent rater assessments to assess the customer delight of participants during the interaction with the robot. Each item was assessed by three independent raters for each participant based on video-recording of the interaction. Therefore, the intraclass correlation coefficient had to be calculated first, to assure high consistency among the ratings of the independent raters, before the construct itself could further be examine in terms of measurement quality. The ratings of customer delight showed a high consistency among the raters with an ICC(3,1) of .72. Table 3.13 gives an overview about the three items of the construct and the quality measures that were all above the required factor loadings, indicating a valid construct for further consideration as dependent variable. Besides customer delight, service marketing frequently relies on customer satisfaction as positive customer response to an interaction with a firm. As this measure is well-established in literature it has already been validated. We adapted the construct from Stock et al. (2017) and Homburg et al. (2009) and relied on the five items as shown in Table 3.14. As expected, all the quality measurement criteria scored above the required threshold values indicating high validity as well as high reliability of this construct.
98
3
Method
Table 3.13 Items and Quality Measures for “Customer Delight” Customer Delight Items
Item-to-Total Correlation Factor Loading
The service representative was looking for 0.67 ways to delight the customer.
0.86
The service representative noticed even the 0.71 smallest things that were important for the customer.
0.88
The service representative treated the customer in a way that made him happy.
0.68
0.86
Construct Quality Measures
Value
Threshold
Intraclass correlation coefficient; ICC(3,1) 0.72
≥0.60
Cronbach’s alpha (α)
≥0.70
0.85
Composite reliability (CR)
0.86
≥0.60
Average variance extracted (AVE)
0.67
≥0.50
Fornell and Larcker criterion
Fulfilled
Table 3.14 Items and Quality Measures for “Customer Satisfaction” Customer Satisfaction Items
Item-to-Total Correlation
Factor Loading
This service experience was fun.
0.73
0.85
The service was helpful to complete the check in.
0.48
0.61
The check in was a positive experience for me.
0.88
0.94
I am really satisfied with this service experience.
0.83
0.91
I really liked the service.
0.72
0.83
Construct Quality Measures
Value
Threshold
Cronbach’s alpha (α)
0.88
≥0.70
Composite reliability (CR)
0.92
≥0.60
Average variance extracted (AVE)
0.70
≥0.50
Fornell and Larcker criterion
Fulfilled
3.2 Data Analysis Methods
99
3.2
Data Analysis Methods
3.2.1
Analysis of Variance and Corresponding Post Hoc Tests
3.2.1.1 Analysis of Variance Most studies of this thesis rely on experimental data with several manipulations. Each manipulative condition is applied to a specific group of participants. Later analysis compares a specific measure that was assessed in all experimental conditions. For the interpretation of the effects caused by the manipulation, analyses focus on comparisons of mean values between groups. If there is a difference between the mean values x and y of one measurement across different manipulation groups, it is essential to assess whether this difference between the two samples is large enough to assume that it also applies to the overall population. This requires analysis methods to evaluate the significance of mean differences across experimental manipulation groups. In case of just comparing two mean values of one measure assessed in two experimental groups, a t-test as inductive and bivariate analysis is appropriate. It tests the null hypothesis that the means of both groups are equal, assuming that both samples are normally distributed (Homburg et al. 2008): H0 : μ1 = μ2 with t =
x−y
(n 1 +n 2 ) (n 1 −1)s12 +(n 2 −1)s22 n 1 n 2 (n 1 +n 2 −2)
There, n 1 and n 2 refer to each sample size, while s1 and s2 refer to the standard deviation of the two groups. The test value of the null hypothesis indicated above is t-distributed with d f = n 1 + n 2 − 2 degrees of freedom. Comparing the calculated test value with the corresponding critical t-value, allows the testing of the null hypothesis (Homburg et al. 2008). However, in many cases experimental studies include more than two experimental groups and require the comparison of more than the mean values of two groups with each other. As t-tests are not suitable for this kind of analysis, this requires the analysis of variance method. In contrast to a t-test, this method distinguishes between dependent and independent variables. In terms of experimental data, the affiliation with a manipulation group might be considered as independent variable and the dependent variable is analyzed regarding the mean difference. Therefore, variance analysis is considered as a type of dependency analysis, examining the effects of a nominal independent variable on a metric dependent variable to assess whether there are significant differences between
100
3
Method
different groups. Thereby, the variance of a dependent variable is explained by the influence of one or more independent variables. Depending on the number of independent variables, there are different types of variance analyses. Single-factor analyses of variance just include one independent variable, whereas multifactor analysis of variance includes two or more independent variables. Both types of analyses of variance are abbreviated as ANOVA. If there is more than one dependent variable, multivariate analyses of variance are referred to as MANOVA. In case of additional consideration of the error caused by a covariate, the analysis of covariance (ANCOVA) is the appropriate method for one dependent variable and the multivariate analysis of covariance (MANCOVA) is applied in case of multiple dependent variables (Hermann and Landwehr 2008). The subsequent description of the sequential analysis of variance process focuses on a multi-factor analysis of variance including two independent and one dependent variable to give an exemplary overview about the process. In general, the analysis of variance requires four major steps as depicted in Figure 3.1.
Step 1:
Specification of the Model
Step 2:
Partitioning of the Sum of Squares
Step 3:
Quality Assessment
Step 4:
Interpretation of Results
Figure 3.1 Sequential Process of the Analysis of Variance. (see Homburg 2014)
The first step focuses on the specification of the model. In the case of a twofactor design, the model can be described as following (Homburg and Krohmer 2006): Yghk = μ + αg + βh + (α ∗ β)gh + εghk with αg and βh representing the main effects of the two independent variables x 1 and x2 , with μ as mean value of the dependent variable in the overall population and (α ∗ β)gh as interaction effect between x A and x B . Thereby, Yghk is the k th observation value in group g, regarding x A and regarding x B in group h with εghk as residual. For multi-factor analysis of variance, it is important to distinguish
3.2 Data Analysis Methods
101
between direct effects and moderating effects. Moderating effects affect the direct effect of an independent variable on the dependent variable. For multi-factor designs with multiple independent variables, there might be even interaction among more than two variables, although the interpretation of these effects is complex (Koschate 2002). The partitioning of the sum of squares is the next step and includes the partitioning of the total sum of squares SS Y into components related to the sum of squares within the groups SS e and the sum of squares between the groups. The sum of squares between the group is separated in the sum of squares caused by factor A (SS A ), the sum of squared caused by factor B (SS B ) and the sum of squared residuals caused by the interaction between A and B (SS Ax B ). The partitioning can be displayed as following: SS Y = SS e + SS A + SS B + SS Ax B Step 3 concerns the quality assessment of the model. First an F-test should be conducted regarding the whole model of analysis to test whether it provides a significant contribution to the explanation of the dependent variable. Subsequently, further F-tests can be calculated for each effect of interest to check whether the considered effect is significant. For example, the F-Test for the direct effect of factor A on the dependent variable Y calculates as: F=
SS A /(G − 1) SS e /(K − G)
with G as number of groups and K as the sum of the number of observations in the individual groups (Homburg and Krohmer 2006). Further, the strength of the effect from the independent on the dependent variable can be measured with the quality measure η2 (Pierce et al. 2004). It ranges between 0 and 1, with high values indicating a strong effect on the dependent variable. In the final step of the interpretation of results, the F-values are interpreted first. The F-tests for the individual effects indicate what direct effects and what interaction effects apply on the dependent variable. Further analysis can be facilitated through graphical representation and the reporting of the mean values in the individual groups (Homburg and Kromer 2006). Beyond these measures, there are further post hoc tests to examine the significance of mean value differences.
102
3
Method
3.2.1.2 Post Hoc Tests A significant analysis of variance indicates differences between included groups regarding the effect on the dependent variable. However, in most cases it is essential to get more detailed results including pairwise comparisons between specific groups and to identify where exactly significant differences occur (Ruxton and Beauchamp 2008). The multiple application of t-tests to assess mean differences is not appropriate, as multiple testing leads to an alpha error accumulation, meaning that with an increasing number of hypothesis tests, the probability of false positive results increases (Brown 2005). This is called p-hacking (Bruns and Ionnidis 2016) and increases false positive rates and leads to results that are not replicable as they are just based on data randomness (Simmons et al. 2011). Therefore, literature identified a large variety of post hoc tests to conduct pairwise comparisons subsequent to an analysis of variance that adjusts results by correcting associated significance levels. There are three main criteria to consider when choosing a post hoc test: control for the type I error rate, control for the type II error rate and test robustness for violation of assumptions. Type I (α) errors refer to false positive results, where a true null hypothesis is mistakenly rejected. Conservative tests apply strict criteria leading to a low probability of type I errors. Type II (ß) errors describe the probability to fail rejecting a false null hypothesis. High probability of type II errors is associated with low statistical power. Therefore, the adjustment of a test is always a trade-off between being conservative enough to reduce type I errors and still having enough statistical power to detect existing group differences. Further, the robustness of a test states how it performs under the violations of assumptions such as deviations from normality, unequal group sizes and different group variances (Homack 2001). Although there is a large variety of different post hoc tests that were developed over the years, is worth to mention a few special post hoc tests that are widely relied on and have special characteristics (Field 2013): the LSD (least significant difference) test is also referred to as Fisher’s method (Fisher 1935) is the oldest method and does not control for the alpha errors and therefore it is too liberal in its original form. In contrast, Bonferroni and Tukey tests are both really conservative tests and therefore lack of statistical power. While the Bonferroni test has more statistical power for analyses with fewer pairwise comparisons (Toothaker 1993), the Tukey test (also called honestly significant differences HSD) offers more power for comparisons of many groups (Tukey 1953). Scheffé’s test (Scheffé 1953) is a little less conservative, although still one of the more conservative tests in the field with higher statistical power. It is especially applicable for supplemental analysis of comparisons that have not been defined in advance, as it provides
3.2 Data Analysis Methods
103
an F-value for each combination of means and is very sensitive for complex comparisons (Brown 2005). In case of doubts regarding equal variances across groups according to Levene’s test, it is recommended to rely on the Games-Howell procedure or to the more conservative Dunnett post hoc procedures (Scheffé 1953). For equality of variances, but unequal group sizes, it is recommended to apply Gabriel’s post hoc test for small differences of group size and the Hochberg GT2 test for large differences of group sizes. This thesis relies on analyses of variance followed by post hoc tests of Bonferroni and Scheffé. As group sizes were equal based on the equal assignment of participants to the experimental groups and Levene’s test did not raise doubts regarding equal variances across groups, there were no special requirements for post hoc tests with high robustness for the violation of these assumptions. Although considered a very conservative, Bonferroni’s test (Toothaker 1993) is still having enough statistical power to provide clear results, confirming proposed effects as significant, such as in study 3. In study 4, one of the proposed hypotheses could not be accepted, although relying on the Scheffé post hoc test (Scheffé 1953) with its higher statistical power.
3.2.2
Multiple Regression Analysis
Multiple regression analysis is one of the most widely applied multivariate analysis method in marketing and belongs to the multivariate dependency analysis methods (Cohen et al. 2003) based on metric scaled measures (Hair et al. 2010). While bivariate regression analyses just examine the effect of one independent variable on one dependent variable, multiple regression analysis analyzes the effect of various independent variables on a dependent variable (Homburg et al. 2008). Regression analysis is able to show the effects of several independent variables (such as type of service representative, robotic behavior, service appropriateness, and various expectation levels) on an outcome variable, such a customer response. Further, regression analysis can show moderation effects based on interaction effects (Aiken and West 1991). This thesis relies on multiple regression analysis in several studies, such as for the examination of moderating effects of robot anxiety (study 1) but also extending the multiple regression analysis by including higher-order polynomial terms and conducting a polynomial regression analysis (see section 3.2.3) in study 3. The process of the regression analysis contains four steps that are visualized in Figure 3.2 and described below.
104
3
Step 1:
Specification of the Model
Step 2:
Estimation of Parameters
Step 3:
Quality Assessment
Step 4:
Interpretation of Results
Method
Figure 3.2 Sequential Process of the Regression Analysis. (see Homburg and Krohmer 2006)
The first step of a regression analysis concerns the specification of the model. The basic multiple regression analysis expresses a relation between the independent variables xi and the dependent variable y that is based on a linear model of the following structure (Chatterjee and Price 1991): y = b0 +
n i=1
bi xi + e
with b0 as regression constant and bi as regression coefficients describing the effect of the corresponding independent variable xi on the dependent variable y. The model includes n independent variables that explain the dependent variable together with the remaining residual e, describing influences on y that are not included in the model. For the examination of moderating effects, the model has to be adjusted, as the relation between the independent variable x m and the dependent variable y is moderated by the value of the moderator z. Therefore, the basic regression model from above has to be extended by the direct effect of the moderating variable z and the interaction term between the moderator z and the interacting variable xm (Aiken and West 1991): y = b0 +
n i=1
bi xi + (bn+1 z + bn+2 xm z) + e
Step 2 considers the estimation of parameters, as they describe how strong every independent variable affects the dependent variable. The estimation of parameters is based on the principle of least squares (Cohen et al. 2003). They are estimated based on the values of k samples, minimizing the sum of squared errors between the empirical values of the dependent variable y j and the corresponding values generated from the regression model y j (Skiera and Albers 2008). This can be
3.2 Data Analysis Methods
105
described as follows: k
e2 j=1 i
=
k j=1
yj − y j
2
→ min
where y j represents the value of the dependent variable of the sample j and y j represents the estimation of the regression model for y j . Compared to the bivariate analysis, the estimation of regression parameters cannot be calculated manually for multiple regression analyses, as it requires the application of numerical methods. As the resulting regression coefficients b j cannot be directly compared due to different underlying scales, results are often reported relying on the standardized regression coefficients β j (Skiera and Albers 2008). Step 3 of the process relates to the quality assessment of the results. There are global quality measures to assess quality of the whole regression model, as well as quality measures to test the significance of each regression coefficient. The global quality of a regression model is the coefficient of determination R2 . It describes the amount of variance of the independent variable that the regression model explains with the independent variables (Hair et al. 2006). Therefore, it is calculated as:
k R = 2
j=1 y j k j=1 y j
−y −y
2 2
with y as mean value of the dependent variable y, ranging from 0 to 1, with high values indicating a higher amount of variance explained by the regression model (Skiera and Albers 2008). As the inclusion of further independent variables into the regression models increases the determination coefficient R2 , literature 2 established a way to adjust the determination coefficient R based on the degrees of freedom depending on the number of samples and the number of included regression coefficients (Pindyck and Rubenfeld 1991). Further, an F-test should be calculated to assess the quality of the regression model, indicating that R2 is significantly different from 0 (Schuchard-Ficher et al. 2013). An increase of the F-value also shows whether the inclusion of further independent variables in the regression model leads to a significant increase of explanatory power of the overall model (Cohen et al. 2003). The F-value is defined as following, based on the number of coefficients n and the number of samples k (Skiera and Albers 2008):
106
3
Method
R 2 /n F= 1 − R 2 /(k − n − 1) Besides the global quality criteria referring to the whole regression model, the significance of each regression coefficient b j is important for the further interpretation of the results. This test for significance is conducted as t-test based on k − n − 1 degrees of freedom (Homburg and Kromer 2006). The corresponding null hypothesis tests, whether the regression coefficient is significantly different from 0. Finally, quality assessment includes to make sure, that the preconditions of regression analyses are met, such as having no multicolinearity among the independent variables, having residuals with normal distribution, an expected value of 0, a constant variance (homoscedasticity), and not being correlated (no autocorrelation) as shown in Schuchard-Ficher et al. (2013). The interpretation of results is the final step of a regression analysis. The level of the regression coefficients is the most relevant measurement for further interpretation, as long as all quality criteria from the previous paragraphs are met. A regression coefficient bi can be interpreted as the increase of the dependent variable y, as the independent variable xi increases by one unit. In order to compare the regression coefficients, it is essential to standardize them first and compare the adjusted regression coefficients βi to receive values that are not affected by scale differences (Skiera and Albers 2008).
3.2.3
Polynomial Modeling and Surface Response Method
Study 3 examines the effects of disconfirmation between expected and perceived robotic behavior (see section 1.3) based on expectation disconfirmation theory (see section 2.3.1.2). This thesis follows the advice of Venkatesh and Goyal (2010), relying on polynomial modeling in combination with surface response analysis. The following sections give an overview on the basics of this approach (section 3.2.3.1), describe the sequential process (3.2.3.2) and point out the benefits of applying polynomial modeling in combination with surface response analysis.
3.2.3.1 Basics of the Surface Response Method The surface response approach is based on prior polynomial modeling. Therefore, a polynomial regression is conducted first (Shanock et al. 2010). Instead of analyzing the results of the polynomial regression directly, the surface response
3.2 Data Analysis Methods
107
method subsequently examines the response pattern that can be diagrammed as visual three-dimensional representation (Harris et al. 2008). Polynomial Modeling is predestined to examine linear and non-linear disconfirmation effects on an outcome variable (Edwards and Harrison 1993) and is recently favored for the examinations of the expectation disconfirmation paradigm (Venkatesh and Goyal 2010). As mentioned in section 2.3.1.2, disconfirmation contains two individual component measures (e.g. ex ante expectations and ex post experiences). Polynomial modeling enables non-linear analyses, as well as complex relationships between two component measures and one outcome variable by taking different levels of disconfirmation into account (Edwards and Harrison 1993). Therefore, this method is able to provide a more accurate description of the relationship of X as predictor 1 (e.g. expected ISB) and Y as predictor 2 (e.g. perceived ISB) on the dependent variable Z (e.g. customer delight). Polynomial modeling requires a hierarchical analysis starting with a linear regression, analyzing the effect of X and Y on the outcome Z. Subsequently, higher order terms such as the squared terms of X2 and Y2 and the interaction term XY for quadratic relationships. The hierarchical analysis continues to include higherorder polynoms, as long as the variance explained by the next highest order (e.g. cubic terms) is significant. Surface Response Method (SRM) builds upon the results of polynomial modeling by offering a variety of visual interpretations that are supported by statistical tests (Venkatesh and Goyal 2010). It allows a graphical interpretation of how the two component measures describe the surface of the outcome and further statistical tests are able to identify the effects as significant (Edwards 2002). Surface response analysis especially focuses on potential stationary points and corresponding principal axes, on surface curvatures (convex versus concave), as well as on the slope and curvature along lines of special interest (Edwards and Parry 1993). Two important lines of interest are the line of perfect agreement (where X = Y) and the line of incongruence (where X = −Y). Special characteristics of these lines show how agreement affects the outcome (Shanock et al. 2010), depending on the level of agreement (slope for X = Y), how the degree of discrepancy affects the outcome and how direction matters (slope for X = Y) and have the potential to detect non-linear effects (curvature along the lines of interest).
3.2.3.2 Sequential Process of the Surface Response Method The approach of polynomial modeling in combination with surface response analysis is conducted in several steps, as identified from Shanock et al. (2010). Figure 3.3 gives an overview about the required steps.
108
3
Step 1:
Determination of Appropriateness
Step 2:
Check for Occurrence of Discrepancies
Step 3:
Conducting a Polynomial Regression
Step 4:
Calculation of Surface Values
Step 5:
Compilation of the Graph
Step 6:
Interpretation of Surface Values and the Graph
Method
Figure 3.3 Sequential Process of the Surface Response Method. (see Shanock et al. 2010)
The determination of appropriateness of the underlying research design for the surface response analysis is the first step. It has to be ensured, that the two independent component measures are measured on the same numeric scale and fulfill the assumptions of multiple regression analysis. Further, it has to be determined if the research question concerns the effects of two independent component measures on one outcome variable. Exemplary research questions examine how agreement, the degree or the direction of discrepancy relate to an outcome, such as person-environment fit (Edwards and Parry 1993), external versus self-perception (Gibson et al. 2009), and pre-exposure versus post exposure data (Venkatesh and Goyal 2010). In step 2 it is essential to check for occurrence of discrepancies. An overview about the ratio of discrepancies and their direction is needed to identify the practical value of potential SRM results. In case of no discrepancies or unilateral discrepancies in one direction, the results will be much less meaningful. According to Fleenor et al. (1996), standardized variables of predictor 1 that differ more than half a standard deviation from predictor 2 can be considered as discrepant. The percentage of values in agreement and discrepant values in either direction should be checked to increase transparency and to support following result interpretations (Shanock et al. 2010). As the requirements of the previous two steps are fulfilled, the next step includes conducting a polynomial regression. In advance, the two predicting values should be centered around the midpoint of the scale (Edwards 1994), to facilitate interpretation and reduce potential multicollinearity (Aiken and West 1991). Then,
3.2 Data Analysis Methods
109
higher order terms such as the quadratic terms (X2 and Y2 ) and the interaction term (XY) and possibly higher-order terms (see section 3.2.3.1) are calculated and the polynomial regression is conducted. Exemplary, the according quadratic polynomial equation is: Z = b 0 + b1 X + b 2 Y + b3 X 2 + b4 X Y + b5 Y 2 + e with Z as the dependent variable, X as predictor 1 and Y as predictor 2. The results of the polynomial regression analysis will not be examined as usual, but rather used for the calculation of surface test values. The calculation of surface test values a1 , a2 , a3 and a4 from the unstandardized beta coefficients for the centered predictor variables is needed for further statistical analysis of the surface response method: a1 = (b1 + b2 ); a2 = (b3 + b4 + b5 ); a3 = (b1 − b2 ); a4 = (b3 − b4 + b5 ) where a1 describes the slope of the line of perfect agreement (X = Y), a2 describes the curvature along the same line, a3 describes the slope of the line of incongruence (X = −Y), and a4 describes the curvature along the same line. Further, the significance tests for the surface values can be calculated as follows: ta1 =
ta2 =
S E 2b1 + S E 2b2 + 2cov b1 b2
a2
S E 2b3 + S E 2b4 + S E 2b5 + 2cov b3 b4 + 2cov b4 b5 + 2cov b3 b5 ta3 =
ta4 =
a1
a3
S E 2b1 + S E 2b2 − 2cov b1 b2
a4
S E 2b3 + S E 2b4 + S E 2b5 − 2cov b3 b4 + 2cov b3 b5 − 2cov b4 b5
110
3
Method
Step 5 concerns the compilation of the graph to facilitate and enhance possible interpretation of the results on a visual basis. Therefore, the response surface is plotted as three-dimensional surface, based on the results of the polynomial regression. The final step is the interpretation of surface values and the graph. The interpretation depends on the underlying research questions, but in most cases the focus lies on questions as how the agreement between the predictors relates to the outcome and how the degree, as well as the direction of discrepancy relates to the outcome. These questions can be addressed by analyzing the surface test values as described in Table 3.15.
3.2.3.3 Benefits of the Surface Response Method In combination with the surface response method, polynomial modeling offers various benefits compared to difference scores and compared to traditional moderated regression. Compared to difference scores, this approach provides much more differentiated interpretations. The combination of polynomial modeling with surface response analysis was introduced specifically to address the problems associated with the application of difference scores (Edwards 1994). Difference scores are calculated simply through subtracting one measure from the other. The reduction from two individual measures to only one new measure leads to a significant loss of information. For example, the difference score is not able anymore to make a point regarding the contribution of the individual independent measures on the outcome, such as which of the factors have a stronger influence on the variance of the output. This is where the surface response method comes in and helps to reduce the problems of ambiguous interpretations of results, based on difference scores (Shanock et al. 2010). Difference scores are not able to examine many types of non-linear effects as precisely as SRM and the effects of the level of agreement on the outcome can also not be measured. Several studies (Atwater et al. 1998; Edwards 2002) could show that the reduction of two dimensions to only one difference score dimension is associated with a significant loss of information, that is necessary for a precise interpretation, which is why surface response method is superior to difference scores. Another benefit of the surface response method becomes apparent compared to traditional moderated regression. This comparison has not received a lot of attention in literature, but is also worth to consider (Shanock et al. 2010). While traditional moderated regression analysis just delivers a two-dimensional picture of the underlying effects, surface response analysis enables a three-dimensional view on the relationship between two predictors and one outcome variable. Therefore, surface response analysis provides more detailed information on the
3.2 Data Analysis Methods
111
Table 3.15 Interpretation of Surface Test Values Surface Test Value
Interpretation
Effects along the line of perfect agreement (X = Y) a1 is significant
Linear slope: agreement between X and Y relates to Z in a linear way
- a1 is positive
Positive slope: Z increases as both X and Y become higher
- a1 is negative
Negative slope: Z decreases as both X and Y become lower
a2 is significant
Non-linear slope: agreement between X and Y relates to the Z in a non-linear way
- a2 is positive
Convex surface: Z increases more sharply as X and Y become higher
- a2 is negative
Concave surface: Z decreases more sharply as X and Y become lower
Effects along the line of incongruence (X = −Y) a3 is significant
Linear slope: discrepancy between X and Y relates to Z in a linear way
- a3 is positive
Positive slope: Z is higher when the discrepancy is (X > Y) rather than (Y > X)
- a3 is negative
Negative slope: Z is higher when the discrepancy is (Y > X) rather than (X > Y)
a4 is significant
Non-linear slope: discrepancy between X and Y relates to Z in a non-linear way
- a4 is positive
Convex surface: Z increases more sharply as the degree of discrepancy becomes higher
- a4 is negative
Concave surface: Z decreases more sharply as the degree of discrepancy becomes higher
relationships between the various combinations of the two predictors on the outcome. For example, the SRM gives detailed information how the agreement of the predictors affect the outcome, depending on the absolute level where they are in agreement. Traditional moderated regression results do not allow such detailed assessments of the relationships and might not discover non-linear relationships along the line of perfect agreement (Shanock et al. 2010). Further SRM is able to examine the effects of increasing discrepancy on the outcome that cannot be examined with two-dimensional interaction graphs of a moderated regression (Shanock et al. 2010).
4
Study 1: A Service Robot Acceptance Model: Customer Acceptance of Humanoid Robots During Service Encounters
This study theoretically develops and tests a Service Robot Acceptance Model (SRAM), relying on the confirmation-disconfirmation paradigm, role theory, and a qualitative study with 63 users. This model assumes that users intuitively rely on alternative references to assess the interaction with service robots. This in turn affects a user’s service robot acceptance that is defined as the extent to which the user perceives the interaction with a service robot positively in terms of its functional, informational and relation abilities. Our experimental laboratory study with 90 student participants and the humanoid robot of the type NAO shows that the acceptance of service robots depends on the reference category the user relies on prior to the interaction with a service robot. Furthermore, robot anxiety affects the relevance of different reference categories for service robot acceptance.
4.1
Introduction
The number of robots being used in businesses is augmenting rapidly and service robots are increasingly becoming part of our daily lives. According to the intended application, robots are classified into industrial robots and service robots. In contrast to industrial robots, a service robot is a “robot that performs useful tasks for humans or equipment excluding industrial automation” (ISO 8373). There are two categories of service robots. Personal service robots are used for non-commercial tasks usually by laypersons, whereas professional service robots Based on a publication in IEEE International Conference on Pervasive Computing and Communications (PERCOM) 2017 together with Stock, Ruth.
© The Author(s), under exclusive license to Springer Fachmedien Wiesbaden GmbH, part of Springer Nature 2021 M. Merkle, Humanoid Service Robots, Neue Perspektiven der marktorientierten Unternehmensführung, https://doi.org/10.1007/978-3-658-34440-5_4
113
114
4
Study 1: A Service Robot Acceptance Model
are used for commercial tasks operated by properly trained operators (ISO 8373), although “people who interact with [these] robots are increasingly unlikely to be technically trained experts and thus more likely to use casual intuitive approaches to the interaction” (Mathur and Reichling 2016). This study focuses on the second category of professional robots such as shop floor robots in retailing (Nestlé 2016) or receptionist robots in hotels (Rajesh 2015). Although service robots gain more and more importance in our daily life, research on user acceptance of these robots is scarce. This is surprising because “in order to introduce social robots successfully, we must first understand the underlying reasons whereupon potential users accept these robots” (De Graaf and Allouch 2013, p. 1476). Most studies examined the acceptance of robots in general. For example, extant research investigated the impact of demographic variables, such as interest in science, technological affinity, and robot experience on users’ robot acceptance (Arras and Cerqui 2005; Kuo et al. 2009; Nomura et al. 2006). To our knowledge, only three studies examine user acceptance of humanoid service robots in particular (De Graaf and Allouch 2013; Reich and Eyssel 2013). “In a laboratory-based user interaction study with 60 participants, De Graaf and Allouch (2013) apply the NAO robot”. The authors identify usefulness, adaptability, enjoyment, sociability, companionship, and perceived behavioral control as important drivers of user acceptance of service robots. The study by Reich and Eyssel (2013) test a three-dimensional model of users’ psychological anthropomorphism as the “tendency to imbue the real or imagined behavior or nonhuman agents with humanlike characteristics, motivations, intentions or emotions” (Epley et al. 2007, p. 864). They identify demographic and personality characteristics as important antecedents of this tendency. This proposed research contributes to the knowledge on human-robot interaction by adding insights into users’ acceptance of a service robot. In this study, we attempt to understand the intuitive mechanisms that are responsible for a user’s acceptance of a service robot, i.e., service robot acceptance. Extant research has shown that individual users intuitively rely on various reference categories to deal with new products or services provided by a company (McKinney et al. 2002). For example, users have been shown to rely on experiences with similar products or services (Oliver and DeSarbo 1988; Tse and Wilton 1988; Yi 1988). Accordingly, users’ acceptance of a service robot may depend on his or her prior experiences with similar or comparable situations. This study strives to provide insights into potential reference categories, responsible for service robot acceptance, and to answer the first research question: Which reference categories are responsible for users’ acceptance of a service robot? Service robot acceptance is defined as the extent to which the user perceives the interaction with a service robot positively in terms of its functional, informational and relation abilities; and is conceptualized
4.1 Introduction
115
as a three-dimensional construct, encompassing a functional, an informational, and a relational component. Consistent with extant literature on the technology acceptance model (TAM) (Solomon et al. 1985), we conceptualize functional service robot acceptance as the extent to which a service robot is perceived as useful and easy to use by a user. Beyond the well-known functional component (Davis 1989), we further add an informational and a relational component. This is because “robots are no longer merely features of our technological environment, but are beginning to penetrate our social sphere” (Mathur and Reichling 2016, p. 22). Rather, in contrast to industrial robots, their purpose of a service robot is the delivery of information and interaction with users (Stock 2016). Informational service robot acceptance captures the extent to which users consider the information, provided by a service robot as being rich in terms of quantity and quality. Relational service robot acceptance is the extent to which a user feels understood by and trusts a service robot. To examine this question, we rely on the confirmation-disconfirmation (C/D) paradigm, role theory (Solomon et al. 1985) and the technology acceptance model (Davis 1989). Furthermore, it is likely that acceptance of service robots by users depends on contingency factors. Accordingly, we attempt to make a first step toward a better understanding of factors that affect the relationship between a user’s reference in terms of a service robot and the various dimensions of service robot acceptance. Robotic research so far argues that robot anxiety is pivotal for a user’s perception of robots in general (Nomura et al. 2004; Nomura et al. 2006; Nomura et al. 2008; Nomura et al. 2012). In this study, we focus on robot anxiety as contingency factor to more deeply understand the impact of a user’s reference on service robot acceptance. Robot anxiety refers to “emotions of anxiety or fear preventing individuals from interaction with robots having functions of communication in daily life, in particular, communication in a human-robot dyad” (Nomura et al. 2006, p. 372). Accordingly, the second research question is: How does robot anxiety affect the relationship between a user’s reference categories in terms of a service robot and service robot acceptance? This research is based on a qualitative study with 63 participants, which was analyzed through content analysis. Furthermore, we tested the Service Robot Acceptance Model (SRAM) in a laboratory experiment with 90 participants.
116
4.2
4
Study 1: A Service Robot Acceptance Model
Theoretical Background
As antecedents of service robot acceptance, we rely on the expectation disconfirmation theory and role theory.
4.2.1
Expectation Disconfirmation Theory
As the basics of expectation disconfirmation theory and the underlying mechanisms have already been introduced in section 2.3.1.2, this section briefly applies the general theory to the research question of this study and links the theory to the proposed service robot acceptance model. The concept has recently been applied in IS research (McKinney et al. 2002) and predicts that consumers form different standards to which they compare a current performance of a company. This in turn will result in a confirmation (i.e., met comparison standard) or disconfirmation (i.e., a discrepancy between the comparison standard and the individual’s perceptions). Important implications of the expectation disconfirmation theory relate to the comparison standard. Extant literature distinguishes three comparison standards: ideal imaginations about a product and/or service, expectations, and experiences (Fournier and Mick 1999). While expectations relate to an anticipated performance level (Herrmann and Johnson 1999; Spreng et al. 1996), experiences are based on a user’s experiences with the same or similar product or service (Cadotte et al. 1987). An ideal standard reflects the user’s imagination of an ideal product and/or service. Based on this theory, we predict that users are likely to compare their perceived performance with three comparison standards, i.e., an ideal performance of the service robot, their expectations toward a service robot’s performance, and prior experiences during similar situations, i.e., service encounters. The outcome of the comparison process is likely to affect the user’s acceptance of the service robot.
4.2.2
Role Theory
Another theoretical basis for exploring how customers assess the interaction with a humanoid robot is role theory (Solomon et al. 1985). The basic principle of this theory has been introduced in section 2.3.2.2 and the following section focuses on the application of role theory to the context of the service robot acceptance model.
4.3 Qualitative Study on Social Robot Acceptance (Study I)
117
In service settings, a commitment to an effective role performance could incorporate that giving good service will matter. Specifically, a customer’s a priori expectations regarding a service could rely on the following scripts: On the one hand, the user can rely on his or her own expectations towards an ideal service (Solomon et al. 1985). On the other hand, a user could rely on the expected complementary behavior of the other party. Role theory provides important insights to more deeply understand the underlying mechanisms of user acceptance of service robots during the service encounter. Specifically, users are likely to form their perceived script toward a service encounter with a robot based on two aspects, i.e., the expectations toward an ideal service provided by a service robot and the expected complementary behavior of the service representative. To understand a customer’s acceptance of humanoid robots, we compare the ideal services with a customer’s expectations toward a human service employee and a self-service technology (SST). We choose these role models because currently, personal services are mostly provided by human service employees (Stock and Bednarek 2014) or SSTs (Bolton and Saxena-Iyer 2009).
4.3
Qualitative Study on Social Robot Acceptance (Study I)
Research on underlying psychological mechanisms of users’ robot acceptance is scarce. To investigate this phenomenon, we therefore conducted a preliminary qualitative study. The purpose of this qualitative study was to more deeply understand the three theoretically identified expectation categories by customers during the human-robot interaction. Furthermore, we attempted to identify other potential categories.
4.3.1
Semi-Structured Interviews
With qualitative interviews, we gained deeper insight into various expectations by users toward the frontline service representative at the service encounter during the course of conversation (Graebner and Eisenhardt 2004). One researcher interviewed 62 participants for 15 min on average. To ensure some standardization across interviews, we used an interview guide with standard, open-ended questions for all respondents (Jehn 1997). We also allowed idiosyncratic questions if necessary, for clarification and added details.
118
4
Study 1: A Service Robot Acceptance Model
The interviews started with a general question about whether the person has any experience or imagination about humanoid robots. A more general question aims at providing the basis for more sensitive questions (Kyale 1996; Morse and Richards 2002). These subsequent questions were sufficiently focused but also allowed the participants to present their perspectives without being forced into a specific answer (Booms and Nyquist 1981; McMurry 1981). Information about the specific expectation category was elicited with two focused questions: “Imagine you would get a service from a service robot, e.g., in a hotel, restaurant, or at the airport: Which reference or experience would you rely on to figure out how to behave toward the robot or what to expect from the robot during the service encounter?” and “Why did you choose this particular reference or experience you mentioned?”
4.3.2
Sample and Analytical Strategy
The participants were between 19 and 76 years of age, earned between $18,000 and $140,000 annually, and came from various sectors, such as banking/insurance, transportation, IT, and public sector. To determine the empirical relevance of the investigated scripts, we quantitatively assessed the answers to the question about what the participants guessed when they thought about getting a service from a service robot; each time a respondent mentioned an issue, we counted it. We determined the number of respondents who mentioned a construct at least once during the interview and the frequency with which each construct was mentioned across all interviews.
4.3.3
Results
The results of the semi-structured interviews indicated that the participants rely on various alternative expectations that represent the basis for the acceptance of the service by a service robot. They most frequently referred to a human service employee as mental reference, followed by an SST. The respondents who referred to a human service employee as reference cited that this was their so far most present experience of a service delivery. Respondents also mentioned SST as reference point to form their expectations toward a service robot. Mostly, they believed that the humanoid service robot would not be different from any computer, programmed by humans. A third group of participants referred to their own goals they wanted to achieve or the ideal service. In sum, we could confirm the
4.4 Testing the Social Robot Acceptance Model (Study II)
119
theoretically derived expectation three expectation categories, i.e., the imagination about an ideal service from a service robot (14 participants), a service employee (22 participants), and an SST (30 participants). Several participants mentioned multiple, overlapping expectation categories as basis for their a priori thoughts about a potential service interaction with a service robot.
4.4
Testing the Social Robot Acceptance Model (Study II)
4.4.1
Basic Effects
In Figure 4.1, we depict our model of users’ acceptance of service robots. Role theory and our qualitative data indicate that users largely rely on their own ideal standard for a service and the experiences of complementary behaviors during the service encounter (i.e., by an SST or a human service employee). Interestingly, already five decades ago, researchers indicated that a service script as perceived by a customer could range from those of an equal partner vis-à-vis the customer to those of a virtual automation (Solomon et al. 1985). Finally, a customer’s expectations toward a service can be based on an imagined or learned conception of the prototypical service experience (Davis 1989; Venkatesh et al. 2003). To conceptualize a SRAM, we rely on the well-established TAM, which focuses on the cognitive evaluation of technologies (Davis 1989; Fishbein and Ajzen 1975). According to the TAM, a customer’s adoption of a new technology depends on its perceived usefulness and its perceived ease of use (Malhotra et al. 2008). Perceived usefulness has been adapted from the construct “usefulness” (Davis 1989, Gerlach et al. 2014) and is defined as the extent to which a user perceives that the service robot or the service employee contributes to the fulfillment of his or her needs. Perceived ease of use is based on the construct from the TAM (Malhotra et al. 2008). While the TAM primarily focuses on functional aspects of new technology usage, we also attempt to capture the “informational and relational side” of robots during the service encounter. The informational component is captured as informativeness, defined as the extent to which the robot provides helpful information that supports the customer to achieve his or her goals. The relational component captures benevolence and understanding. Benevolence is a user’s belief that the robot is interested in the customer’s welfare and will not take unexpected actions that would have a negative impact on the user (Kumar et al. 1995). Understanding is the extent to which a service robot understands a customer’s needs and expresses emotions of understanding.
120
4
Study 1: A Service Robot Acceptance Model
Customers‘ Role Expectations
Robot Acceptance
Functional Component Ideal Standard
Robot Anxiety as Moderator
Ideal Service by Social Robot
Ease of use
Usefulness Informational Component
Other party Self Service Technology
Frontline Service Employee
Informativeness of interaction Relational Component Benevolence
Understanding
Figure 4.1 Service Robot Acceptance Model (SRAM) During Service Encounter
4.4.2
Robot Anxiety as Moderator
The effect of different role expectations on a customer’s service robot acceptance may also depend on a customer’s attitudes toward robots. For example, a customer’s robot anxiety has been argued to affect the interaction with a humanoid robot (Nomura et al. 2004) and its acceptance (Nomura et al. 2008). Generally, “anxiety is defined as a feeling of mingled dread and apprehension about the future without a specific cause for the fear” (Nomura et al. 2008). In this study, we focus on state anxiety as anxiety, derived from specific situations. Accordingly, robot anxiety is defined as a strong overwhelming fear, deriving from the interaction with a robot. In this study, we argue that robot anxiety may weaken the acceptance of a service robot compared with a human service employee or an SST. Specifically, we
4.5 Experimental Study
121
examine robot anxiety as moderator of the customer expectation-robot acceptance relationship. The tested SRAM including robot anxiety as moderator is depicted in Figure 4.1.
4.5
Experimental Study
4.5.1
Experimental Setting
The mechanical basis for the robot for the second study was the NAO Academic Edition. The 57 cm high humanoid robot is autonomous, programmable with 25 degrees of freedom, and produced by Aldebaran. The NAO has several technical features that make the interaction with the robot appear human-like: For expression, it has a voice synthesizer, LED lights, and two speakers. Furthermore, the robot has a series of sensors, such as four microphones, two cameras, a sonar distance sensor, two IR emitters and receivers, nine tactile sensors, and eight pressure sensors. We used the NAO for two reasons. First, earlier robotics research has shown that people tend to interact naturally with this type of robot because of its familiarity (Fong et al. 2003). Second, the NAO has been most widely used for academic purposes (De Graaf and Allouch 2013). For example, previous research has used the NAO in various human robot interaction settings, such as simulated body movements (Xu et al. 2014), storytelling and conversation (Csapo et al. 2012; Han et al. 2012), smart home interactions with human beings (Louloudi et al. 2010) and emotion transitions for therapeutic treatments (Miskam et al. 2014). Our experiment was carried out in a research lab associated with the authors’ university. A room in the lab was outfitted to resemble a “hotel lobby”, intended to welcome incoming guests to check-in. The room was stocked with furnishings and objects likely to be familiar to test subjects and appropriate for a hotel lobby—e.g., a reception desk, a seating group, a bookshelf with books, and small shelf with information about the city near the reception desk. Figure 4.2 depicts the experimental laboratory setting. A separate room, with no view of the experimental setting, was used for pre and post-session surveys with research subjects. At the beginning of the experiment, the subjects had to wait for 30 seconds and were told that the service representative had to finish another process and they had to wait for a minute. The purpose was to get the participant aware of the setting and to adjust to the situation. Furthermore, initial mood stages could be captured based on video recordings as described in more detail below. For all
122
4
Study 1: A Service Robot Acceptance Model
sessions, the experimental room was kept free from external sounds, and room lighting and room temperature were maintained at normal residential levels.
Notes: left side: operator controlling the robot behind a curtain; right side: human-robot interaction
Figure 4.2 Experimental Setting of the Study
During the experiment, each participant in the role of a customer had to interact separately with the service robot or the confederate in the role of a service employee. All visual displays and sounds were recorded by external HD cameras, positioned in the room and on Nao’s body. The experimenter was not visible to the participants; at a hidden station, the experimenter was observing video streams from the cameras (Figure 4.2). We applied the Wizard of Oz design (Gould et al. 1983; Kelley 1984) that has been widely used to study human-robot interaction (Dahlbäck et al. 1993; Xu et al. 2014). Accordingly, the participants were told that the service robot acted autonomously, whereas the robot was directed by the operator behind a curtain (Hennig-Thurau et al. 2006).
4.5.2
Experimental Design
For manipulation purposes, the three different comparison references, i.e., ideal, SST, service employee, were tested for three different groups in a between-subject design. In the first setting, the participants only interacted with a service robot,
4.5 Experimental Study
123
the second group dealt with an SST before interacting with a service robot, and the third group interacted with a service employee prior to the interaction with a service robot (Figure 4.3).
Notes: left: human-SST interaction; right: human-robot interaction
Figure 4.3 Check-in Experiences Prior to Human-Robot Interaction
To avoid learning effects, we applied a between subject design (Atzmüller and Steiner 2010). Before entering the hotel lobby, the subjects were requested to check in at a hotel they previously booked in Boston upon entering the hotel lobby, with the help of a service robot or a service employee. The instructions differed with respect to the service representative (service robot, human service employee, or SST) and the level of induced innovative service behavior (neutral vs. high) by the service representative and subjects. – Condition (1) contained the interaction with a service robot (ideal service expectations treatment). – Condition (2) contained two interactions, first with a service employee and afterwards with a service robot (human service employee treatment). – Condition (3) contained two interactions, first with an SST and afterwards with a service robot (SST treatment). To ensure that each subject understood his or her assigned task after reading the instructions, they were asked to write down aforementioned tasks. In case of any
124
4
Study 1: A Service Robot Acceptance Model
issues or differences, the experimenter corrected their understanding, and asked them to again provide a written summary of their task to double check that the instructions were properly understood.
4.5.3
Experimental Subjects and Confederates
Participants in the experiment were undergraduate and graduate students from a medium-sized university. The final sample contained data from 90 participants, 41.1% of whom were women. The mean age of the respondents was 26.0 years (SD = 8.3); ages range from 15 to 59 years. Although the use of student samples is sometimes considered a limitation in marketing research, drawing on student samples in experimental designs, especially those involving role playing, is well accepted for examining causal relationships (Barsade 2002; Hennig-Thurau et al. 2006). As the focus of the study was on customer acceptance of service robots, students were seen as customers. As an incentive, all participants got a financial reimbursement of $11 for completing the study. Additionally, they could join in a lottery for three Amazon coupons of $500, $200, and $100, respectively. To increase the level of immersion, intensively trained actors were used as confederates to play the roles of service employees in the experiment. To reduce any confounding effects due to their personal communication style and to standardize the interaction with the company, standardized service scripts were used (the detailed scripts can be requested from the authors), similar to those commonly used for hotels in business practice. To increase the realism of the experiments, it was referred to an existing service offer by a hotel. The service representative checked in the participant as a customer, according to a standardized script.
4.5.4
Data Collection and Analysis Methods
Recall that the primary purpose of this experiment was to identify and record data related to: (1) subjects’ (in the role of a customer) perceptions and responses to a service robot’s behavioral cues during the service encounter and (2) comparing this with the customer expectations toward an SST or a service employee. Data for participants in all three experimental conditions were collected both immediately prior to and immediately following subjects’ 10-minute experiences in the hotel setting.
4.6 Results
125
We relied on self-ratings by the subjects to assess the items for the service robot acceptance, i.e., functional component (ease of use, usefulness), informational component (informativeness of interaction), and relational component (benevolence, understanding). To assess customer satisfaction, a three-item scale was adapted from Homburg and Stock (2004). The three-item scale for customer delight was inspired by Finn (2005) and Paul (2000). Service robots’/service employees’ innovative service behavior was assessed with a four-item-scales used by Stock et al. (2017). All measures meet the required quality criteria as shown in section 3.1.2.
4.6
Results
The first step to analyze the experimental data was the manipulation check (Table 4.1). The assessment whether the three different treatment groups, i.e., interaction with the SFR only (group G0), interaction with a human service employee prior to the robot interaction (group G1), and interaction with an SST prior to the robot interaction (group G2) was assessed based on t-test.
Table 4.1 Manipulation Check: t-Test for Mean Differences Reference
G0 ideal services as role expectations
G1 human service employee as role expectation
G2 SST as role expectation
Mean difference (G0 –G1 )
Mean difference (G0 –G2 )
Ideal expectation1
.88
.77
.78
SST1
.48
.44
.73
2)
−.24*
Human1
.63
.84
.63
−.21*
2)
.11*
.16*
Note: 1) Measured as ranking; 2) Reference group: ideal services as role expectations; n(G0 ) = 30; n(G1 ) = 30; n(G2 ) = 30; SST = self-service technology; * p ≤ .01.
Results show that all groups G0, G1 and G2 could be successfully manipulated. Group G0 shows significantly higher mean values on “ideal” as expectation category, compared with group G1 and group G2. Furthermore, group G1 scored significantly higher on service employees as expectation category than group G0 and group G2 show significantly higher mean reference values on SST than group
126
4
Study 1: A Service Robot Acceptance Model
G0. Thus, our manipulation of group G1 to frame a service employee as expectation category toward the service robot was successful. Likewise, group G2 was manipulated successfully as this group’s main reference level was the SST. Table 4.2 shows the t-test results for the three components of robot acceptance, i.e., functional, informational, and relational. The mean values for usefulness (M = 5.18, group G0; M = 5.16, group G1; M = 5.35, group G2) were not significantly different. Interestingly, the functional component, derived from the TAM, does not depend on the user’s expectation category. In other words, functional component of robot acceptance is not affected by a prior interaction with the SST or the service employee. In contrast, significant differences for the informational and relational component can be found. For example, informativeness scored significantly lower for group G1 (M = 5.67) compared with group G0 (M = 6.28). Thus, user acceptance of the informational component was significantly lower for group G1 , i.e., the participants that interacted with the service employee prior to the service robot interaction.
Table 4.2 Service Robot Acceptance: t-Test for Mean Differences in User Perception Reference
G0 ideal services as role expectations
G1 human service employee as role expectation
G2 SST as role expectation
Mean difference (G0 –G1 )
Mean difference (G0 –G2 )
(a) Functional Component Ease of use1
5.39
4.93
4.98
.46
.41
Usefulness1
5.18
5.16
5.35
.022
−.17
5.67
6.13
.61*
.14
(b) Informational Component Informativeness of interaction1
6.28
(c) Relational Component Benevolence1
5.09
4.10
4.35
.99*
.74*
Understanding1
4.52
3.62
4.12
.90*
.40
Note: 1) Measured with a 7-point Likert scale: 1 = strongly disagree, 7 = strongly agree; n(G0 ) = 30; n(G1 ) = 30; n(G2 ) = 30; SST = self-service technology; * p ≤ .05.
Regarding benevolence as part of the relational component, group G1 (M = 4.10) and group G2 (M = 4.35) both rated significantly lower compared to group
4.6 Results
127
G0 (M = 5.09). This means that participants who interacted with either the service employee or the SST prior to the interaction with the service robot both perceived the service robot as less benevolent than the group G0 interacting directly with the service robot, comparing it mostly with their own expectations as reference. Table 4.3 depicts the moderating effects of robot anxiety on the three components of robot anxiety. The power of the moderating effects varied strongly between group G1 and group G2 each compared to the group G0 that interacted directly with the service robot. For participants that interacted with an SST prior to the service robot interaction there was a strong moderating effect of robot anxiety on the functional component (ease of use, usefulness) of service robot acceptance. In this case, robot anxiety weakened the basic effect on the functional component. By contrast the participants interacting with a service employee prior to the service robot interaction, robot anxiety had a negative moderating effect on the informational component compared to the group interacting directly with the robot without a prior treatment.
Table 4.3 Moderating Effects: Linear Regression with Robot Anxiety as a Moderator Manipulation Service Employee1
SST1
Ease of use1
β = −.30
β = −4.01*
Usefulness1
β = −.27
β = −2.74*
β = −.63*
β = −.38
Benevolence1
β = −.36
β = .22
Understanding1
β = −.43
β = −1.89
(a) Functional Component
(b) Informational Component Informativeness of interaction1 (c) Relational Component
Note: 1)
Compared to the group G0 with ideal expectations; 2) Measured with a 7-point Likert scale: 1 = strongly disagree, 7 = strongly agree; N = 120; * p < .05.
These findings match with our assumption that robot anxiety may weaken the acceptance of a service robot compared with a human service employee or an SST.
128
4.7
4
Study 1: A Service Robot Acceptance Model
Discussion
In recent years, service companies increasingly started to apply service robots in their shop floors. However, so far we hardly have insights whether and when customers would accept service robots.
4.7.1
Research Implications
Information systems research has a long tradition in examining antecedents and underlying mechanisms of technology acceptance. In particular, research on the well-known TAM (Davis 1989) has suggested that a user’s technology acceptance encompasses the components ease of use and usability (Davis 1989). This study extends the components of the technology acceptance model by adding two additional components, i.e., the informational and the relational component. These components are important due to the human appearance of social robots, which makes information exchange and relationship building during the human-robot interaction more important than within a human-computer interaction. In terms of the antecedents of service robot acceptance, we rely on the confirmation-disconfirmation paradigm (Oliver and DeSarbo 1988; Tse and Wilton 1988; Yi 1988) and role theory (Solomon et al. 1985) to extract different categories of role perceptions or experience responsible for an individual’s acceptance of a service robot during the service encounter. Specifically, we compared customer expectations towards ideal services with individuals, which largely rely on their prior experiences with a service employee or an SST. Our results reveal that these role expectations or experiences are particularly important for the informational and the relational component of service robot acceptance, but not for the functional component. Thus, during a human-robot interaction other components of technology acceptance seem to matter for human-computer interactions. From a conceptual perspective, this research applies the confirmation disconfirmation paradigm and introduces them to literature on human-robot interaction. Concepts provide valuable insights on the acceptance of service robots by human users. Further research should more strongly rely on theories and concepts with roots in psychology or other disciplines to more deeply understand user responses to service robots. So far, extant research largely relies on laboratory experimental settings to examine user acceptance of robots in general. To our knowledge, this study is the first of type that examines user acceptance of social robots in real life hotel context. Particularly, our subjects in the experiment had to book a hotel room,
4.7 Discussion
129
supported by a service robot, a service employee or an SST. Further research could rely on our experimental design to further understand user acceptance of service robots in real life contexts, such as shop floors and check-in counters. To our knowledge, literature on robotics acceptance has essentially focused on human-robot interactions, but largely neglected human-human interactions in comparison. A major purpose of this study is to examine various comparison categories and their effects on user responses to service robots, including human-human interactions, human-computer interactions, and an ideal expectation standard of human users towards service robots.
4.7.2
Limitations and Areas of Future Research
This is the first of type study that examines technology acceptance in the context of the service robot-user dyad in a real life context. Therefore, to validate our results, different types of service robots will be used in several service scenarios with various participant groups besides our student sample. Finally, this study is restricted to an experimental setting. Future research could examine our or similar research questions in a natural setting after service robots have been more largely diffused in organizations.
5
Study 2: A Cross-Country Comparison of Attitudes toward Humanoid Service Robots
So far, researchers know very little about what people actually expect from humanoid robots during a human-robot interaction. Therefore, this study surveyed 610 non-experts from Germany (133), the US (174), and India (303) and asked them to rate the following attributes regarding humanoid robots: empathy, expertise, reliability, and trust. This paper develops hypotheses, connecting robot attributes to the four cultural dimensions suggested by Hofstede—individualism, masculinity versus femininity, power distance, and uncertainty avoidance. The results show, that India rates all the attributes the highest, and that Germany and the US rate all aspects rather similarly with the largest difference regarding reliability.
5.1
Introduction
Humanoid robots are becoming more popular and are used more and more as service robots in human-robot interactions. Therefore, we chose the customer interaction with a service robot as exemplary situation although our results refer to human-robot interaction in general. Ivanov, Webster, and Berezina (2017) give an overview of where service robots such as the humanoid robot Pepper are already applied today. Their examples include restaurants, hotels, theme and amusement parks, airports, and other public spaces. Their paper is a great example for the fact that facing humanoid robots will soon be unavoidable.
Based on a publication on 52nd Hawaii International Conference on System Sciences (HICSS) 2019 together with Homburg, Nadine.
© The Author(s), under exclusive license to Springer Fachmedien Wiesbaden GmbH, part of Springer Nature 2021 M. Merkle, Humanoid Service Robots, Neue Perspektiven der marktorientierten Unternehmensführung, https://doi.org/10.1007/978-3-658-34440-5_5
131
132
5
Study 2: A Cross-Country Comparison
When looking at current research, there are many studies that focus on the acceptance of humanoid robots. However, there is very little known about the expectations towards robots, which becomes more important with the increased application of humanoid service robots. In this paper, we surveyed what participants expect from humanoid robots in terms of expertise, the extent to which they trust a robot, how reliable they expect a robot to be, and how much empathy they expect from a robot. We chose these aspects as they were identified as important for human service employees by prior service literature (Homburg and Stock 2001, 2005). For future research, these results are important to take into consideration, as this study gives an overview about what users expect from humanoid robots. This might have an impact on the results of experimental studies, which will be further examined in the discussion. Furthermore, this study provides more detailed insights about specific cultures and their connection to technology. This is important for future research depending on which country the study will be conducted in. Moreover, this research cannot only be used for future research including humanoid robots and service encounters but for general human-robot interactions. For example, the attribution of trust is also important in terms of health care. With the decrease in the number of health care professionals (Super 2002), it is important to figure out, whether robots might be able to solve this problem. This is especially important in terms of trust. Broadbent, Stafford, and MacDonald (2009) for example, give an overview on “literature about human responses to healthcare robots”. This study could help future research in this area, especially when conducting research in specific countries. For example, when conducting robotic experiments in one of the countries surveyed, the results of this study could be used to explain further results in this area. There are already some studies about the acceptance of or attitude toward different robots in different countries, which are visualized in the literature review (Table 5.1). In these studies, acceptance and attitudes towards robots were the main outcome variables examined based on different factors. The literature review indicates that the US, European, and Asian countries account for a substantial part of overall robotic research. In terms of the countries considered in this study, literature already provides first insights into the impact of cultural differences as described subsequently. Li, Rau, and Li (2010) found that Germany scored the lowest on trust compared to China and Korea. Furthermore, Rau, Li, and Li (2009) found that more people in China would rather accept a recommendation from a robot than would people in Germany.
5.1 Introduction
133
Table 5.1 Literature Review about Robots in Different Countries Author/s (Year)
Countries (# Participants)
Examined Variables
Major Findings
Bartneck et al. (2005)
China (44), Germany (109), Japan (135), Mexico (21), Netherlands (41), UK (58), USA (59)
Attitudes towards interaction with robots, attitudes towards social influence of robots, attitudes towards emotions in interaction with robots
Interaction: Mexico highest, USA lowest; Social influence: China highest, USA lowest; Emotions: Japan highest, Mexico lowest
Broadbent, Stafford, and MacDonald (2009)
USA (N/A), Japan (N/A), France (N/A), Germany (N/A), Korea (N/A)
Acceptance and attitudes of robots in the healthcare sector
French more accepting than Germans; Japanese thought that humanoid robots are more human like; different roles for Japanese and Americans
Evers et al. (2008) USA (31), China (27)
Acceptance of choices (comparing humans and robots)
US: higher trust with both; China: more comfortable with both
Green, MacDorman, Ho, and Vasudevan (2008)
USA (479), Japan (237)
Attitude towards robots Both countries prefer depending on people over robots experience (USA more than Japan)
Kaplan (2004)
Japan, Western Countries
Review of Japanese and western culture influencing myths and novels regarding artificial beings
-
Li, Rau, and Li (2010)
China (36), Korea (36), Germany (36)
Robot appearance and task as factors, on robot’s likeability. Engagement with, trust in and satisfaction with the robot.
German: lowest on all 4 scales Chinese and Korean results rather similar, Korea lower trust
Ouwehand (2017)
Netherlands, Japan: comparative case analysis
The extent to which elderly are willing to accept robots into their lives
Thesis, that culture has an influence on the acceptance of social assistive robots (continued)
134
5
Study 2: A Cross-Country Comparison
Table 5.1 (continued) Author/s (Year)
Countries (# Participants)
Examined Variables
Major Findings
Rau, Li, and Li (2009)
China (16), Germany (16)
Effects of communication styles and cultures on accepting recommendations from a robot
Chinese participants would rather accept recommendations than German participants
Salem, Ziadee, and Sakr (2014)
English (44) and Arabic native speakers (48)
Acceptance and anthropomorphization of humanoid robots
Arabic native speakers more positive toward humanoid robots
In the US for example, MacDorman, Vasudevan, and Ho (2009), found that people from the US generally preferred people over robots. Evers et al. (2008) conducted an experimental study and found that US participants reported higher trust in robots and were more compliant with robotic assistants than Chinese participants were. To our knowledge, no study about the acceptance of robots in India has been conducted so far. The findings from this literature review will be further discussed in the hypotheses section. This study was conducted in Germany, the US, and India to take the different cultural aspects of these countries into account. Especially in India, a country with increasing economic power, there could be major potential for the use of humanoid robots. This aspect will be expanded in the discussion when taking the results and potential use of robots into account.
5.2
Theoretical Background
To be able to compare the culture of each country, the Cultural Dimensions Theory of Hofstede (1980; 2003) was chosen. Section 2.3.2.4 introduced the theory in detail, including its four dimensions to describe the culture of countries: Power distance, Individualism, Masculinity versus Femininity, and Uncertainty avoidance. Hofstede and Hofstede (2005) rated countries on a scale of 1–100 for each of these dimensions. The ratings for each country can be found at: https:// www.hofstede-insights.com. Next, we will define the four dimensions and build our hypotheses.
5.2 Theoretical Background
135
91
77 67
66
65 62
56 48
46 40
40 35
India
Germany Individualism
Masculinity
Power distance
United States Uncertainty avoidance
Figure 5.1 Hofstede’s Cultural Dimensions for India, Germany, and the US. (based on Hofstede 1980)
Power distance. “… that is, the extent to which the less powerful members of organizations and institutions (such as the family) accept and expect that power is distributed unequally.” (Hofstede and McCrae 2004, p. 62) Uncertainty avoidance. “… deals with a society’s tolerance for ambiguity. It indicates to what extent a culture programs its members to feel either uncomfortable or comfortable in unstructured situations.” (Hofstede and McCrae 2004, p. 62) Individualism. “… versus its opposite, collectivism, refers to the degree to which individuals are integrated into groups.” (Hofstede and McCrae 2004, p. 63) Masculinity. “… versus its opposite, femininity, refers to the distribution of emotional roles between the sexes, another fundamental problem for any society to which a range of solutions are found.” (Hofstede and McCrae 2004, p. 63) Especially interesting when looking at the rating for each country (Figure 5.1) are the differences in individualism in all three countries as well as the rather similar rating of masculinity with the maximum difference being 10.
136
5.3
5
Study 2: A Cross-Country Comparison
Hypotheses
In our study, we looked for mainly four variables: Empathy, Reliability, Expertise, and Trust. After referring to the definition of each, we will introduce our hypotheses, based on the Cultural Concept by Hofstede and on prior studies as shown in the literature review.
5.3.1
Empathy
Empathy is “the capacity to clearly project an interest in others and to obtain and reflect a reasonable complete and accurate sense of another’s thoughts, feelings, and experiences” (Bush et al. 2001). Prior research already studied the gender differences regarding empathy. Christov-Moore et al. (2014), for example show that there are “behavioral and neural differences in affective empathy between males and females.” Females tend to be more empathic than males (Hoffman 1977). Transferred to Hofstede’s cultural dimension of masculinity versus femininity of a society, we assume that cultures with a higher level of masculinity (and therefore low level of femininity) attribute lower empathy to a robot. H1: India (lower masculinity) has higher expectations toward a robot’s empathy during a human-robot interaction than Germany and the US.
5.3.2
Reliability
Reliability “is defined as the extent to which a salesperson assures that promises made to customers are met (Parasuraman et al. 1994) and that customer instructions are precisely followed” (Homburg and Stock 2005, p. 402). The higher the reliability of a service, the lower is the uncertainty about the reactions and behaviors of the service representative. Therefore, we suggest that reliability is connected to uncertainty avoidance, as low uncertainty avoidance means that a culture is more open toward humanoid robots. They might question the reliability less than a culture of high uncertainty avoidance. H2: Germany (highest uncertainty avoidance) has a lower score on reliability towards the robot during a human-robot interaction than the US and India (lower uncertainty avoidance).
5.3 Hypotheses
5.3.3
137
Expertise
Expertise “is defined as the presence of knowledge and ability to fulfill a task” (Homburg and Stock, p. 394). Cultures with a high level of uncertainty avoidance prefer to be on the safe side and expect guaranteed expertise. We suggest that these cultures with high uncertainty avoidance are more likely to question the expertise of new things that they have little experience with. Therefore, they would be rather skeptical when it comes to humanoid robots. This would be a similar phenomenon to the one described in the context of reliability. Comparable to the previous section, we assume that countries with high uncertainty avoidance would associate less expertise with the robot. H3: Germany (highest uncertainty avoidance) has a lower score on expertise towards the robot during a human-robot interaction than the US and India (lower uncertainty avoidance).
5.3.4
Trust
Definition: “most researchers agree that trust is a personal characteristic that refers to “a willingness to rely on an exchange partner in whom one has confidence” (Moorman et al. 1993, p. 82). We assume that trust is significantly based on uncertainty avoidance. Cultures that are generally more open towards new things will have an easier way of trusting them. Therefore, we assume that the higher the uncertainty avoidance the lower the trust and the lower the individualism the higher the trust. Furthermore, prior studies like Li, Rau, and Li (2010), which is also mentioned in our literature review, already examined the trust in robots compared to humans in Germany, Korea, and China. In this study, Germany scored the lowest on trust. This would support our hypotheses, as Germany with the highest uncertainty avoidance would have to score the lowest on trust. H4a: Germany (highest uncertainty avoidance) has a lower score on trust than the US and India (lower uncertainty avoidance). Finally, the higher the femininity (and the lower the masculinity) the higher the trust.
138
5
Study 2: A Cross-Country Comparison
H4b: India (lowest masculinity) has the highest score on trust towards the robot during a human-robot interaction.
5.4
Empirical Basis
5.4.1
Data Collection
To address our research objectives we conducted a cross-country survey study with data from the US, from Germany and from India. Data from the US and from Germany were collected with paper pencil questionnaires at public places such as shopping malls and train stations. For the Indian data, we relied on Amazon’s MTurk to find participants for our study. We further asked the participants to provide their city of residence within the questionnaire to make sure that the MTurk respondents were from India. Recent studies raised quality concerns about data gathered via MTurk (Ipeirotis et al. 2010). Therefore we included two control questions such as “please click on disagree if you read this question carefully” to check whether the participants were reading the questions carefully and giving conscientious answers (Finin et al. 2010). 256 out of 866 filled questionnaires had to be excluded from this study due to wrong answers to our control questions. However, many studies already relied on Amazon “MTurk as a potential mechanism for conducting research in psychology and other social sciences” (Buhrmester et al. 2011, p. 5), verifying demographic declarations (Rand 2012), validating psychometric properties of MTurk responses (Buhrmester et al. 2011) and were able to replicate classic paper pencil findings with MTurk (Horton et al. 2011; Suri and Watts 2011). The questionnaire showed a picture of the Pepper robot from Softbank Corp, to give participants an example of a service robot. This type of robot was chosen, as Pepper is already in widespread use in the service context (Softbank 2018). After the participants watched the picture, they were asked to provide demographic data and rate humanoid robots in the categories empathy, expertise, reliability and trust. The constructs are measured by multiple items that were adapted from service literature (Homburg and Stock 2005) to fit for a service robot. Empathy is assessed with the use of six items based on scales suggested in extant research (Davis 1983; Hogan and Hogan 1984; Parasuraman et al. 1991). The expertise of the robot is measured on an eight-item scale that was developed based on the scales of Behrman and Perreault (1982), Doney and Cannon (1997), and Homburg and Stock (2005). The constructs reliability (Parasuraman et al. 1988) and trust (Doney and
5.5 Results
139
Cannon 1997) were also adapted to fit for a robot. All these items were rated on a seven-point Likert sale from ‘strongly disagree’ to ‘strongly agree’.
5.4.2
Characteristics of the Sample
As current research provides evidence, that demographics as age (Ezer et al. 2009) and gender (Kuo et al. 2009) affect the acceptance of robots, we strived to reach a representative average population sample. Our sample of 303 participants includes 338 men and 272 women whose average age was 38.2 years ranging from the age of 7 until 95 years. Moreover, the sample represents a range of occupations and a variety of different experience levels with robots.
5.5
Results
First, we describe the findings for the examined countries in detail. Subsequently we compare the three countries with one another.
5.5.1
Findings for the Three Countries
Table 5.2 shows the mean differences among the robot attributes of the 174 US participants. The confidence in the expertise and the reliability of the robot scores significantly higher than for empathy and trust. The attributes for reliability and expertise are on the same level. Table 5.2 Mean Differences Among the Robot Attributes in the US 1
2
3
1 Empathy1
--
2 Expertise1
1.18*
--
3 Reliability1
1.16*
-.02
--
.66*
-.52*
-.50*
4
Trust1
Note: 1 Measured on a 7-point Likert scale; p < .05
Table 5.3 shows the mean differences among the robot attributes for the 133 German participants taking part in our study. German participants report high
140
5
Study 2: A Cross-Country Comparison
attributions of expertise whereas the scores for trust (M = -.67*) and empathy (M = -1.76*) are significantly lower. Table 5.3 Mean Differences Among the Robot Attributes in Germany 1
2
3
1 Empathy1
--
2 Expertise1
1.76*
--
3 Reliability1
1.52*
-.24*
--
4 Trust1
1.09*
-.67*
-.43*
Note:
1 Measured
on a 7-point Likert scale; p < .05
Robots reached high scores for all of the categories from our 303 Indian participants, although the attribute for the robots’ empathy scores slightly lower than the other attributes. The mean differences among the robot attributes in India are shown in Table 5.4. Table 5.4 Mean Differences Among the Robot Attributes in India 1
2
3
1 Empathy1
--
2 Expertise1
.26*
--
3 Reliability1
.28*
.02
--
4 Trust1
.26*
.00
-.02
Note:
1 Measured
5.5.2
on a 7-point Likert scale; p < .05
Comparison of the Countries
Comparing the responses of the Indian participants with those of US participants and German participants there is a clear trend: Indians attribute robots significantly (p < .05) higher values for empathy, expertise, reliability and trust (see Table 5.5). Participants from the US and Germany rate the robot on the same level regarding its expertise. German participants rate the robot slightly lower with respect
5.5 Results
141
to reliability and trust, whereas the biggest gap occurs regarding the evaluation of the robot’s empathy. US participants attribute the robot significantly higher values regarding empathy (M = 3.98) compared to the German participants (M = 3.39). Regarding empathy, expertise, reliability, and trust, the results from Germany and the US are rather similar while the results from India are higher in all four sections. (Table 5.5) The biggest difference occurs regarding empathy. While Indian respondents assume that robots have a rather high degree of empathy, German and US respondents are not of the opinion that robots have a high empathy. All three countries rated the robot high on expertise with India being a little higher than Germany and the US. Regarding reliability, Germany rated the robot the lowest and India the highest. However, all three countries think of the robot as rather reliable. While India shows high trust in robots, Germany and the US rated trust lower, with Germany trusting robots the least. These results clearly show that India as an increasing economic power generally rates robots higher that countries like Germany and the US. Therefore, there is a very high potential for robots in the Indian market, due to high trust and openness. The results also point out, that conducting robot studies in India is not the same as conducting a study in one of the other countries surveyed. One can suggest that experiments in India can be conducted more easily due to high trust while in Germany they can be rather difficult in comparison. This could also be due to factors like “the German fear”. Table 5.5 Comparison Between the Countries US1
GER2
IND3
Mean Difference US—GER
Mean Difference US—IND
Mean Difference GER—IND
1 Empathy4
3.98 (1.48)
3.39 (1.27)
5.22 (1.12)
.59*
-1.24*
-1.83*
2 Expertise4
5.16 (1.16)
5.15 (.92)
5.48 (.83)
.01
-.32*
-.33*
3 Reliability4
5.14 (1.57)
4.91 (1.13)
5.50 (.93)
.23
-.36*
-.59*
4 Trust4
4.64 (1.33)
4.48 (1.06)
5.48 (.88)
.16
-.84*
-1.00*
Note: 1 United States; 2 Germany; 3 India; 4 Measured on a 7-point Likert scale; *Mean Difference is significant at p < .05; N(US|GER|IND) = (174|133|303).
142
5.5.3
5
Study 2: A Cross-Country Comparison
Connection to Hofstede
5.5.3.1 Empathy Regarding empathy, H1 is supported. Compared to the other surveyed countries, India has the lowest masculinity (56) and the highest rate for empathy (M = 5.22). In addition, Germany, with the highest masculinity (66), rates lowest on empathy (M = 3.39). However, because the difference in masculinity between India, Germany, and the US is rather small compared to their difference in empathy, we assume that there are other major factors influencing empathy.
5.5.3.2 Reliability Regarding reliability, H2 was supported. Germany rates overall high, in spite of a high score in uncertainty avoidance (65) and in line with our hypothesis it rates the lowest on reliability (M = 4.91) from these three countries.
5.5.3.3 Expertise For expertise, H3 was also supported. However, the difference between the ratings is rather small (M = .33) while the difference between the ratings in uncertainty avoidance is rather large ( = 25) (especially between India and Germany). We therefore suggest that even though uncertainty avoidance is a factor, it influences the rating of expertise rather little.
5.5.3.4 Trust In Hypotheses H4a and H4b we proposed that uncertainty avoidance and masculinity have a negative effect on trust in the robot. Both hypotheses were supported, as Germany with the highest uncertainty avoidance (65) shows the lowest score of trust in the robot (M = 4.48) and India with the lowest masculinity (56) has the significantly (p < .05) highest trust (M = 5.48) in robots. Overall, all of our hypotheses were supported. However, there are several surprising aspects, for example, to which extend the aspects were rated in spite of the Hofstede dimensions.
5.6
Discussion
5.6.1
Research Implications
The starting point was relating Hofstede’s Cultural Concept to the aspects surveyed about humanoid robots through our hypotheses. The major findings were
5.6 Discussion
143
that India scored the highest on all four aspects, while the US and Germany rate the aspects lower and rather similar. This study is one of several studies (Table 5.1) to research robot acceptance in different cultures. To our knowledge, this is the second research that applies Hofstede’s Cultural Concept with roots in management to the field of robotics. There was already a thesis by Anouk Ouwehand (2017) to compare the acceptance for social assertive robots in the Netherlands compared to Japan. However, this paper used a comparative case analysis. This research reveals that it also provides valuable insights for the understanding of cultural differences in the perception of robots. So far, extant literature essentially relies on plausibility considerations. Future research could rely more intensively on managerial culture approaches, such as Hofstede (1984), GLOBE (House et al. 2004), or Trompenaars’ approach (Trompenaars and Woolliams 2002). The results of this study are especially important for future research regarding all kinds of humanoid robot studies done in these countries. Especially interesting are the results for India, as it shows the potential for using robots. Furthermore, this study shows that studies done in one country are not necessarily representative for another country. Here, from what can be seen from the results of this study, Germany and the US are rather comparable, while India is not. When surveying these aspects, the results show that the cultural concept by Hofstede can be used as a starting point for hypotheses. As robots develop it can be assumed that more aspects will be surveyed in the future about robot acceptance as robots become for skilled for example. Therefore, more aspects can be correlated to Hofstede. Furthermore, it would be interesting to look at the developments in these aspects in different countries over the years, as robots advance. When researching these aspects in more countries, researchers could also find a general approach for robot or technology acceptance in a country. With this general approach, a new Hofstede dimension could be added, as there are six so far. However, limitations of this research include applying these results to nonhumanoid robots, as participants were specifically asked to rate a humanoid robot and were give a picture of the Pepper.
5.6.2
Managerial Implications
The results of this study are especially important for future research regarding all kinds of humanoid robots studies done in these countries. Especially interesting
144
5
Study 2: A Cross-Country Comparison
are the results for India, as it shows the potential for using robots. As India rates all aspects the highest from all three countries, humanoid robots could be tasked with many more things than in Germany or the US. For example, India has high trust in robots. Therefore, humanoid robots in India could be used for tasks where higher trust is needed, e.g., elder care. However, in Germany (with a lower trust in robot), one could suggest using humanoid robots as regular service robots. Here, humanoid robots in elder care might be less accepted due to a lower level of trust. This is for example also something Broadbent, Stafford, and MacDonald (2009), from our literature review studied and they found that Germany is less accepting of robots in health care than for example a country like France.
5.6.3
Limitations and Further Research
This study examines how cultural dimensions influence robot attributes. However, the attribution of robots is not only defined by the cultural background. Correlations indicate that prior experience with robots increases attribute levels of empathy (r(610) = .26, p < .001) and trust (r(610) = .12, p < .001) with the robot. The participants’ age decreases the attribute of empathy (r(610) = -.64, p < .001) and trust (r(610) = -.35, p < .001) with the robot. Future research could find out more about how different factors such as age affect these attributes in different countries. Furthermore, the original theory of Hofstede proposed the four dimensions applied in this study. By now, two new dimensions have been added to the concept. Future research could examine the effects of the new Hofstede dimensions long-term orientation and indulgence versus self-restraint. For this paper however, we chose to use only the original dimensions for out hypotheses and for the explanation of the results. As more research is done on the new dimensions, future research could also connect these results to the two new dimensions. Even though the participants were asked to rate the attributes during a service interaction, the attributes are not limited to service interactions with humanoid robots. However, the results of this research are limited to humanoid robots and cannot be implied for non-humanoid robots, due to lager differences between the robot types.
6
Study 3: Beyond the Call of Duty: The Impact of Innovative Service Behavior by Robots on Customer Delight
Innovative service behaviors offer a promising way to excite customers during service encounters; such behaviors increasingly are being performed by frontline service robots (FSRs), functioning as service representatives. This research tests whether innovative service behavioral cues provided by an FSR instead of a human service representative might be able to delight customers too. The empirical findings result from two experimental studies in a hotel setting with either an FSR or a human frontline employee (FLE) as a service representative. The first experimental study (132 participants) compares a human–human interaction with a human–robot interaction in a failure-free service encounter; the second (137 participants) simulates a service failure. The FSR’s innovative service behavior increases customer delight, though confirming their expectations is more important than perceptions of innovative service behavior. Furthermore, even after a service failure, FSRs can achieve customer responses comparable to those evoked by FLEs.
6.1
Introduction
Innovation is key to firm growth (Ahlstrom 2010; Hyytinen and Toivanen 2005) and profitability (Plambeck and Taylor 2005; Wuyts et al. 2004), which generally requires creative employees who can help generate competitive advantages (Coelho and Augusto 2010). Service firms thus invest in encouraging their Based on a publication on International Conference on Information Systems (ICIS) 2018 together with Stock, Ruth.
© The Author(s), under exclusive license to Springer Fachmedien Wiesbaden GmbH, part of Springer Nature 2021 M. Merkle, Humanoid Service Robots, Neue Perspektiven der marktorientierten Unternehmensführung, https://doi.org/10.1007/978-3-658-34440-5_6
145
146
6
Study 3: Beyond the Call of Duty
employees’ innovative service behaviors (Chang et al. 2011; Moosa and Panurach 2008), asking frontline employees (FLEs) to “generate new problem-solving ideas and transform these into uses during the service encounter” (Stock 2015, p. 574). Further, reduced constraints have been shown to increase the number of solutions besides traditional problem solving (Stock-Homburg et al. 2020). Through innovative service behaviors, FLEs go beyond the call of duty (Chebat and Kollias 2000), exceed customers’ expectations (Ho and Gupta 2012), and delight customers (Slåtten and Mehmetoglu 2011; Stock et al. 2017). Customer delight is “a profoundly positive emotional state generally resulting from having one’s expectations exceeded to a surprising degree” (Rust and Oliver 2000, p. 86), and it can increase customer loyalty (Coelho et al. 2011). In addition, innovative service behaviors can address heterogeneous customer needs (Dubinsky et al. 1986) and changing customer requirements (Coelho et al. 2011). One option for providing service in innovative ways is to install frontline service robots (FSRs), which socially interact with and physically support customers (Ivanov et al. 2017). Such FSRs are appearing in diverse service industries, including tourism and hospitality (Gockley et al. 2005; Ivanov et al. 2017; Pan et al. 2015; Pinillos et al. 2016), healthcare (Lee et al. 2017; Mirheydar and Parsons 2013; Piezzo and Suzuki 2017), education (Barakova et al. 2015; Conti et al.2017; Kanda et al. 2007), and retail (Kanda et al. 2010; Sabelli and Kanda 2016; Shiomi et al. 2013). Extant studies provide insights about FSRs’ functioning in real field settings to help researchers optimize their functionalities. However, these observational studies are inherently nonrandom, so they cannot support conclusions about causal relationships, such as service outcomes (Austin 2011; Rosenbaum 2005). Because “creative tasks that were previously reserved for humans will be increasingly automatable in the coming years” (Brynjolfsson and Mitchell 2017, p. 1533), innovativeness is highly relevant to various human–robot interactions (HRI), including those at the service frontline. Research on computational creativity examines the “application of artificial intelligence to the … generation of creative output” (Van Nort and Hogeveen 2017, p. 3). Machine-learning algorithms in robots can create creative systems and support computational creativity behaviors, such as novel real-time choreographies (Augello et al. 2017) through neural networks (Crnkovic-Friis and Crnkovic-Friis 2016), artistic performances (Manfrè et al. 2017), new product ideas (Christensen et al. 2017), and assisting humans’ creativity (Osipyan et al. 2017). Despite a wealth of research pertaining to HRI though (Conti et al. 2017; Gockley et al. 2005; Kanda et al. 2007, 2010; Lee et al. 2017; Pan et al. 2015; Piezzo and Suzuki 2017; Pinillos et al. 2016; Sabelli and Kanda 2016; Shiomi et al. 2013), service research has not sufficiently addressed service robots (Xiao and Kumar 2019). Mende et al. (2019) offer an
6.1 Introduction
147
initial study using virtual embodiments, but the participants in their study watched video recordings of FSRs instead of experiencing a true face-to-face interaction. Therefore, we create face-to-face customer interaction scenarios for this study, with a physically embodied robot and a vignette approach, involving a realistic hotel check-in encounter, to examine how customers respond to an FSR’s innovative service behavior. In particular, we examine whether an FSR can delight customers by expressing a form of innovative service behavior and going beyond the call of duty. With our first research question, Can an FSR delight customers with innovative service behavior? we determine whether the positive effect of innovative service behavior on customer delight also applies when FSRs perform these behaviors. Insights about customer responses to FSRs also might increase understanding of humans’ psychological responses to robotic behaviors. In particular, humans’ psychological responses to HRI arguably might be comparable to or different from those evoked by human–human interactions. Thus, the second research question asks, How does customer delight differ in response to innovative service behavior in human–robot interactions as opposed to human–human interactions? In “regular” service situations, FSRs likely can provide successful service, but the implications for service failure situations have not been examined, to the best of our knowledge. Yet it is virtually impossible for companies to completely avoid service failures (Berry and Parasuraman 1991), which tend to result in negative word of mouth or customer switching and thus diminished firm performance (Bitner et al. 1994; Tax and Brown 1998). Therefore, we include a potential moderating effect of service failure in the innovative service behavior–delight relationship and a third research question: How does a service failure affect the innovative service behavior–delight relationship during human–robot interactions? Finally, we analyze interactions with FSRs in depth to specify the effects of unexpected innovative service behaviors by a robot on customer delight. In a prestudy with 120 participants who rated their expectations of FSRs’ innovativeness, we note that 68.3% of them indicated they did not expect an FSR to be innovative. The unexpected innovative service behavior cues exhibited by the robot thus may trigger customer delight, by evoking excitement and pleasure (Barnes et al. 2013). Our last research question asks, How does the discrepancy of customer expectations regarding robotic innovative service behavior affect customer delight? To answer this question, we analyze both positive and negative discrepancy between expected and perceived innovative service behavior and the associated effects on customer delight during the service interaction. By providing new insights into the relationship between FSRs’ innovative service behavior and customer delight, we contribute to extant research in several
148
6
Study 3: Beyond the Call of Duty
important respects. First, this study is the first to examine face-to-face human– robot service encounters in a natural experimental setting. Second, we advance theoretical understanding of the psychological mechanisms that define HRI during service encounters. Specifically, we introduce theoretical concepts with roots in social psychology, including the expectancy disconfirmation paradigm and cognitive script theory, to HRI. Third, the results provide insights about customer delight in response to innovative service behaviors but also about customer reactions to service failures. Fourth, the experimental design encompasses multiple data sources to test the predicted causal relationships and reduce common method variance (i.e., self-reports by participants and ratings by three independent raters). In turn, we offer recommendations for managers and firms that are considering how to include FSRs in service encounters. They can install them and delight customers, but they need to establish ways to encourage innovative service behaviors, perhaps by relying on learning algorithms for creative systems (Christensen et al. 2017; Manfrè et al. 2017; Osipyan et al. 2017), in both failure-free and failed service encounters.
6.2
Theoretical Background
6.2.1
Innovative Service Behavior
Innovative work behavior refers to “an initiative from employees concerning the introduction of new processes, new products, new markets or combinations of such into the organization” (Amo and Kolvereid 2005, p. 8). Several studies identify antecedents of innovative work behavior, such as leader behavior (Janssen and van Yperen 2004) and job characteristics (Oldham and Cummings 1996), at the individual level or the organizational level. At the individual level, key determinants include altruism (Hu et al. 2009), FLE boreout (Stock 2015, 2016), intrinsic motivation (Coelho et al. 2011), role conflict/ambiguity (Coelho et al. 2011), self-efficacy combined with optimism (Michael et al. 2011), personal networks due to active knowledge sharing (Hu 2009), and ethical leadership (Dhar 2016). On an organizational level, knowledge sharing (Kim and Lee 2013), team support, coordination, cohesiveness (Hu et al. 2009), and leader-member exchange combined with job autonomy (Garg and Dhar 2017) encourage innovative behaviors. Job autonomy, variety, and feedback (Coelho and Augusto 2010) and indirect factors such as work motivation and job stress (Li and Hsu 2016) also can influence innovative service behavior.
6.2 Theoretical Background
149
In terms of the outcomes of such behaviors, only a few studies exist. They indicate that innovative service behavior positively affects customer delight (Stock et al. 2017) and may increase loyalty (Coelho et al. 2011), as well as prompt positive performance outcomes (Yuan and Woodman 2010).
6.2.2
Expectation Disconfirmation Theory
In section 2.3.1.2 this thesis provided a general overview about expectation disconfirmation theory and the underlying mechanisms. The following section demonstrated the applicability to this study and links the theory to expectations and perceptions of ISB and the effect on customer delight. This study relies on expectation disconfirmation theory to predict customer responses in service interactions as this theory has its roots in marketing research. It was already applied to explain how customers respond to services provided by a firm, according to the interplay of customer expectations and actual perceptions of the service (Oliver and DeSarbo 1988; Tse and Wilton 1988; Yi 1990). In information systems research this theory has already been applied to the use of new technologies (for an overview, see Venkatesh and Goyal 2010). For example, McKinney et al. (2002) develop an approach to measure web consumer satisfaction; Bhattacherjee (2001b) examines users’ motives to continue using a system; and Bhattacherjee and Premkumar (2004) explain how beliefs and attitudes about the use of information technologies change over time. Various outcome variables derive from the expectation disconfirmation process, such as technology adoption (Brown et al. 2012; Venkatesh and Goyal 2010), online shopping behavior (Hsu et al. 2006), and web customer satisfaction (McKinney at al. 2002). This study applies expectation disconfirmation theory to explore the relationship between an FSR’s innovative service behavior and customer delight. We focus on customer delight as the dependent variable for several reasons. First, it is a positive emotional state that can result from a positive disconfirmation, in that a customer’s expectations have been exceeded to a surprising degree (Rust and Oliver 2000). Innovative service behavior might generate such a positive surprise (Chebat and Kollias 2000). Second, Stock et al. (2017) argue that at the individual level, innovative service behavior is more important for inducing customer delight than for achieving customer satisfaction.
150
6.2.3
6
Study 3: Beyond the Call of Duty
Script Theory
As the basics of script theory (Tomkins 1978) and the underlying mechanisms have already been shown in section 2.3.2.1, the following section applies script theory to the focus of this study and lays the foundation for further hypothesis development in section 6.3. In service interactions, customers’ scripts serve as standards for evaluating their satisfaction with a provider and its performance (Bitner et al 1994; Mohr and Bitner 1991), as well as their buying behavior (Erasmus et al. 2002; Taylor et al. 1991). Falces et al. (2002) show that satisfaction is influenced by the customer’s script, and Taylor et al. (1991) propose using customers’ role schemata to understand their decision making. We thus anticipate that script theory has important implications for understanding customer responses to HRI during service encounters, such that customers may transfer scripts from well-known service encounters to both human–human interactions and to HRI.
6.3
Framework and Hypotheses
6.3.1
Study Framework
We examine the relationship between innovative service behavior and customer delight during HRI, compared with human-human interactions, in a hotel setting and with different outcomes for the customers. In the framework in Figure 6.1, the dependent variable is customer delight, which implies an extremely positive and emotional state, generated because the customer’s expectations have been greatly and surprisingly exceeded (Rust and Oliver 2000). The independent variables are perceived and expected innovative service behaviors. The former includes inventing new solutions and introducing new ideas to customers and inspiring them (Stock et al. 2017); the latter reflects predictions of some imaginary level of innovative service behavior (Shackle 2012). Therefore, we compare customer responses to the perceived innovative service behavior of an FSR compared with a human FLE, using service failure as a potential. We also focus more specifically on HRI to investigate customer delight in response to the interplay of perceived and expected service innovativeness.
6.3 Framework and Hypotheses
151
Customer Response
Physical representation of the service representative
Frontline Service Robot
Frontline Employee
H1aI Perceived Innovative Service Behavior Perceived Innovative Service Behavior
H3b Service Failure
H1b
H3b
H1aII
Expected Innovative Service Behavior
H3a
Disconfirmation Positive Disconfirmation
H2b Negative Disconfirmation
Confirmation
Customer Delight
H2a
H2c
Figure 6.1 Framework of the Study
6.3.2
Hypotheses
6.3.2.1 Customer Responses to FSRs’ Innovative Service Behavior, Failure-Free Encounter With their innovative service behavior, FLEs can inspire customers and enhance the standard service (Friedman 2002; Ottenbacher and Harrington 2009). Creative ideas and the innovative service behavior may go “beyond the call of duty for customers” (Chebat and Kollias 2000, p. 72). In line with Stock et al. (2017), we anticipate a positive relationship between an FLE’s innovative service behavior and customer delight in human–human interactions. Then, by applying the computers are social actors (CASA) paradigm (Reeves and Nass 1996), we argue that these findings can transfer to HRI, because people display social reactions to computers and other media. That is, people respond to computers as if they were social actors and tend to apply the same social rules they would adopt in an interaction with a human (Nass et al. 1994), such that they mindlessly attribute agency, personality, and intentionality to the technology (Carpenter et al. 2009). The CASA paradigm is well-established and helps predict mindless human behaviors (Moshkina et al. 2014), as applied to HRI involving pictures of robots (Eyssel and Hegel 2012), video studies (Benninghoff et al. 2013), telepresent robots (Edwards et al. 2016), or real physical robots (Bartneck et al. 2006; Tay
152
6
Study 3: Beyond the Call of Duty
et al. 2014). Such studies examine robots in roles as instructors (Edwards et al. 2016) and coworkers (Bartneck et al. 2006); we seek to extend the field to HRI in service encounters, in which a humanoid robot, with an appearance designed to resemble a human (Zimina et al. 2016), provides services using human-like voices and gestures. Therefore, in line with the CASA paradigm (Reeves and Nass 1996), we predict that customers will treat the FSR as a social actor (see Nass and Moon 2000) and apply the same social rules and behaviors (Moshkina et al. 2014; Nass et al. 1994) that they would exhibit in an interaction with a human FLE. Then, we can hypothesize that innovative service behavior increases customer delight, whether that behavior is exhibited by a human FLE or an FSR (Slåtten and Mehmetoglu 2011; Stock et al. 2017). Thus, H 1a :
A customer’s delight is positively affected by (i) an FLE’s innovative service behavior and (ii) an FSR’s innovative service behavior.
We further presume a relationship between the service representatives’ appearance and the corresponding familiarity. Uncanny valley research indicates that humans generally perceive a human more positively than a robot, independent from its human likeness (Mori 1970/2005). Accordingly, the interaction with an FLE is likely to lead to higher familiarity as compared to the FSR (Mori et al. 2012). We applied two physical representations of the service representative: the humanoid robot Pepper from Softbank and a human FLE. According to the uncanny valley paradigm, the FLE leads to higher familiarity as compared to the less humanlike FSR (Mori et al. 2012). This paradigm provides us insights to more deeply understand customers’ responses to the two types of service representatives (FSR and FLE), as the degree of human-likeness is an important robot perception dimension (Belk 2016; Broadbent 2017). That means that customers in the interaction with a FLE experience higher familiarity compared with customers in the interaction with the FSR (Mori et al. 2012). This prediction is in line with Mende et al.’s (2019) findings that customers feel more comfortable with a human FLE than a service robot. In service interactions, customer familiarity is positively correlated with satisfaction (Söderlund 2002) and combined with arousal or surprise it is posited as the concept of customer delight (Bowden 2009; Rust and Oliver 2000). H 1b :
The positive effect of an FSR’s innovative service behavior on customer delight is weaker than that of a comparable FLE’s innovative service behavior.
6.3 Framework and Hypotheses
153
6.3.2.2 HRI and Disconfirmation of Expectations In accordance with expectation disconfirmation theory (Oliver and DeSarbo 1988), we argue that customers likely compare their expectations of an FSR’s innovative service behavior with actually perceived behavior by the FSR. The expectations, according to script theory (Abelson and Schank 1977; Tomkins 1978), develop through prior experiences in a customer role, which inform people’s service expectations in comparable situations when they again act as customers. Their comparison of these expectations with the actual service provided might result in confirmation, positive disconfirmation, or negative disconfirmation. As we noted previously, confirmation implies the customer’s expectations of innovative service behavior are consistent with the perceived innovative service behavior exhibited by the FSR during the service encounter (Bloemer and Odekerken-Schroder 2002). It leads to positive customer reactions, potentially including delight (Bhattacherjee 2001a; Chen et al. 2010; Spreng et al. 1996). Disconfirmation instead means some inconsistency between the expected and perceived innovative service behavior of the FSR (Bloemer and Odekerken-Schroder 2002), whether it is better (positive disconfirmation) or worse (negative disconfirmation) than expected (Churchill and Surprenant 1982). However, “both, positive and negative disconfirmation result in a state of dissonance resulting in psychological discomfort to the users of the system” (Venkatesh and Goyal 2010, p. 288), so disconfirmation negatively affects customer reactions (Yi 1990). Venkatesh and Goyal (2010) also assert that inconsistency due to disconfirmation leads to psychological discomfort (Festinger 1957), and the extent of this discomfort depends on the degree of inconsistency (Szajna and Scamell 1993). Accordingly, we predict that a disconfirmation of an expectation of innovative service behavior will decrease customer delight in service encounters with the FSR. We also seek to specify the two types of disconfirmation in more detail. In the case of a negative disconfirmation, the perceived level of innovative service behavior falls below prior expectations, so there may be an additional negative effect that goes beyond the discrepancy itself, in line with evidence that negative disconfirmation has a stronger negative effect on customer responses than positive disconfirmation (Hsu et al. 2016; Qazi et al. 2017). That is, a lack of expected innovative service behavior is worse than more innovative service behavior than expected. A negative disconfirmation lacks any possibility of pleasure (Oliver 1980) and evokes disappointment without mitigating the negative effects of dissonance (Lankton et al. 2016). In contrast, positive disconfirmation, such that the perceived innovative service behavior exceeds customer expectations (Oliver and DeSarbo 1988), could be appreciated to some extent as a “pleasant surprise”
154
6
Study 3: Beyond the Call of Duty
(Rust and Oliver 2000) that reduces some of the negative effects of the disconfirmation. Therefore, negative disconfirmation should lead to a stronger decrease in customer responses and delight than positive disconfirmation (Hsu et al. 2016; Qazi et al. 2017). Innovative service behavior is a positive characteristic of service encounters (Chang et al. 2011; Moosa and Panurach 2008) and generally drives customer delight, so we predict that, despite the negative effect of disconfirmation, unexpectedly innovative service behavior has a positive effect on customer delight. This positive effect also reduces the negative effect that results from the expectation disconfirmation. This argument implies that expectation confirmation will lead to positive customer reactions (Bhattacherjee 2001a; Chen et al. 2010). However, confirmation is not limited to a single point (Shanock et al. 2010). Congruence and confirmation can range from customers with low expectations combined with low perceptions to customers with high expectations and high perceptions (Shanock et al. 2010). Along this range, all matched combinations of expectations and perceptions are equivalent and produce confirmation (Shanock et al. 2010), which increases positive customer reactions (Chen et al. 2010), due to the lack of psychological discomfort that would be associated with disconfirmation (Venkatesh and Goyal 2010). When we combine this prediction with H1a , which anticipates that higher levels of innovative service behavior increase customer delight with the FSR (Friedman 2002; Ottenbacher and Harrington 2009; Stock et al. 2017), we propose that the absolute level of confirmation exerts a positive effect on the degree of customer delight. H 2a : H 2b :
H 2c :
Disconfirmation between expected and perceived robotic innovative service behavior decreases customer delight. Customer delight decreases more for negative disconfirmation than for positive disconfirmation between expected and perceived innovative service behavior. When perceived and expected robotic innovative service behaviors are congruent, customer delight is higher at higher levels of innovative service behavior.
6.3.2.3 Customer Responses to FSRs’ Innovative Service Behavior, Service Failure Service firms aim to provide failure-free services, but service failures are omnipresent, due to the special characteristics of services (Berry and Parasuraman 1991). Customers must face service failures occasionally (McCollough et al. 2000; Swanson and Kelley 2001), so we consider how a service failure might affect the
6.4 Empirical Studies
155
strength of the link between a service robot’s innovative service behavior and customer delight. According to script theory, customers rely on schemata about service interactions (Taylor et al. 1991), most of which do not feature failures (Reichheld and Sasser 1990), so customers build scripts without service failures. When such a failure occurs, the situation varies from the script, so the customer experiences a discrepancy that evokes mental discomfort, due to cognitive dissonance (Festinger 1957). This discomfort likely reduces the customer’s delight with the service interaction, in line with evidence of distinct declines in positive customer responses after a service failure (Lapré 2011; McCollough et al. 2000). Therefore, we expect a decline in customer delight after a service failure (Rust and Oliver 2000), but in line with H1b, we also predict different customer responses to HRI than to human–human interactions. Specifically, after a service failure, customers may respond more critically to a service robot than a human FLE, because they feel less familiar with it according to uncanny valley paradigm. Even if the FSR exhibits innovative service behavior, customer delight may be lower during HRI than human–human interactions. H 3a : H 3b :
6.4
After a service failure, customers are less delighted with the service representative than they would be in a failure-free service encounter. After a service failure, customers are less delighted with an innovative FSR compared with an innovative FLE.
Empirical Studies
To address our research questions, we conducted three independent studies. With a prestudy survey of potential customers, we determined what qualities and abilities they believed FSRs could display, including the extent to which customers attribute innovative behavior to FSRs. Thus we can assess whether customers’ existing attributions toward FSRs allow for the possibility of disconfirmation due to exhibitions of innovative behavior (Christensen et al. 2017; Osipyan et al. 2017). Then in Study I, we investigate successful service delivery in a hotel setting; participants took the role of a customer who needed to check in to a hotel with the help of the humanoid robot Pepper. This failure-free setting provides the basis for investigating the basic predictions of expectation disconfirmation theory. Finally, Study II
156
6
Study 3: Beyond the Call of Duty
features a similar hotel scenario after a service failure, to examine how the customer’s desire to gain confirmation and avoid disconfirmation might change when the interaction already starts with negative disconfirmation due to the failure.
6.4.1
Prestudy
To understand potential customers’ expectations of a service robot’s innovative service behavior, we surveyed 120 respondents with a paper-and-pencil study. These participants (52.5% men) had an average age of 40.0 years (standard deviation = 18.1 years). The sample had diverse prior experiences with robots (38.3% none, 44.2% little, 10% medium, 2.5% some, and 3.3% a lot of experience) and represented various educational levels (55.0% graduates, 45.0% non-graduates). The 4-item scale of innovative service behavior, adapted from Stock et al. (2017), asked participants to indicate what qualities and abilities they think a humanoid service robot and a human FLE can display. We captured their answers on a 7point Likert scale (1 = strongly disagree; 4 = neither agree nor disagree; 7 = strongly agree). The participants expressed relatively low expectations about an FSR’s innovative service behavior (M = 3.08, SD = 1.60), such that 58.3% of them did not expect a robot to be innovative (M ≤ 3.0), 32.6% were not sure (3.0 < M ≤ 5.0), and only 9.1% expected a service robot to be innovative (M > 5.0). Their expectations of the innovative service behavior of an FLE (M = 5.55, SD = 1.21) were significantly higher (M = 2.47, p < .01). That is, only 5.0% of participants did not expect an FLE to be innovative (M ≤ 3.0), 25.8% were not sure (3.0 < M ≤ 5.0), and 69.2% expected the FLE to be innovative (M > 5.0). Because creative algorithms for generating new product ideas (Christensen et al. 2017), creating news reports (Brooks 2014), and assisting human creativity (Osipyan et al. 2017) already exist, we confront participants of Studies 1 and 2 with an unexpectedly innovative FSR. With the prestudy results in mind, we expect that an FSR’s innovative service behavior will prompt disconfirmation for customers and investigate the customer-related outcomes according to expectation disconfirmation theory.
6.4 Empirical Studies
6.4.2
157
Studies I and II: Common Experimental Setting and Method
Both of the experimental studies took place in a research lab associated with the authors’ university. A room in the lab was outfitted to resemble a hotel lobby and welcome arriving guests; it was stocked with furnishings and objects likely to be familiar to customers of hotels, such as a reception desk, seating group, bookshelf with books, and small shelf near the reception desk with information about the city. Figure 6.2, Panel A, depicts the experimental laboratory setting. In a separate room, with no view of the experimental setting, participants completed the preand post-session surveys. At the beginning of the experiment, they had to wait for 30 seconds, then were told that the service representative had to finish another task, so they would have to wait for another minute. This delay helps ensure that participants become aware of the setting and adjust to the situation; we also captured their initial excitement levels through video recordings, as described in more detail subsequently. For all sessions, the experimental room was free of external sounds, and the room lighting and temperature were maintained at normal residential levels. As Figure 6.2, Panels B and C, depicts, during the experiment, each participant interacted separately with either an FSR or a confederate in the role of an FLE. All visual displays and sounds were recorded by external HD cameras, positioned in the room and on the robot’s body. No experimenters were visible to the participants, but at a hidden station, the experimenter constantly observed the video streams.
6.4.2.1 Mechanical Basis and Manipulation Preparation The mechanical basis for the FSR was the humanoid robot Pepper, a 120 cm robot with 20 degrees of freedom, produced by Softbank. Robots of this type have been used in various HRI settings to study simulated body movements (Domingues et al. 2011; Xu et al. 2014), storytelling and conversation (Csapo et al. 2012; Gelin et al. 2010; Han et al. 2012), smart home interactions with human beings (Louloudi et al. 2010), and emotion transitions for therapeutic treatments (Miskam et al. 2014; Shamsuddin et al. 2012). Pepper’s platform offers simple, moderate facial features, so we manipulated the vocal and bodily expressions offered by Pepper, using Pepper’s graphical programming tool Choregraphe (Pot et al. 2009). We identified appropriate behaviors that would make Pepper appear innovative in two steps. First, we relied on extant innovation research (Chang et al. 2011; Moosa and Panurach 2008; Rego et al. 2014) to identify typical innovative behavior cues (Lee and Ashton 2004; Peterson and Seligman 2004) and included them in the robotic service script. Second, we conducted test runs with 10
158
a
6
surveillance screens
Study 3: Beyond the Call of Duty
reception desk
pepper robot
entrance to experimental setting
participant
camera 1 wizard of Oz
curtain
b
camera 2
c
Figure 6.2 Service Representative and Experimental Setting, Studies I and II
potential customers and 5 innovation researchers. They assessed Pepper’s verbal and bodily expressions as indicators of innovative service behavior during a service encounter and provided feedback for improving them, which we implemented. Figure 6.3 shows some sample innovative service behaviors expressed by the robot in the experiments.
6.4 Empirical Studies
Showing interest (l) and thinking about new solutions (r)
159
Coming up with new ideas (l) and explaining it to the customer (r)
Interrupting standard processes (l) proposing new ways of doing things (r)
“Can find something of interest in any situation” (Peterson and Seligman 2004)
“Is able to come up with new and different ideas” (Peterson and Seligman 2004)
“Come up with something new” (Lee and Ashton 2004)
“Quickly think up new ideas” (Hogan and Hogan 1995)
“Be full of ideas” (Lee and Ashton 2009)
“Love to think up new ways of doing things” (Hogan and Hogan 1995)
Figure 6.3 Sample Gestures by Pepper to Express Innovative Service Behaviors
6.4.2.2 Wizard of Oz Method We adopted a Wizard of Oz method (Gould et al. 1983; Kelley 1984). Participants were told that the FSR acted autonomously, but it actually was being operated by an experimenter, hidden behind a curtain. The Wizard of Oz method has been used frequently to design and collect language corpora in speech-based systems (Dahlbäck et al. 1993). It also has been employed in projects involving HRI (Xu et al. 2014).
6.4.2.3 Experimental Designs For Studies 1 and 2, we use a between-subjects design to avoid learning effects (Atzmüller and Steiner 2010). When participants entered the room, they were asked to imagine that they needed to check in to a previously booked hotel, with the help of an FSR or an FLE. The description of the interaction indicated that they had previously asked for a room without carpeting, due to their dust allergy, that was stocked with an electric kettle. The manipulation relied on four alternative sets of manipulations, only one of which each experimental subject received (Table 6.1). The manipulations differed with respect to the service representative (FLE vs. FSR) and the level of exhibited innovative service behavior (neutral vs. high). To ensure that each participant understood the assigned task, after reading the instructions, they wrote down their task. If there were any errors, the experimenter corrected their understanding and asked for another written summary of the task, as a second check of their comprehension.
160
6
Study 3: Beyond the Call of Duty
Table 6.1 Four Experimental Conditions Behavioral Manipulations
Type of Service Representative
Neutral Service Behavior
Innovative Service Behavior
FSR
Condition (1) The FSR acts in a neutral way, not showing any innovative cues or coming up with additional ideas.
Condition (2) The FSR expresses innovative service behavior by non-verbally and verbally expressing innovations (e.g., trying to come up with helpful solutions, creating new ideas, showing a vivid imagination, thinking up new ways of doing something)
FLE
Condition (3) The FLE acts in a neutral way, not showing any innovative cues or coming up with additional ideas.
Condition (4) The FLE shows innovative service behavior such as trying to come up with helpful solutions, creating new ideas, showing a vivid imagination, thinking up new ways of doing something.
Note: FSR = frontline service robot, FLE = frontline employee
6.4.2.4 Measures and Interrater Reliability Because we sought to observe customer responses to a service robot’s innovative service behavior, we collected data both immediately prior to and immediately after the participants’ 10-minute interactions with the service representatives. They provided self-reports on digital questionnaires. To assess participants’ expectations of innovative service behavior, we relied on the 4-item scale from the prestudy. Prior to the experiment, we also gathered demographic information and some potential control variables, such as technological affinity and prior experience with robots. Directly after their interaction with the FSR or FLE, we asked participants to indicate their perceptions of the innovative service behavior exhibited. The measures all achieved acceptable values for Cronbach’s alpha and composite reliabilities. In addition to these self-reports, we relied on three external, independent raters who observed participants’ behaviors on video and assessed their exhibited delight, using excitement measures (Barnes et al. 2013). The videos showed participants during the interaction with the service representative. We trained the
6.4 Empirical Studies
161
coders to rate levels of excitement, based on facial expressions, gestures, and intonation (Barsade 2002). However, these raters had no knowledge of the experimental conditions or study objectives. Such uses of independent raters have been well established in management research and offer a means to reduce the risk of common method bias (Baer et al. 2004; Barsade 2002; Ford and Gioia 2000; Perry-Smith and Shalley 2014). We checked the consistency of the raters’ assessments according to their interrater reliability, using the interclass correlation coefficient (ICC). In both studies, the three raters achieved a high degree of reliability. In the failure-free setting of Study I, the average ICC was .72 (95% confidence interval [CI] [.62 to .79], F(129,258) = 3.51, p < .01). In the service failure setting of Study II, we also identify high consistency, with an average ICC of .71 and a 95% CI [.63 to .78], F(182,364) = 3.49, p < .01).
6.4.2.5 Data Collection and Analysis As noted, in both Studies I and II, we collected participant data immediately prior to (expectations) and immediately following (perceptions) the 10-minute interactions in the hotel setting. Although extant studies usually adopt a direct measure of disconfirmation with traditional difference scores, this approach suffers from weak explanatory potential; it cannot assess effects according to varying degrees of agreement or discrepancy, nor can it account for nonlinear effects due to information loss (Shanock et al. 2010). Brown et al. (2014, p. 749) assert that “the influence of confirmation on the outcome is dependent on the absolute levels of confirmation,” so for example, Brown et al. (2012) examine how information systems expectations, experiences, and use might be interrelated by relying on polynomial models. Polynomial models offer two important advantages (Venkatesh and Goyal 2010): First, they can reveal interaction effects of expectations and perceptions that affect the confirmation/disconfirmation level. Second, these models “represent additional inflections and curvatures in the underlying surfaces and test such complex relationships” (Venkatesh and Goyal 2010, p. 287). Therefore, we use polynomial regression to assess the effects of experienced and perceived innovative service behavior on the dependent variable (customer delight), but instead of just interpreting the polynomial regression, we also turn to surface response analysis coefficients to examine the potentially curvilinear effects in detail (Shanock et al. 2010). Surface response analysis supports both visual and statistical tests and produces an interpretation of polynomial coefficients, which can be difficult to interpret (Venkatesh and Goyal 2010). Therefore, this method offers a more accurate understanding of the relation between innovative service behavior disconfirmation and customer delight (Edwards 1994).
162
6
Study 3: Beyond the Call of Duty
6.5
Innovative Service Behavior in a Failure-Free Service Encounter (Study I)
6.5.1
Participants
A total of 132 volunteer students, enrolled in business, psychology, or engineering courses, participated in this study. According to G*Power 3.1.92 software (Faul et al. 2007, 2009), a total sample size of N = 128 would produce an effect size of f = .3, an alpha error of α = .05, and a power of (1—β) = .8 for our four experimental groups. Thus, the realized sample size is sufficient to test the hypotheses. The participants (52.3% men) had a mean age of 21.8 years (SD = 5.69). They were required to put themselves into a customer role. As an incentive, all participants received $12 for completing the study, and then they could enter a lottery for three Amazon coupons, worth $500, $200, and $100.
6.5.2
Manipulation Check and Descriptive Statistics
We tested the effectiveness of the between-subjects manipulation by asking each participant to assess the FSR’s or FLE’s innovative service behavior, then comparing the mean scores they provided in the neutral and innovative conditions. The manipulations of innovative service behavior appear successful. The FSR that expressed innovative service behaviors was considered significantly more innovative (M = 6.35, SD = .82) than its neutral counterpart (M = 5.33, SD = 1.32; M = −1.02, p = < .01). Similarly, the innovative FLE (M = 6.65, SD = 1.21) appears significantly more innovative (M = −.86, p < .01) than the neutral FLE (M = 5.79, SD = 1.55). The descriptive statistics are in Table 6.2.
6.5.3
Results
Table 5.6 presents the results from the four experimental conditions; according to an analysis of variance (ANOVA), the effect of innovative service behavior on observed delight (F(3,129) = 9.323, p < .01) is significant. In particular, in support of H1a , customer delight is significantly higher in response to an (i) innovative FLE and (ii) innovative FSR than their neutral counterparts. Even in an HRI, innovative service behavior leads to greater customer delight. We thus explicitly compare the effects of an FLE’s innovative service behavior versus those of an FSR’s innovative service behavior on customer delight, using an ANOVA for
6.5 Innovative Service Behavior in a Failure-Free Service Encounter (Study I)
163
Table 6.2 Descriptive Statistics, Reliabilities, and Intercorrelations for HRI Variables
M (SD)
α / CR / AVE
1
1 Customer Delight (rater assessment)
4.24 (1.25)
.69 / .77 / .54
(.73)
2 Expected ISB
4.55 (1.50)
.91 / .92 / .74
−.07
(.86)
3 Perceived ISB
5.65 (1.22)
.83 / .86 / .60
.28*
.49*
(.77)
4 Participant Age
20.37 (6.12)
-- / -- / -- / --
.05
−.24
−.15
–
5 Experience with Robots
2.00 (1.19)
-- / -- / -- / --
.19
.15
−.10
.18
–
6 Technological Affinity
4.77 (1.40)
.79 / .89 / .73
.08
−.10
.01
.19
.13
2
3
4
5
6
(.85)
Note: The elements on the diagonal (values in parentheses) represent the square root of the variance extracted (AVE) and the values outside the diagonal represent the correlations between constructs. ISB = innovative service behavior; M = mean, SD = standard deviation, α = Cronbach’s alpha. N = 66. * p ≤ .05.
observed delight (F(3,129) = 9.323, p < .01) and a corresponding Bonferroni post hoc test (Table 6.3). As the results in Table 6.4 show, whereas we anticipated a stronger effect of the human–human interaction, we do not find any such differences, whether the service representatives are neutral or innovative. Thus, we must reject H1b . Innovative service behavior increases customer delight, regardless of whether the service representative is an FLE or an FSR. For the tests of H2 , we also address customers’ expectations of innovative service behavior. Many customers still lack much experience with service robots, so their expectations and perceptions likely vary widely. In our sample, more than 90% of participants expressed different ratings of expected and perceived innovative behavior. Specifically, 12.5% of participants indicated higher expectations of the robot’s innovative behavior than they actually perceived; 78.1% of participants instead rated their actual experience with the robot as involving more innovative behavior than they had expected. Because the disconfirmation thus occurs in both positive and negative directions, we conduct polynomial regression analyses to
164
6
Study 3: Beyond the Call of Duty
Table 6.3 Bonferroni Post Hoc Test on Delight, Depending on Service Behavior Neutral Service Behavior Setting (I)
Innovative Service Behavior Setting (J)
Mean Difference (I—J)
p-Value
FLE Interaction
Delight (Third Rater)
3.73
4.75
−1.02*
0.03
FSR Interaction
Delight (Third Rater)
3.68
4.80
−1.12*
0.01
Note: Delight was measured on a 7-point Likert scale: 1 = not at all, 7 = extremely. N(FSR setting) = (33|33); N(FLE setting) = (33|33); FSR = frontline service robot, FLE = frontline employee. * p ≤ .05.
Table 6.4 Bonferroni Post Hoc Test on Delight, FLE versus FSR FLE Interaction (I)
FSR Interaction (J)
Mean Difference (I—J)
p-Value
Neutral Service Behavior
Delight (Third Rater)
3.73
3.68
0.05
1.00
Innovative Service Behavior
Delight (Third Rater)
4.75
4.80
0.05
1.00
Note: Delight was measured on a 7-point Likert scale: 1 = not at all, 7 = extremely. ISB = innovative service behavior, FSR = frontline service robot, FLE = frontline employee. N = 132. * p ≤ .05.
test whether and how they might predict ratings of customer delight. The ANOVA results suggest a predictive effect for the variance in observed customer delight (R2 = .32, p < .01), so we next conducted a response surface analysis. Table 6.5 contains the results from the polynomial regression and the response surface analysis, and Figure 6.4 offers a graphical representation of the significant polynomial model. The results show that the degree of disconfirmation between the expected and perceived innovative service behavior of the FSR has a significant curvilinear relationship with customer delight (a2 = .52, t = 3.27, p < .01). As the graph
6.5 Innovative Service Behavior in a Failure-Free Service Encounter (Study I)
165
Table 6.5 Polynomial Regression and Response Surface Analysis Variable Regressed onto Customer Delight Polynomial regression analysis
Constant (b0 )
4.06*
Expected ISB (b1 )
−.21 (.16)
Perceived ISB (b2 )
.59 (.20)*
Expected ISB squared (b3 )
.18 (.12)
Expected ISB × Perceived ISB (b4 )
.43 (.22)*
Perceived ISB squared (b5 ) Response surface tests
b (SE)
−.08 (.14)
R2
.32*
a1 (slope along x = y)
.38 (.18)*
a2 (curvature along x = y)
.52 (.16)*
a3 (slope along x = −y)
−.80 (.31)*
a4 (curvature on x = −y)
−.33 (.41)
Note: ISB = innovative service behavior; SE = standard error; b = regression coefficient. Polynomial regression model: Z = b0 + b1 X + b2 Y + b3 X2 + b4 XY + b5 Y2 + e, with X (expected ISB), Y (perceived ISB), and Z (customer delight). Interpretation of the polynomial model with surface response analysis a1 = b1 + b2 ; a2 = b3 + b4 + b5 ; a3 = b1 —b2 ; a4 = b3 –b4 + b5 . N = 66. * p ≤ .05
in Figure 6–4 shows, toward the left and right, where disconfirmation increases and customer delight decreases, we find an inverted U-shape along the dashed disconfirmation line, in support of H2a , because the level of customer delight decreases for both positive and negative disconfirmation. Also in line with H2c , we find that confirmation of expectations significantly increases customer delight; delight increases when both expected and perceived innovative service behavior increase (a1 = .38, t = 2.14, p < .05). That is, the highest level of customer delight occurs at the front right corner of the graph, where expected and perceived innovative service behavior are both high, and then decreases toward the back, where both expected and perceived innovative service behavior are in agreement but low. The direction of the disconfirmation matters though (a3 = −.80, t = − .25, p < .05), so as we predicted in H2b , customer delight is lower when expected innovative service behavior is high and perceived innovative service behavior is low, compared with when expectations are low but perceived innovative service behavior is high. Finally, in the bottom left corner of Figure 6.4 (high expected, low perceived innovative service behavior), delight is very low. At the bottom
166
6
Study 3: Beyond the Call of Duty
right corner (high perceived, low expected innovative service behavior), delight is high, though not as great as a condition marked by confirmation between the expected and perceived innovative service behavior (center).
Z Customer Delight
-4 0 -4
-2
0
Y Perceived Innovative Service Behavior
2
4
4
X Expected Innovative Service Behavior
Notes: The solid black line at the bottom represents the line of confirmation, where expected and perceived innovative service behavior match. The area in the front-left triangle on the bottom represents the area of negative disconfirmation, where perceived innovative service behavior is lower than expected innovative service behavior. The right-back triangle represents the area of positive disconfirmation, where perceived innovative service behavior exceeds the expected innovative service behavior.
Figure 6.4 Effects of Innovative Service Behavior on Customer Delight
In summary, customer delight reaches the highest levels when customers obtain confirmation of their expectations. When innovative service behavior is greater, the level of delight increases even more. Negative disconfirmation strongly decreases customer delight. Positive disconfirmation imposes a negative impact on customer delight, because the perceived innovative service behavior deviates strongly from expectations, as we predicted in H2a . If the service robot behaves too innovatively and excessively exceeds customers’ expectations, it can decrease customers’ delight in the HRI.
6.7 Discussion
6.6
167
Innovative Service Behavior after a Service Failure (Study II)
Study II simulates a service encounter marked by a service failure. The instructions explained that participants had booked a suite for a special rate, but the reserved suite had been given to another guest, so the service representative offered the customer a standard room. The 137 student participants in Study II (62.0 men) had a mean age of 23.3 years (SD = 6.29) and were recruited from the same pool as Study I participants.
6.6.1
Manipulation Check and Descriptive Statistics
The manipulation was successful; participants clearly recognized the level of innovative service behavior exhibited, such that the FSR in the innovative service behaviors condition was considered significantly more innovative (M = 4.85, SD = 1.05) than the neutral version (M = 3.88, SD = 1.27; M = −.97; p < .01). The innovative FLE also was considered significantly more innovative (M = 4.64, SD = 1.37) than the neutral FLE (M = 1.65, SD = .99; M = −2.99; p < .01). The descriptive statistics are in Table 6.6.
6.6.2
Results
Table 6.7 provides the results, including the mean values on the measures. Compared with the results of the failure-free Study I, we note an overarching decline in customer delight for all service behaviors and service representatives (Mneutral FLE = -−.93, p < .01; Minnovative FLE = −1.15, p < .01; Mneutral FSR = −.29, p < .01; Minnovative FSR = −.38, p < .01), as proposed in H3a . However, the levels of customer delight evoked do not differ significantly for FSRs and FLEs. The results even seem to indicate that customers might be slightly more delighted with the neutral FSR than the neutral FLE. We thus reject H3b , because customers are not more delighted with an FLE following a service failure.
6.7
Discussion
To predict some outcomes of the expanding introductions of service robots in service encounters and explore corresponding customer reactions, we conducted
168
6
Study 3: Beyond the Call of Duty
Table 6.6 Descriptive Statistics, Reliabilities, and Intercorrelations (Service Failure) α / CR / AVE 1
Variables
M (SD)
1 Customer Delight (rater assessment)
2.87 (1.31) .85 / .86 / .67
(.82)
2 Expected ISB
4.65
.87 / .90 / .69
.17
(.83)
3 Perceived ISB
(1.24)
.89 / .91 / .72
.50*
.32*
4 Participant Age
4.32
-- / -- / -- / --
−.09 .05
5 Experience with Robots
(1.27)
-- / -- / -- / --
−.01 −.01 .10
−.18
.91 / .94 / .83
−.10 −.03 −.03
−.36* .25* (.91)
6 23.32 Technological Affinity
2
3
4
5
6
(.85) −.33* – –
Note: The elements on the diagonal (values in parentheses) represent the square root of the variance extracted (AVE) and the values outside the diagonal represent the correlations between constructs. ISB = innovative service behavior; M = mean, SD = standard deviation, α = Cronbach’s alpha. N = 71. * p ≤ .05.
Table 6.7 Bonferroni Post Hoc Test on Delight, FLE versus FSR (Service Failure) FLE Interaction (I)
FSR Interaction (J)
Mean Difference (I–J)
p-Value
Neutral Service Behavior
Delight (Third Rater)
3.73
3.68
0.05
1.00
Innovative Service Behavior
Delight (Third Rater)
4.75
4.80
0.05
1.00
Note: Delight was measured on a 7-point Likert scale: 1 = not at all, 7 = extremely. ISB = innovative service behavior, FSR = frontline service robot, FLE = frontline employee. N = 137. * p ≤ .05.
6.7 Discussion
169
two experimental studies to (1) empirically test the effects of innovative service behavior cues on customer delight, (2) compare human–human interactions with HRI, (3) compare failure-free interactions with service failures, and (4) examine any discrepancy in the effects of customer expectations on customer delight.
6.7.1
Research Contribution
Although extant innovation research reveals that innovative service behavior is crucial to establish a fruitful customer relationship (Coelho and Augusto 2010; Coelho et al. 2011; Hu et al. 2009; Kim and Lee 2013; Scott and Bruce 1994), previous marketing research has neither examined customer expectations and responses to physical FSRs nor investigated desirable robotic behaviors for service encounters. To the best of our knowledge, this study is the first to examine the effects of robotic innovative service behavior cues at the service encounter and compare them with the outcomes of human–human interactions. According to creative computing research (Brooks 2014; Christensen et al. 2017; Osipyan et al. 2017), algorithms for innovative service behavior already exist, making our study setting realistic. In turn, we derive several relevant implications. First, we demonstrate the benefits of integrating multiple theoretical perspectives, spanning uncanny valley paradigm (Mori 1970/2005) and script theory (Tomkins 1978), as well as the CASA paradigm (Reeves and Nass 1996) and expectation confirmation theory. The CASA paradigm anticipates that people mindlessly treat robots like humans (Reeves and Nass 1996), despite their asocial nature, but uncanny valley paradigm focuses on different levels of familiarity between interactions with humans and robots (Mori et al. 2012). We thus develop differentiated hypotheses for human–human interactions versus HRI, and we establish strong support for the CASA paradigm. It is somewhat surprising that people exhibit mindless treatment of robots engaged in both neutral and innovative service behavior. We thus challenge uncanny valley paradigm, because the physical dissimilarity between humans and robots does not appear sufficiently critical for customers to perceive less familiarity with an FSR and become less delighted with it. Second, we reveal that people do not believe FSRs are able to engage in innovative service behavior (see the prestudy). Noting that they can, we actively decided to confront customers with innovative service behavior by robots that go “beyond the call of duty” (Chebat and Kollias 2000), for which we combined the expectancy disconfirmation paradigm with polynomial models of innovative service behavior in HRI. We find that confirmation evokes no negative effect and
170
6
Study 3: Beyond the Call of Duty
results in the highest levels of customer delight in our studies. But both positive and negative disconfirmations of innovative service behavior expectations have negative influences on customer delight; that is, any disconfirmation, even in a positive direction, will exert a negative effect on customer delight. Unrealistically high expectations, combined with relatively lower perceptions of the innovative service behavior, cast doubt on the abilities of the FSR. Even if customers perceive some innovative service behavior, they continue to focus on what is missing, relative to their expectation, so they experience less delight. Moreover, when the service representative’s perceived innovative service behavior exceeds expectations, it still lowers customer delight. This finding contrasts with prior findings that customers are more delighted in the case of positive disconfirmation (Oliver et al. 1994). To explain this conflicting finding, we note that most studies that apply the expectancy disconfirmation paradigm focus solely on the direction of disconfirmation, not its degree (Szajna and Scamell 1993). With our polynomial model and surface response analysis, we can consider both the degree and the direction of disconfirmation, which produces more nuanced results. Third, we examine customer reactions to FSRs during a service failure. It is critical to handle service failures with care, because they can lead to extreme customer reactions, such as dissatisfaction, loss of loyalty, and negative word of mouth (Dabholkar and Spaid 2012; Lin et al. 2011). Neither robotic research nor service research has previously examined the effects of robot service failures on customer responses yet; as we show, customers appear to respond similarly to both FSRs and FLEs in a failure context. Fourth, our findings introduce a rarely researched topic to marketing research, namely, FSR in encounters with customers. Moreover, robotic research thus far has largely focused on programming and improving human-like behaviors, not human expectations about or responses to physically embodied robots in reallife service encounters. Even service research has barely examined HRI, despite its increasing prevalence in service settings and demands from firms for deeper knowledge (Xiao and Kumar 2019). Our experimental studies involving face-toface interactions with physically embodied robots help address this research gap by examining HRI in a realistic service encounter.
6.7 Discussion
6.7.2
171
Managerial Implications
Our study also contributes to managerial practice by providing valuable insights about desirable robotic behaviors. Organizations already are using FSRs to interact with customers (Barakova et al. 2015; Ivanov et al. 2017; Mirheydar and Parsons 2013), and creative computing research already offers creative algorithms (Augello et al. 2017; Christensen et al. 2017; Crnkovic-Friis and Crnkovic-Friis 2016; Van Nort and Hogeveen 2017). Thus, we identify a completely new field for the potential application of FSRs in service encounters, and we show that customers largely accept FSRs. Their responses are generally positive; for clearly predefined tasks such as a hotel check-in interaction, the range of customer delight with FSRs is comparable to that with human FLEs (see Table 6.4). In particular, our two experimental studies confirm that customers appreciate innovative service behaviors, whether by an FSR or human FLE. Thus, companies should include such behavioral cues when programming their FSRs. With regard to service failures, which affect customer responses, retention, and profitability (Holloway and Beatty 2003), we find it surprising that companies would adopt FSRs without knowing how customers will respond to the inevitable service failures. This study provides pertinent insights, by showing that customers are as likely to forgive a robot as a human FLE after a service failure. They also experience a comparable (reduced) level of delight with the FSR and FLE after a service failure.
6.7.3
Research Directions
This study represents a first step toward a better theoretical understanding and empirical test of the effects of FSRs’ innovative service behavior on customer responses, including customer delight. It also tests interactions with a physically embodied robot in a laboratory-based but real-life service setting. However, the setting is clearly experimental, so continued research might test our research questions in a natural setting, especially after FSRs have diffused more widely in organizations. We include customer delight as our dependent variable and focus exclusively on the outcomes of innovative service behaviors. Additional research might examine other outcomes, such as trust, loyalty, and functionality (e.g., technology acceptance model), along with other behavioral cues. An extension beyond those tests might involve comparisons of the different cues and their effects and outcomes. Further research could examine acceptance of social robots within
172
6
Study 3: Beyond the Call of Duty
organizations too, such as robotic assistants, and then the resulting effects on organizational constructs like culture and leadership. Finally, we investigate customer responses to successful service provision and service failures by FSRs. Continued research should take the next step to examine the effects of robotic service recovery efforts on customer responses, after a service failure occurs.
7
Study 4: Customer Responses to Service Robots: Comparing Human-Robot Interaction with Human-Human Interaction
This paper investigates how service failures affect customers by comparing human-robot interactions with human-human interactions. More specifically, it compares customers’ satisfaction in a service robot interaction depending on a service failure with the customers’ satisfaction in a frontline service employee interaction. On a theoretical basis, extant literature on the uncanny valley paradigm proposed that service robots would create lower satisfaction than human frontline employees would. However, I find that service robots could keep up with human frontline employees. Based on an extensive literature research on service failures, I propose that customer satisfaction after a service failure declines far less for a human frontline employee compared with a service robot. Nevertheless, I find evidence that service robots create even higher customer satisfaction than human frontline employees after the exactly similar service failure. I base my findings on an experimental laboratory study with 120 student participants and the service robot “Pepper” from Softbank Corp.
7.1
Introduction
Digitalization of services is changing the way companies interact with their customers nowadays. Within the last several years, electronic services were revolutionized, so that today’s world is increasingly characterized by technology-facilitated Based on a publication on 52nd Hawaii International Conference on System Sciences (HICSS) 2019.
© The Author(s), under exclusive license to Springer Fachmedien Wiesbaden GmbH, part of Springer Nature 2021 M. Merkle, Humanoid Service Robots, Neue Perspektiven der marktorientierten Unternehmensführung, https://doi.org/10.1007/978-3-658-34440-5_7
173
174
7
Study 4: Customer Responses to Service Robots
transactions. An increasing amount of customers interacts with technologies to create their own service, instead of interacting with a human frontline employee (FLE). Self-service technologies (SSTs) are „technological interfaces that enable customers to produce a service independent of direct service employee involvement (Meuter et al. 2000). Regular SSTs such as automated teller machines (ATMs), ticket machines, airport check-in kiosks, and internet based services such as online banking, already established on a large scale. In recent years, the number of service robots as new service technology skyrocketed. Sales of service robots for professional use sold in 2015 increased by 25% and sales value increased to $ 4.7bn in 2016 (IFR). In contrast to conventional SSTs, this service technology comes along with a physical appearance, and is thus much more similar to human FLEs than the SST. Service robots are already used in many industries such as retail, healthcare and hospitality industry. The French supermarket Carrefour (Carrefour 2017) installed Pepper robots on the shop floor to give customers information on promotions and discounts and hotels as Hilton and Marriott are already experimenting with robots at the reception. At the Hospital in Liège, this service robot is already working at pediatrics (CHR 2016). Even within the robotization, there is the trend to design service robots with an increasingly human appearance, such as the android robot Erica that is almost not distinguishable from a human anymore (Warburton 2017). Although these service robots are becoming increasingly human, studies suppose that customers would be more satisfied with an FLE instead of a service robot (Mori 1970). As of yet there is no empirical proof of this assumption carried out in an experimental real-life scenario. Thus, the first research question addressed in this study is: (1) Do service robots really create lower customer satisfaction than human FLEs? Self-service technologies work quite well for standardized activities and routine procedures. Nevertheless, these services might fail from time to time and lead to service failures. Service failures are “activities that occur as a result of customer perceptions of initial service delivery behaviors falling below the customer’s expectations” (Holloway and Beatty 2003, p. 93). However, these failures are not only occurring in the interaction with a self-service technology but also at the service encounter with an FLE. From service literature, we know that good service recovery is important for firms to maintain customers satisfied and loyal (Bitner et al. 1990). For human FLEs there are already many studies on how to deal with a service failure (see Table 7.1). However, a large number of firms is still struggling with service recovery (Michel et al. 2009) of FLEs.
7.2 Literature
175
This proposed research strives to provide insights on the effects of a service failure on customer satisfaction, comparing human FLEs with service robots, and to answer the second research question: (2) How does a service failure affect customer satisfaction with a service robot compared to an FLE?
7.2
Literature
So far, many studies have investigated service failures, service recovery and corresponding customer responses. However, the vast majority of this research stream relates to traditional service encounters in a human-human interaction (HHI). As digitalization of services moved forward, studies started to examine service failures in the interaction with service technologies more and more frequently. According to this development, I give an overview of the most relevant recent literature in Table 7.1. First, I analyzed service failure studies regarding HHI, as it is important to know which findings from traditional service literature can be transferred to the more technology-based types of service that are increasing in the context of service digitalization. Bonifield and Cole (2007) found that service failures triggers negative customer emotions and therefore affects purchase behavior in a negative way. Besides the service failure itself, the perceived controllability of the service failure plays a major role. Customer reactions are significantly more negative when the firm could have prevented the failure (Choi and Mattila 2008). The level of satisfaction is also negatively correlated with the degree of service failure severity (Weun at el. 2004). Subsequently firms and researchers came up with service recovery strategies to cushion the negative effects of service failures. However, customer satisfaction is always lower after a service failure and recovery than for an appropriate service (McCollough et al. 2000). After a service failure, customers have high recovery expectations and even high recovery performance is not enough to satisfy customers as if there was no service failure (McCollough et al. 2000). Finally, the effect of service failure recovery also depend on the context of the interaction. Leisure customers are more satisfied by recovery than business customers are (Lewis and McCann 2004) and if customers have high expectations of relationship continuity they have lower recovery expectations (Hess et al. 2003). Second, I give an overview of service failure studies focusing directly on services that are provided through technologies. Many studies already focused on e-commerce and online retailing and a few studies already examined the interaction with self-service technologies apart from web-based services. There seems
176
7
Study 4: Customer Responses to Service Robots
to be a difference between online and offline SSTs, as online customers blame themselves more and expect less service failure recovery than offline customers (Harris et al. 2006). In line with the results from HHI (Hess et al. 2003), dissatisfied SST customers are less likely to complain about a service failure if they already had many appropriate service interactions with the SST (Holloway and Beatty 2003). Service recovery might lead to customer satisfaction but this still does not ensure repurchase intentions (Holloway and Beatty 2003). In case of a service failure recovery, it is important that the SSTs provide immediate recovery to reduce negative attributions and increase customer satisfaction (Dabholkar and Spaid 2012). Employee assistance might help to solve the problem, but it even increases the negative attribution to the SST (Dabholkar and Spaid 2012). The extent of the service recovery activities depend on the customer assessment of fairness. If the customer perceives distributive justice in a way that the outcome of the recovery is fair, this increases repurchase intentions (Lin et al. 2011). In comparison to FLEs, customers may prefer to use an SST if it solves a need, is easy to use, avoids service personnel, safes time and money, and provides a better availability (Meuter et al. 2000). Nevertheless, to my knowledge there is no research so far examining service robots in the context of service failures. On the one hand, many results of the service failure research can be easily transferred to the interaction with service robots. On the other hand, service robots differ significantly from other service technologies through their physical appearance.
7.3
Conceptual Background
As antecedent of satisfaction with a service robot, this study relies on the uncanny valley paradigm, as it shows the relation between the appearance of robots and the corresponding acceptance. The basics of uncanny valley paradigm have already been presented in section 2.3.1.3, whereas this section links the uncanny valley paradigm to the specific research questions of this study and provides the basis for further hypothesis development in section 7.4. As this study compares a service robot that is far away from an almost humanlike appearance (see “humanoid robot” in Figure 7.1) with a human FLE (see “healthy person” in Figure 7.1), the uncanny valley itself is not of interest here. Nevertheless, this paradigm claims that a more human-like appearance leads to a higher familiarity. The uncanny valley paradigm provides us insights to more deeply understand customers’ responses to the two different service representatives (service robot
Examination of the differences in consumers´ attributions of blame for service failures and its effect on their expectations for recovery in both online and offline settings
Harris, 2006 Online Service Mohr, and Failure, Customer Bernhardt Attributions and Expectations
Content Effects on negative customer/user attributions to the service provider for services using technology-based self-service technologies
Year Title
Dabholkar 2012 Service Failure and Spaid and Recovery in Using Technology-Based Self-Service: Effects on User Attributions and Satisfaction
Author/s
Framework • Failure recovery (yes/no) • Anxiety level • Source of failure (customer/kiosk) • Source of failure (customer/kiosk) • Employee assistance (yes/no)
Survey • Different service (N = 342) scenarios (bank, Non-Student airline) • Service medium Adults (online, offline) • Attribution of blame
Laboratory Experiment (N = 368) Student Sample
Data
Table 7.1 Literature Review: Customer Responses to Self-Service Technology Failures Customer Response
• Online subjects blame themselves more for the service failure than the offline subjects • Online subjects expect less service failure recovery than offline • More customers complain, the greater the service failure is (continued)
Negative attribution to store
• Customer satisfaction with the failure/recovery experience • Negative attribution to kiosk
7.3 Conceptual Background 177
Holloway and Beatty
Content
Main effects and interaction effects of the dimensions of service recovery justice
• Investigation of consumer responses to online retailer service recovery following a service failure • Existence of service recovery paradox within the context of online retailing?
Only 5 to 10% of dissatisfied customers choose to complain following a service failure
• Examination of the service recovery management of online retailers • Types of service failures which happen during online shopping
Data
Laboratory Experiment (N = 225) Student sample
Critical Incident Study (N = 295) Online Shoppers
Framework
• Distributive justice • Procedural justice • Interactional justice (effects on customer satisfaction in online retailing)
• Delivery problems • Website design problems • Customer service problems • Payment problems • Security problems
Customer Response
z customer satisfaction z negative WOM z repurchase intention
• Distributive justice has a positive influence on repurchase intention • Interaction between types of justice influences:
• Not all dissatisfied customers complain as they already ordered successfully many times • Many customers were not satisfied by the retailer’s recovery effort • Even satisfaction with the recovery effort does not ensure repurchase
7
Lin, 2011 Consumer Wang, and Responses to Chang Online Retailer´s Service Recovery After a Service Failure
Year Title
2003 Service Failure in Online Retailing—A Recovery Opportunity
Author/s
Table 7.1 (continued)
178 Study 4: Customer Responses to Service Robots
7.4 Hypotheses Development
shinwakan (comfort level)
179
uncanny valley
still moving
+
healthy person humanoid robot (e.g. Pepper) stuffed animal
industrial robot
human likeness
50 %
100 % corpse
_
prosthetic hand
zombie
Figure 7.1 Uncanny Valley Paradigm. (MacDorman and Ishiguro 2006, p. 299)
versus FLE), as the degree of human likeness is an important robot perception dimension (Belk 2016; Broadbent 2017; Mori et al. 2012). Specifically that means that customers in the totally human-like FLE interaction experience higher values of familiarity, whereas customers interacting with the less human-like service robot experience lower values of familiarity with the service representative.
7.4
Hypotheses Development
According to the uncanny valley paradigm, I propose that customers experience a much higher familiarity with the human FLE than with the service robot as the robot is much less human-like (see Figure 7.3). In service interactions, customer familiarity leads to customer satisfaction (Söderlund 2002). Therefore, I propose that customers interacting with an FLE experience high levels of satisfaction. Accordingly, I propose that customers interacting with the service robot feel less familiar with it as service representative and therefore experience lower levels of satisfaction compared to the customers of the FLE (see Figure 7.2).
180
7
Study 4: Customer Responses to Service Robots
Service Robot
lower satisfaction
Human Frontline Employee
higher satisfaction
satisfaction
human-likeness
Service Failure
Figure 7.2 Study Framework
As most of our participants are already used to service interactions with FLE but still have only little experience with service robots, this may further increase the familiarity with the FLE compared to the rather unknown and eerie service robot. Thus, I propose: H1: Customer satisfaction is higher for the interaction with a human FLE compared to the interaction with a service robot. I assume that this effect is robust enough to withstand even an unpleasant service encounter after a service failure. Although literature proves a distinct decline of customer satisfaction after a service failure (McCollough et al. 2000), I still propose that the human FLE leads to higher values of customer satisfaction compared to a service robot after a similar failure. Moreover, most customers were still little experienced regarding the interaction with service robot. This may lead to a certain degree of anxiety toward the communication capability of the service robot (Nomura et al. 2006) in case of an unscheduled failure that might require a more intense discussion with the service representative. Compared with the human FLE, the conversation with the robot might be inflexible and the robot might be unable to understand complex situations. A service failure might be a complex situation where customers might not want to rely on a service robot but rather on a human FLE. They might be less satisfied with a service robot in that situation.
7.5 Data Collection
181
Söderlund (2002) found the opposing effect that familiarity is associated with more extreme customer responses like a stronger decline in customer satisfaction after a low service performance. However, his familiarity was related to the type of service and not linked to familiarity with the service representative. Therefore, I chose a hotel check-in and assured that all participants were familiar with such a hotel check-in. In this study, the familiarity refers to the service representative itself and our manipulated severe service failure goes far beyond rather lower level of performance. Therefore, I propose that the effects of anxiety toward the complex communication with the robot and the unfamiliarity with the robotic technology outweigh the situational effects and assume that: H2: In case of a service failure, customers are more satisfied with a human FLE compared to a service robot.
7.5
Data Collection
7.5.1
Mechanical Basis and Manipulation Preparation
As mechanical basis for the experiments, I used the Pepper robot from Softbank. This robot is already widely applied in retail and hospitality industry (Ivanov et al. 2017). As Figure 7.3 shows this robot is clearly distinguishable from a human appearance, even though it is already a humanoid robot. Therefore, this robot clearly ranges on the left side of the uncanny valley. I relied on the Wizard-of-Oz method (Dahlbäck et al. 1993; Kelley 1984), applying a remote-controlled robot in this experiment. The robot operator followed a standardized service script that was designed based on a real hotel situation. The robot communicated via voice, gestures and showed pictures of the hotel rooms on its tablet. I prepared a different script for each manipulation group.
7.5.2
Experimental Setting
To run the experiments in a setting as realistic as possible, the setting of a hotel reception was built up that resembles a realistic hotel situation, which was guided by the design of established experimental studies (Stock and Merkle 2017; Stock and Merkle 2018). Before the participants (N = 120, average age of M = 22.5, SD = 5.2; 43% female) started with the interaction, I briefed them in a separate room regarding
182
-
7
Study 4: Customer Responses to Service Robots
120 cm 28 kg 20£ of freedom 360£ movement Wi-Fi connection 12h of autonomy
4 directional microphones HD cameras
touch screen
tactile sensors
tactile sensors
Figure 7.3 Pepper Robot as Mechanical Basis. (Tanaka et al. 2015)
their task during the interaction, informed them that they were taking part in a scientific experiment, and asked for demographic data. After this instruction, the participants were guided to the hotel lobby. There they had to complete the check in with the service representative, which was either a human FLE or a service robot respectively. During the interaction with the robot, the participants had no knowledge about the operator and were told that the robot acts autonomously. Subsequent to this interaction, the participants filled out the post-experimental questionnaire, rated the level of satisfaction they experienced with the service representative and took part in a small interview with the experimenter.
7.5.3
Experimental Design
In this experimental study, I applied a between-subject design to avoid learning effects. The participants were randomly assigned to one of the four experimental conditions. There were two types of service representatives: a well-trained human service employee and a service robot. Both of them acted according to a detailed
7.6 Results
183
service script. However, there were two different service scripts: one contained an appropriate service where the customer could check in without any complications, whereas the other service script contained a service failure. The failure refers to the reservation. The previously booked suite was not available anymore and instead the participant received a much smaller, less comfortable room that was far away from the accompanying friends. Pictures of both room sizes were presented by the service representative to give the participant an idea how much smaller and less comfortable the new room was. appropriate service service robot service failure
appropriate service human FLE service failure Figure 7.4 Experimental Conditions
However, the service representative (service robot or FLE) took responsibility for the service failure admitting that they made a mistake. There was no way to get a better room or compensation in the setting. Figure 7.4 gives an overview about the four experimental conditions of this study.
7.6
Results
7.6.1
Customer Satisfaction with the Service Robot Compared to the Frontline Employee
As first step, Table 7.2 shows the results from the two experimental conditions with appropriate service by the FLE and the service robot. Subsequent to the interaction, the participants were asked to rate their satisfaction with the service
184
7
Study 4: Customer Responses to Service Robots
representative. Customer satisfaction was assessed through a five-item scale that was developed based on extant service literature (Cannon and Perreault 1999; Homburg and Stock 2005). Table 7.2 Differences in Customer Satisfaction Customer Satisfaction in the setting with a …1
N
Mean Value Standard Deviation
Service Robot
30 6.08
0.90
Frontline Employee
30 5.79
0.96
Note: Customer satisfaction was measured on a 7-point Likert scale: 1 = not at all, 7 = extremely. * p ≤ .05.
In total, I had 30 participants interacting with the service robot delivering an appropriate service. They experienced a high level of satisfaction with the robot (M = 6.08). Based on the uncanny valley paradigm, hypotheses 1 assumed that human FLEs might cause higher levels of customer satisfaction as they are more human-like. However, the 30 participants interacting with the FLE experienced an overall satisfaction that was on a similar level (M = 5.79) with the service robot. This value is even slightly lower than the satisfaction with the robot ( = .29) although the difference is not significant (p = .19). Thus, hypothesis 1 is not supported as the service robot leads to comparable levels of customer satisfaction as the FLE. It has already been pointed out that service robot and FLE provided a comparable service based on the same service script.
7.6.2
Customer Satisfaction After a Service Failure
In Table 7.3, I added the customer responses after a service failure occurred during the interaction with the service representative. In line with our expectations and extant literature, the service failure led to decreased levels of customer satisfaction. Customers’ satisfaction with the robot declined ( = 1.39) after the service failure and reached a significantly (p < .05) lower level (M = 4.69). For the interaction with the human FLE I also observed a significant (p < .05) decline ( = 3.38) in customer satisfaction (M = 2.41). However, hypothesis 2 focused on the different levels of customer satisfaction after the service failure comparing customer responses on the FLE with the
7.6 Results
185
Table 7.3 Scheffé’s Post Hoc Test for Mean Differences in Customer Satisfaction Customer Satisfaction after …1
Service Robot (I)
Frontline Employee (J)
Mean Difference (I—J)
p-Value
Appropriate Service
6.08 (0.90)
5.79 (0.96)
0.30 (0.34)
.860
Service Failure
4.69 (1.89)
2.41 (1.49)
2.27* (0.34)
.001
Note: Customer Satisfaction was measured on a 7-point Likert scale: 1 = not at all, 7 = extremely. N = 120. * p ≤ .05.
responses on the service robot. Even after a service failure, the participants were rather satisfied (M > 4.0) with the robot’s service, than dissatisfied. In contrast, those participants who interacted with an FLE were clearly dissatisfied (M < 4.0) in the interaction with the service failure. Comparing the levels of satisfaction after a service failure, the analysis of variance showed that the effect of the type of service representative (service robot or FLE) on customer satisfaction was significant, F (3, 117) = 46.545, p < .001, ηp 2 = .516. The results of the Scheffé post hoc test show that customers rate the robot significantly (p < .05) better ( = 2.27) than the FLE. Subsequent to the interaction, I conducted a manipulation check for the both manipulations that were applied in this experimental study: service failure/ appropriate service and FLE/ service robot. Therefore, I interviewed the participants after the experiment and asked whether they just talked to a service robot or an FLE. Second, I asked them whether they experienced a service failure during the interaction. All of the participants in the service failure conditions clearly recognized the service failure. Furthermore, the service failure was included in the service script and had exactly the same extent for the human-robot interaction (HRI) as for the HHI. Our results show that the same service failure leads to much lower customer satisfaction with the FLE than with the service robot.
186
7
Study 4: Customer Responses to Service Robots
7.7
Discussion
7.7.1
Rationale for Satisfaction with Failing Robot
Contrary to my assumptions based on the uncanny valley paradigm, I had to reject both of our hypotheses. In this experimental laboratory study, human FLEs were not able to create higher levels of customer satisfaction. Customers interacting with the service robot experienced similar levels of satisfaction than customers interacting with the FLE. This is surprising as I expected that the interaction with a real human might lead to higher customer satisfaction. It might be the case that in such a standardized rather short and less intense interaction, most customers just focus on the interaction itself and on their task and do not really bond with the service representative. As the service representative itself is out of focus, customer satisfaction does not vary significantly between the FLE and the service robot. Regarding customer satisfaction after a service failure, this experimental study revealed results that are even more surprising. Although the participants experienced exactly the same service failure with the service robot as with the FLE, I found that customers were significantly more satisfied with the service robot than with the FLE. How come that the customers were so much more likely to forgive a service robot compared to a human FLE? After the interaction with the service representative, the participants were interviewed to get an impression how they perceived the service representative during the interaction. In the condition with the service failure, customers described the human FLE as ‘moody’, ‘malicious’, ‘unkind’, ‘limited in empathy’ and ‘deliberately uncooperative’ making them experience an ‘unpleasant situation’. Although some of the participants had similar attributions for the service robot, most of them did not consider it as moody or malicious and some participants just reconciled themselves to the service failure. Statements like ‘accidents happen’ and ‘everybody can make a mistake’ rather remind of human characteristics but were made regarding the service robot. Attribution theory postulates that if certain outcomes of an activity—such as the check-in procedure—are viewed as beyond the service representative’s control, occurring service failures tend to be attributed to external circumstances (Anderson 1991). Customers may assume that FLEs have more scope of action than the service robot, as they are more flexible and can even handle sudden unexpected situations. Customers see much more controllability of the situation by the human FLE than by the service robot, as the robot is naturally tied to its programming with no
7.7 Discussion
187
additional scope of action. Therefore, customers might see less controllability by the robot, as they assume it has no control about the service failure itself. Previous studies showed that the perception of controllability leads to enhanced anger and less satisfaction with the service (Folkes et al. 1987). This might explain the lower satisfaction with the FLE who might be considered to have more control about the situation than the service robot. “With SSTs [and service robots], customers create the service for themselves, so it is possible to accept more of the responsibility for the outcome” (Meuter et al. 2000, p. 53; Mills et al. 1983; Zeithaml 1981) and therefore be less dissatisfied in case of a service failure. One may also argue with different expectations customers have regarding the human FLE compared with the service robot. Despite experiencing the same service failure with the FLE as with the service robot, customers might expect service recovery from the FLE as this is already common standard after a service failure. Extant service research shows that service recovery after a service failure might increase customer satisfaction, while the absence leads to dissatisfaction (Tax et al. 1998). However, in our scenario all participants experienced the same situation and ended up with exactly the same hotel room. There was no chance to get a refund, discounts or any other recovery. However, customers might have little experiences with service robots offering service recovery, as companies might not yet have found a way how to proceed recovery via self-service technologies or even service robots. Therefore, customers might not have the expectation that the service robot provides service recovery. According to expectation disconfirmation theory, the same service failure might lead to a higher disconfirmation regarding the FLE compared with the service robot, as expectations toward the FLE were already higher from the beginning. The higher the level of disconfirmation, the lower the satisfaction predicted by this theory. In line with the definition of a service failure as an activity, “that occur as a result of customer perceptions of initial service delivery behaviors falling below the customer’s expectations or zone of tolerance” (Holloway and Beatty 2003, p. 93; Zeithaml et al. 1993), the customers might have perceived the FLE’s service failure as more severe, as they had different expectations compared to the robot.
7.7.2
Research Implications
Starting point for this study was the observation that companies start to rely increasingly on service robots. From service research we know, that it is crucial to
188
7
Study 4: Customer Responses to Service Robots
handle service failures with great care, as this may lead to extreme customer reactions such as dissatisfaction, loss of loyalty, and negative word of mouth for example (see Table 7.1). Therefore, it is surprising that IS research has not yet examined the effects of robot service failures on customer responses. To my knowledge, this is the first study to examine customer responses on service failures committed by a service robot at the customer encounter. Robotic research is a rapidly growing research stream. However, to my knowledge it has not yet reached the depth to examine service failures comparable to those caused by human FLEs. So far, the focus is more on robot acceptance and on functional failures. In addition, service research did not examine service robots in the context of service failures and customer responses although this is an increasingly present phenomenon in organizations applying service robots with customers. This study contributes to that research gap by examining customer responses on robot service failures. Second, I attempted to more deeply understand the interaction of the uncanny valley paradigm with attribution theory and confirmation-disconfirmation theory and the effects on customer responses regarding service failures in HRI. The results show that customer responses to service robots differ strongly from responses to human FLEs—in a way that is not consistent with extant assumptions from the uncanny valley paradigm.
7.7.3
Managerial Implications
This study contributes to decision-makers in the field of digitalized services. We observe companies relying increasingly on service robots in interactions with customers. Even in the traditional interaction between human FLEs and customers, service failures occur repeatedly. As these failures affect customer satisfaction and therefore affect customer retention, profitability (Holloway and Beatty 2003) and price sensitivity (Stock 2003), it is quite surprising that companies are already applying service robots in the field without knowing customer responses on robotic service failures. At that point, this study shows that service robots meet great acceptance among customers. Under regular circumstances without service failures, service robots are able to induce customer satisfaction on a comparable level as FLEs. Moreover, this study provides insights, that customers are more likely to forgive a robot instead of an FLE after a service failure. This means that after a service failure, customers experience a higher level of satisfaction with the robot compared to the FLE.
7.7 Discussion
189
Thus, companies should consider expanding the application of service robots or comparable digital service technologies in the context of service recovery besides the current approaches in customer recovery (Homburg et al. 2004). It may be worth exploring new ways to deliver reasonable service recovery via these technologies.
7.7.4
Limitations and Areas of Future Research
The results of this study are not in line with the assumptions made relying on the uncanny valley paradigm. Further research should further specify this theoretical paradigm with additional empirical studies in real-life scenarios and various stages along the graph. Previous studies already criticized this paradigm as too simplistic and rather weak in the definition of the dimensions (Bartneck et al. 2009). However, this study did not include perceptions of the appearance of Pepper, which might also influence the opinions of the participants. Furthermore, this study was restricted to customer responses on service failures by service robots. Future research should examine the effect of robotic service recovery on customer responses, as this is supposed to be the next step after a service failure occurred. Finally, the examined data is just based on an experimental study. Future studies should examine comparable research questions in a real-life field study when service robots are more established in organizations. Customers might show more intense reactions in real-life scenarios than in the experimental setting making a first novel experience. Continued robot encounters may change customer satisfaction over time (Söderlund 2002), raising a need for longitudinal studies.
8
Overall Discussion
Starting point of this thesis was the question what customers expect from professional humanoid service robots and how they respond to this new type of service representative at the frontline service encounter. Professional service robots are on the rise and companies are increasingly applying robots at the service encounter with the customer. Based on the growing managerial relevance for service firms, service marketing research just started to consider service robots as promising trend. However, corresponding studies in the field examining customer expectations and customer responses to physical service robots are still scarce. This thesis is essential to evaluate human-customer interactions at the frontline service encounter. So far, the two research streams of service marketing and human-robot interaction were running largely in parallel without developing significant conjunctions. Therefore, this thesis analyzed the topic based on the two following major goals (see section 1.3): – Major Goal 1: Analysis of customer attributions and expectations towards professional service robots. – Major Goal 2: Analysis of customer responses to humanoid professional service robots at the service encounter. Corresponding with these two major goals, this thesis reports four key findings regarding customer expectations and customer responses to the application of humanoid service robots at the frontline service encounter. First, this thesis showed that customers’ expected roles and corresponding references affect their acceptance of service robots (see chapter 4). Second, a cross-cultural study revealed a significant connection between cultural dimensions and corresponding
© The Author(s), under exclusive license to Springer Fachmedien Wiesbaden GmbH, part of Springer Nature 2021 M. Merkle, Humanoid Service Robots, Neue Perspektiven der marktorientierten Unternehmensführung, https://doi.org/10.1007/978-3-658-34440-5_8
191
192
8
Overall Discussion
expected characteristics of service robots (see chapter 5). Third, customer responses to service robots could keep up with human frontline employees, specific behavioral cues could affect customer delight with the robot and discrepancy between expected and perceived robotic behavior negatively affected customers (see chapter 6). Fourth, customers were more likely to forgive a service robot than a human frontline employee after a service failure (see chapter 7). Based on these findings, this thesis derives a series of implications for scientific research (see section 8.1) as well as implications for managerial practice (see section 8.2). Subsequently, an overview is given about the overall limitations of this thesis and recommended areas for future research (see section 8.3).
8.1
Overall Scientific Contribution
This thesis examines customer expectations and customer responses to service robots from a service marketing perspective and derives several implications to expand the current state of research in the field of service marketing. The scientific insights of this thesis are subdivided in methodological contributions, contentual contributions and conceptual contributions. Three areas can be assigned to the methodological contribution. The first methodological contribution refers to the generation of a unique data sample for empirical analysis that goes far beyond the current standard in service marketing research on service robots. This thesis is based on experimental studies in real-life service scenarios with live face-to-face interactions between customers and physically embodied service robots. In total the data basis consists of 471 extensive service interactions (82 in study 2, 132 and 137 in study 3, 120 in study 4) with a duration between 10 and 15 minutes respectively and is unique in its feature to compare human-human service interactions with human-robot interactions across different service behaviors and service outcomes. The second methodological contribution contains the introduction of polynomial models and surface response analysis to the examination of expectation disconfirmation theory in human-robot interactions. Although extant studies usually adopt direct measures of disconfirmation with traditional difference scores, this approach suffers from weak explanatory potential and cannot assess effects according to varying degrees of agreement or discrepancy, nor can it account for nonlinear effects due to information loss (Shanock et al. 2010). However, the influence on the outcome does depend on the absolute level of confirmation or disconfirmation and the method of polynomial models is able to detect these effects (Brown et al. 2014). This thesis relies on polynomial models
8.1 Overall Scientific Contribution
193
because of two important advantages: First, they can reveal interaction effects of expectations and perceptions that affect the confirmation/disconfirmation level. Second, these models represent potential non-linear effects and allow to test for such complex effects (Venkatesh and Goyal 2010). To interpret the results of the polynomial regression, this thesis turns to surface response analysis to examine the curvilinear effects in detail (Shanock et al. 2010). Surface response analysis supports both, visual and statistical tests and produces an interpretation of polynomial coefficients, which can be difficult to interpret (Venkatesh and Goyal 2010). Therefore, this method offers a more accurate understanding of the relationship between disconfirmation of expectations and customer responses. This methodological approach is not yet widespread in disconfirmation research, but should be applied more broadly in future research. The contentual contribution consists of three major areas. The first contentual contribution refers to customer expectations toward service robots. This thesis addressed two major effects forming customer expectations regarding service robot characteristics (see chapter 1) and roles of service robots during the interaction (see chapter 2). Based on cultural dimensions theory, this thesis revealed a connection between customers’ robot attributions (empathy, expertise, reliability, trust) and the cultural dimensions (individualism, masculinity, power distance, uncertainty avoidance) by comparing expectations across different countries. This is the first study to verify such an interrelationship that is worldwide applicable, as cultural dimensions theory provides global rating for various countries. Considering the extensive criticisms on Hofstede’s cultural dimensions theory (see section 2.3.2.4), the identified effects might also be applied relying on a more refined conceptualization of cultural dimensions such as the GLOBE research project. These results are especially important for future research regarding all kinds of humanoid robot studies, as studies conducted in one country are not necessarily representative for another country, leading to limitation of research results and hampering comparability of international studies. Further, this thesis relied on role theory (Solomon et al. 1985) to extract different categories of role perceptions or experiences that are responsible for customers’ acceptance of service robots during the service encounter. Thereby, study 1 shows that it is essential to consider the basic underlying role script that customers apply to the interaction, as it significantly influences the perception of the service robot. In an experimental study (see chapter 4), this thesis revealed that these role expectations were particularly important for the informational and the relational component of service robot acceptance.
194
8
Overall Discussion
The second contentual contribution is about customer responses and the extant research gap in experimental research with physically embodied service robots in a real life service context. So far, extant research largely relies on laboratory experimental settings to examine user acceptance of robots in general. This thesis is one of the first studies that examines customer acceptance of humanoid service robots in a real life hotel context. It adapts aspects of the well-established technology acceptance model (TAM) to anticipate customer responses, but also shows up shortcomings of the TAM when it comes to social robots that comprise a stronger relational component than traditional technologies considers by the TAM. Further, it shows that customer responses depend on underlying behavioral scripts (2.3.2.1) and role expectations (2.3.2.2) that customers apply to the interaction. Additionally, the physical interaction with a service robot enabled this thesis to discover effects of robot appearance on customer responses based on the opposing theoretical approaches of the CASA paradigm (arguing that customers treat service robots like humans) and the uncanny valley paradigm (arguing that robots do not reach comparable levels of familiarity with the customer). As the studies within this thesis relied on experiments in real-life scenarios with high levels of immersion, this thesis could identify various complex effects of disconfirming customer expectations on customer responses, relying on EDT and cognitive dissonance theory. Further, literature on robot acceptance has essentially focused on human-robot interactions, but largely neglected human-human interactions in comparison. Previous service marketing research has neither examined customer expectations and responses to physical FSRs, nor investigated desirable robotic service behaviors at the service encounter. Therefore, this is the first study to examine the effects of robotic service behavior cues at the service encounter with physically embodied robots and to compare them with the outcomes of human-human interactions. The third contentual contribution concerns the examination of customer reactions to service robots during a service failure. It is critical to handle service failures with care, as they can lead to extreme customer reactions, such as dissatisfaction, loss of loyalty, and negative word of mouth (Dabholkar and Spaid 2012; Lin et al. 2011). Neither HRI research, nor service marketing research has previously examined the effects of robot service failures on customer responses yet. As shown, customers tend to forgive service robots easier than human frontline employees in case of a service failure. While the contentual contributions of this thesis were mainly focused on service robots at the customer encounter based on experimental studies in the hospitality sector, many of the results can be transferred to other social robots such as for example private social robots, as well as to adjacent, research fields such as avatars and conversational agents. Further, the results are not strictly tied
8.1 Overall Scientific Contribution
195
to the hospitality sector and might be transferred to the application of service robots in other industries as well, as examined by studies such as Mende et al. (2019). The conceptual contribution contains three aspects. The first conceptual contribution concerns the conjunction of the two fields of research (see section 2.2), surrounding the overarching research framework of this thesis. So far, service robots have been examined separately in service marketing (e.g. Huang and Rust 2018; Jörling et al. 2019; Mende et al. 2019; Wirtz et al. 2018; Xiao and Kumar 2019) and in human-robot interaction literature (e.g. Kanda et al. 2010; Kirby et al. 2010; Pan et al. 2015; Pinillos et al. 2016; Rodrigues-Lizundia et al. 2015; Sabelli and Kanda 2016; Shiomi et al. 2013; Trovato et al. 2017; Yamazaki et al. 2010). Although both research streams complement each other, service robots are examined independently of each other in both research streams, although both could benefit from a mutual intersection: service marketing research could apply live face-to-face interactions with physical robots instead of pictures and video recordings and HRI research could base its propositions on a broader theoretical basis and increase its managerial relevance. Further, extant research in service marketing is scattered and rather inhomogeneous (Xiao and Kumar 2019). This thesis closes the gap and concatenates both research streams, relying and referring to findings from both research streams and creating an important contribution at the intersection between service marketing and human-robot interaction literature. The second conceptual contribution concerns the theoretical foundation of the overarching research framework. This thesis applies a comprehensive approach, relying on multiple theoretical perspectives from across various research fields, such as information systems and HRI research (TAM, EDT, CASA, uncanny valley paradigm), as well as psychological and socio-psychological research (cognitive dissonance theory, script theory, role theory, cultural dimensions theory). It demonstrates the benefits of integrating multiple theoretical perspectives and provides a research model framework based on a strong theoretical foundation and adapts the theories to fit for human-robot interactions. Further, this thesis provides a detailed overview of adjacent theoretical perspectives (see section 2.3) that can be applied in further future studies on service robots at the frontline encounter. The third conceptual contribution refers to the development of a service robot acceptance model (SRAM) to explain customer acceptance of humanoid robots during service encounters. The theoretically developed SRAM is based on a qualitative study and relies on the technology acceptance model (TAM; Davis 1989) and role theory (Solomon et al. 1985). This thesis extended the components of the
196
8
Overall Discussion
TAM by adding two additional components, i.e., the informational and the relational component. These components are important due to the human appearance of humanoid service robots, which makes the information exchange and relationship building during the human-robot interaction more important than within a human-computer interaction. Further, this new model was empirically tested in an experimental study in a service setting (see chapter 4) and other researchers in hospitality research (Collins 2020; Io 2020; Lin et al. 2019) already seized upon it and made further adjustments (Ghazali et al. 2020).
8.2
Overall Managerial Contribution
Besides scientific contribution for research, this thesis further provides managerial contribution for practitioners and firms in the service domain. As this thesis provides a holistic view on customers interacting with service robots, it derives managerial contributions based on customer expectations and on customer reactions toward professional service robots at the service encounter. Although organizations increasingly apply service robots all over the world at the customer encounter, they hardly know about customer expectations and customer responses to service robots. The managerial contributions of this thesis provides valuable insights for service firms that are structured in six areas of implications. The first managerial contribution concerns the comparison of customer responses toward service robots with responses toward human employees. In several experimental studies, this thesis shows that customers largely accept service robots. Customer responses are generally positive and service robots can catch up with human frontline employees in terms of customer satisfaction and customer delight. These positive customer responses have been shown for clearly predefined tasks, such as a hotel check-in interactions and under the premise of a clear technical implementation and a smoothly running process. This implies that service robots can be applied at the customer encounter for selected fields of application when the functional implementation is done properly. The second managerial contribution is of particular relevance for the specific configuration regarding the programming of the service robot. This thesis provides valuable insights about desirable robotic behaviors at the boundary between service firms and their customers. Specifically, this thesis shows that customers appreciate innovative service behaviors and that creative computing research already offers creative algorithms (Augello et al. 2017; Christensen et al. 2017; Crnkovic-Friis and Crnkovic-Friis 2016; Van Nort and Hogeveen 2017). Thus,
8.2 Overall Managerial Contribution
197
companies should consider these artificial behavioral cues when programming their service robots. The third managerial contribution contains insights into the effects of discrepancies between customer expectations and customer perceptions of service robots on customer responses. This thesis shows that confirmation between customer expectations and perceptions toward the service robot leads to positive customer responses and the level of confirmation is also correlated with positive customer responses. Thus, service firms should generally attempt to set and achieve high expectations regarding the behaviors of service robots. However, it seems crucial to inform customers about the capabilities of service robots to align expectations with the actual abilities of the robots. This will decrease positive and negative disconfirmation to a minimum and therefore increase positive customer responses in the customer-robot interaction at the service encounter. The fourth managerial contribution refers to the cultural differences in terms of customer expectations with corresponding robot acceptance and is especially important for international companies and their roll-out strategies for service robot application. This thesis revealed significant intercultural differences in terms of robot acceptance and therefore, service firms should not just transfer their experiences with customer-robot interactions from one country to another. Service firms should rather consider country-specific attitudes toward service robots. Therefore, this thesis successfully linked Hofstede’s cultural dimensions (Hofstede 1980) that are available for more than 50 countries to customers’ robot attributions. This finding may serve as an initial assessment and service firms can rely on it as a first rough guide to determine how to apply service robots in different countries. In countries with rather high levels of trust in robots such as e.g. India, service robots can be applied on a much broader level also including tasks with higher trust requirements such as for example elder care, compared to countries with lower trust in service robots. The fifth managerial contribution relates to the question of how role assignments and customer mental references affect customer responses to service robots. In an experimental study, this thesis revealed that the majority of customers does not compare service robots to interactions with self-service technologies. Therefore, service firms should resist the temptation to just transfer scripts from self-service technologies to service robots. Customers rather compare service robots to human frontline employees or to their ideal expectations. This implies that besides the functional component, the interaction with a service robot further requires a relational component to trigger positive customer responses.
198
8
Overall Discussion
The sixth managerial contribution concerns the occurrence of service failures, which affects customer responses, retention, and profitability (Holloway and Beatty 2003). It is surprising that companies would adopt service robots without knowing how customers will respond to the inevitable service failures. This thesis provides pertinent insights, by showing that customers are more likely to forgive a service robot as a human frontline employee after a service failure in terms of customer satisfaction. Thus, service firms should consider expanding the application of service robots in the context of service recovery. It may be worth exploring new ways to deliver reasonable service recovery by service robots. In summary, this thesis provides comprehensive implications for managerial practice regarding customer expectations and customer responses to service robots at the frontline encounter. For service firms it is essential to examine the application of service robots thoroughly, to build up a specific service robot implementation strategy and to keep a customer-centric perspective to anticipate potential customer expectations and customer reactions toward service robots. This thesis provides an encompassing basis for the application of professional service robots at the customer frontline encounter.
8.3
Overall Limitations and Recommended Areas for Future Research
Overall, this thesis contributes to extant research and significantly expands the current state of knowledge regarding customer expectations and customer responses toward professional humanoid service robots. Nevertheless, this thesis contains certain limitations and shows various recommended areas for future research. The major limitations of this thesis with its corresponding areas for future research are mainly based on the data basis and the applied method and can be structured in five areas. The first limitation concerns the setting of the experimental studies. Although the experiments were conducted in a real-life scenario, the setting is clearly experimental and laboratory-based. Customers might show less intense reactions in such an experimental setting. Further, the experimental dataset is completely based on studies with student samples. Future research should examine comparable research questions in a natural setting with real-life field studies, especially after service robots have diffused more widely in organizations. This will further enable studies on the acceptance of robots within organizations too, such as robotic assistants and the resulting effects on organizational constructs like culture and leadership.
8.3 Overall Limitations and Recommended Areas for Future Research
199
The second limitation refers to the lack of longitudinal observations or longitudinal experiments. The data this thesis relies on is just based on one-time-only interactions between participants and service robots for a duration of 10–15 minutes. Continued interactions with service robots may change customer responses such as satisfaction or delight over time, raising a need for longitudinal studies. The third limitation is based on the limited selection of variables. This thesis includes customer satisfaction and customer delight as service-related outcome variables and focuses exclusively on the outcomes of innovative service behaviors, service appropriateness and role expectations. Additional research might examine other service-related outcomes, such as customer loyalty and trust, along with other behavioral cues. An extension beyond those tests might involve comparisons of the different cues and their effects and outcomes. The fourth limitation relates to the consideration of service failures in customer-robot interactions. This thesis was restricted to customer responses on service failures by service robots and human employees. Future research should examine the effect of robotic service recovery on customer responses, as this is supposed to be the next step after a service failure occurs. The fifth limitation refers to the application of robots exclusively at the service encounter. Future research could examine service robots in the area of human resources (see Bieling, Stock, and Dorozalla 2015 and Stock 2004 for an overview). For example, social robots could be placed in the context of teams at the boundary between firms and their customers. Former research has emphasized the importance of interorganizational teams at the boundary between firms and customers (Gaitanides and Stock 2004, Stock 2013). Furthermore, social robots could be examined in terms of their potential role as leaders, e.g. by examining implicit assumptions about robotic leaders, relying on implicit leadership theory (Stock and Özbek-Pothoff 2014) or by examining intercultural differences regarding robots in different roles in teams or robotic leaders (Stock and Genisyürek 2012). Further, future research on service robots in service marketing and robotics should be more interrelated, as both research streams examine service robots with complementary aims. Future service marketing research should apply more realistic settings, relying on live face-to-face interactions with physical robots and future HRI research should be based on a stronger theoretical foundation and might consider to increase the focus on service outcomes. In summary, this thesis provides a profound contribution to the field and expands current knowledge on customer expectations and customer responses to
200
8
Overall Discussion
service robots at the frontline encounter. Therefore, it provides important scientific and managerial contributions. As service robots are predicted to rise and as robotic and AI research is progressing rapidly, the topic of service robots at the customer encounter will further gain relevance in the future and service firms will have to deal with service robots strategically. Therefore, further research in this field is desirable from a scientific and from a managerial perspective.
References
Abelson, R., & Schank, R. C. (1977). Scripts, plans, goals and understanding. An inquiry into human knowledge structures, Hillsdale, NJ: Lawrence Erlbaum Associates. Aggarwal, P., & McGill, A. L. (2007). Is that car smiling at me? Schema congruity as a basis for evaluating anthropomorphized products. Journal of Consumer Research, 34(4), 468–479. Ahlstrom, D. (2010). Innovation and growth: How business contributes to society. Academy of Management Perspectives, 24(3), 11–24. Aiken, L. S., West, S. G., & Reno, R. R. (1991). Multiple regression: Testing and interpreting interactions. Sage Publications. Ailon, G. (2008). Mirror, mirror on the wall: culture’s consequences in a value test of its own design. Academy of Management Review, 33(4), 885–904. Alpeyev, P., & Amano, T. (2015). Robots at work: SoftBank aims to bring pepper to stores. Bloomberg Business, June, 30. Åmo, B. W., & Kolvereid, L. (2005). Organizational strategy, individual personality and innovation behavior. Journal of Enterprising Culture, 13(1), 7–19. Anderson, C. A. (1991). How people think about causes: Examination of the typical phenomenal organization of attributions for success and failure. Social Cognition, 9(4), 295–329. Anderson, E. W., & Sullivan, M. W. (1993). The antecedents and consequences of customer satisfaction for firms. Marketing Science, 12(2), 125–143. Andreassen, T. W., van Oest, R. D., & Lervik-Olsen, L. (2018). Customer inconvenience and price compensation: a multiperiod approach to labor-automation trade-offs in services. Journal of Service Research, 21(2), 173–183. Argall, B. D., & Billard, A. G. (2010). A survey of tactile human–robot interactions. Robotics and Autonomous Systems, 58(10), 1159–1176. Arras, K. O., & Cerqui, D. (2005). Do We Want to Share our Lives and Bodies with Robots? A 2000 people survey. LSA-Report-2005–002 [https://infoscience.epfl.ch/record/97585/ files/SurveyPaperArrasCerqui.pdf].
© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Fachmedien Wiesbaden GmbH, part of Springer Nature 2021 M. Merkle, Humanoid Service Robots, Neue Perspektiven der marktorientierten Unternehmensführung, https://doi.org/10.1007/978-3-658-34440-5
201
202
References
Arras, K. O., & Cerqui, D. (2005). Do we want to share our lives and bodies with robots? A 2000 people survey. Technical Report No. 0605–001, Autonomous Systems Lab, Swiss Federal Institute of Technology Lausanne (EPFL). Atwater, L. E., C. Ostroff, F. J. Yammarino, & J. W. Fleenor. (1998). Self-other agreement: does it really matter? Personnel Psychology, 51(3), 577–598. Atzmüller, C., & Steiner, P. M. (2010). Experimental vignette studies in survey research. Methodology: European Journal of Research Methods for the Behavioral and Social Sciences, 6, 128–138. Augello, A., Cipolla, E., Infantino, I., Manfre, A., Pilato, G., & Vella, F. (2017). Creative robot dance with variational encoder. arXiv preprint arXiv:1707.01489. Austin, P. C. (2011). An introduction to propensity score methods for reducing the effects of confounding in observational studies. Multivariate Behavioral Research, 46(3), 399–424. Baer, R. A., Smith, G. T., & Allen, K. B. (2004). Assessment of mindfulness by self-report: the Kentucky inventory of mindfulness skills. Assessment, 11(3), 191–206. Bagozzi, R. P. (1979). The role of measurement in theory construction and hypothesis testing: toward a holistic model, in: Ferrel, O., Brown, S., & Lamb, C. (eds.), Conceptual and theoretical developments in Marketing, Chicago, 15–32. Bagozzi, R. P. (2007). The legacy of the technology acceptance model and a proposal for a paradigm shift. Journal of the Association for Information Systems, 8(4), 244–254. Bagozzi, R. P. (2007). The legacy of the technology acceptance model and a proposal for a paradigm shift. Journal of the Association for Information Systems, 8(4), 3. Bagozzi, R. P., & Fornell, C. (1982). Theoretical concepts, measurements, and meaning. Fornell (ed.), A Second Generation of Multivariate Analysis, (vol. 2), 5–23. Bagozzi, R. P., & Phillips, L. W. (1982). Representing and testing organizational theories: A holistic construal. Administrative Science Quarterly 27(3), 459–489. Bagozzi, R. P., & Yi, Y. (1988). On the evaluation of structural equation models. Journal of the Academy of Marketing Science, 16(1), 74–94. Bagozzi, R. P., & Yi, Y. (2012). Specification, evaluation, and interpretation of structural equation models. Journal of the Academy of Marketing Science 40(1), 8–34. Bagozzi, R. P., Yi, Y., & Phillips, L. W. (1991). Assessing construct validity in organizational research. Administrative Science Quarterly 36(3), 421–458. Barakova, E. I., Bajracharya, P., Willemsen, M., Lourens, T., & Huskens, B. (2015). Longterm LEGO therapy with humanoid robot for children with ASD. Expert Systems, 32(6), 698–709. Barnes, D. C., Collier, J. E., Ponder, N., & Williams, Z. (2013). Investigating the employee’s perspective of customer delight. Journal of Personal Selling & Sales Management, 33(1), 91–104. Barrett, M., Oborn, E., Orlikowski, W. J., & Yates, J. (2012). Reconfiguring boundary relations: Robotic innovations in pharmacy work. Organization Science, 23(5), 1448–1466. Barsade, S. G. (2002). The ripple effect: Emotional contagion and its influence on group behavior. Administrative Science Quarterly, 47(4), 644–675. Bartko, J.J. (1966). The intraclass correlation coefficient as a measure of reliability. Psychological Reports 19(1), 3–11. Bartlett, F. C. (1932). Remembering: A study in experimental and social psychology. Cambridge: Cambridge University Press.
References
203
Bartneck, C. (2002, November). Integrating the occ model of emotions in embodied characters. In Workshop on Virtual Conversational Characters, 39–48. Bartneck, C., Nomura, T., Kanda, T., Suzuki, T., & Kennsuke, K. (2005). A cross-cultural study on attitudes towards robots. Proceedings of the HCI International, 10. Bartneck, C., Reichenbach, J., & Carpenter, J. (2006). Use of praise and punishment in humanrobot collaborative teams. In ROMAN 2006-The 15th IEEE International Symposium on Robot and Human Interactive Communication, 177–182. Bartneck, C., Kanda, T., Ishiguro, H., & Hagita, N. (2007). Is the uncanny valley an uncanny cliff?. IEEE International Symposium on Robot and Human Interactive Communication, 368–373. Bartneck, C., Kanda, T., Ishiguro, H., & Hagita, N. (2009, September). My robotic doppelgänger-A critical look at the uncanny valley. In RO-MAN 2009-The 18th IEEE International Symposium on Robot and Human Interactive Communication (pp. 269–276). IEEE. Bartneck, C., Kuli´c, D., Croft, E., & Zoghbi, S. (2009). Measurement instruments for the anthropomorphism, animacy, likeability, perceived intelligence, and perceived safety of robots. International Journal of Social Robotics, 1(1), 71–81. Bartneck, C., Kanda, T., Ishiguro, H., & Hagita, N. (2009). My robotic doppelgänger—A critical look at the uncanny valley theory. Proceedings of the IEEE International Symposium on Robot and Human Interactive Communication, 269–276. Beane, M., & Orlikowski, W. J. (2015). What difference does a robot make? The material enactment of distributed coordination. Organization Science, 26(6), 1553–1573. Behrman, D. N., & Perreault, W. D. (1982). Measuring the performance of industrial salespersons. Journal of Business Research, 10, 355–370. Belk, R. W. (2013). Extended self in a digital world. Journal of Consumer Research, 40(3), 477–500. Belk, R. (2016). Anthropomorphism and Anthropocentrism. ACR North American Advances. Belk, R. (2016). Understanding the Robot: Comments on goudey and bonnin. Recherche et Applications en Marketing, 31(4), 83–90. Benbasat, I., & Barki, H. (2007). Quo vadis TAM?. Journal of the Association for Information Systems, 8(4), 211–218. Benninghoff, B., Kulms, P., Hoffmann, L., & Krämer, N. (2013). Theory of mind in humanrobot-communication: Appreciated or not? Kognitive Systeme, 1. Bergkvist, L., & Rossiter, J. R. (2007). The predictive validity of multiple-item versus singleitem measures of the same constructs. Journal of Marketing Research, 44(2), 175–184. Bernhardt, K. L., Donthu, N., & Kennett, P. A. (2000). A longitudinal analysis of satisfaction and profitability. Journal of Business Research, 47(2), 161–171. Bernstein, P., Paolone, N., Higner, J., Gerbasi, K., Conway, S., Privitera, A., & Scaletta, L. (2008). Furries from A to Z (Anthropomorphism to Zoomorphism). Society & Animals, 16(3), 197–222. Berry, L. L. and Parasuraman, A. (1991). Marketing Services: Competing Through Quality. New York: Free Press. Bettencourt, L. A., & Brown, S. W. (2003). Role stressors and customer-oriented boundaryspanning behaviors in service organizations. Journal of the Academy of Marketing Science, 31(4), 394–408.
204
References
Bhattacherjee, A. (2001a). Understanding information systems continuance: an expectationconfirmation model. MIS Quarterly, 25(3), 351–370. Bhattacherjee, A. (2001b). An empirical analysis of the antecedents of electronic commerceservice continuance. Decision Support Systems, 32(2), 201–214. Bhattacherjee, A., & Premkumar, G. (2004). Understanding changes in belief and attitude toward information technology usage: A theoretical model and longitudinal test. MIS Quarterly, 28(2), 229–254. Biddle, B. (1979). Role theory: Expectations, identities, and behaviors. New York: Academic Press. Biddle, B. (1986). Recent developments in role theory. Annual Review of Sociology, 12(1), 67–92. Bieling, G., Stock, R., & Dorozalla, F. (2015). Coping with demographic change in job markets: How age diversity management contributes to organizational performance. Zeitschrift für Personalforschung, 29(1), 5–30. Biggs, G., & MacDonald, B. (2003). A survey of robot programming systems. Proceedings of the Australasian Conference on Robotics and Automation, 1–3. Bitner, M. J. (2017). Service research: rigor, relevance, and community. Journal of Service Research, 20(2), 103–104. Bitner, M. J., Booms, B. H., & Tetreault, M. S. (1990). The service encounter: diagnosing favorable and unfavorable incidents. Journal of Marketing, 54(1), 71–84. Bitner, M. J., Booms, B. H., & Mohr, L. A. (1994). Critical service encounters: The employee’s viewpoint. Journal of Marketing, 58(4), 95–106. Blascovich, J., Loomis, J., Beall, A. C., Swinth, K. R., Hoyt, C. L., & Bailenson, J. N. (2002). Immersive virtual environment technology as a methodological tool for social psychology. Psychological inquiry, 13(2), 103–124. Bloemer, J. M. M., & Odekerken-Schröder, G. J. (2002). Store satisfaction and store loyalty explained by customer-and store related factors. Journal of Consumer Satisfaction, Dissatisfaction and Complaining Behavior, 15, 68–80. Bollen, K., & Lennox, R. (1991). Conventional wisdom on measurement: A structural equation perspective. Psychological bulletin 110(2), 305. Bolton, R. N. (1998). A dynamic model of the duration of the customer’s relationship with a continuous service provider: The role of satisfaction. Marketing Science, 17(1), 45–65. Bolton, R., & Saxena-Iyer, S. (2009). Interactive services: A framework, synthesis and research directions. Journal of Interactive Marketing, 23(1), 91–104. Bolton, R. N., McColl-Kennedy, J. R., Cheung, L., Gallan, A., Orsingher, C., Witell, L., & Zaki, M. (2018). Customer experience challenges: bringing together digital, physical and social realms. Journal of Service Management. Bonifield, C., & Cole, C. (2007). Affective responses to service failure: Anger, regret, and retaliatory versus conciliatory responses. Marketing Letters, 18(1), 85–99. Booms, B. H., & Nyquist, J. (1981). Analyzing the customer/firm communication component of the services marketing mix. Marketing of Services: 1981 Special Educators’ Conference Proceedings, James H. Donnelly and William R. George, eds. Chicago: American Marketing Association. Bowden, J. L. H. (2009). The process of customer engagement: A conceptual framework. Journal of marketing theory and practice, 17(1), 63–74.
References
205
Boyd, B. K., Gove, S., & Hitt, M. A. (2005). Construct measurement in strategic management research: Illusion or reality?. Strategic Management Journal, 26(3), 239–257. Breazeal, C. (2003). Emotion and sociable humanoid robots. International Journal of HumanComputer Studies, 59(1–2), 119–155. Breazeal, C. (2005). Socially intelligent robots. Interactions, 12(2), 19–22. Brissett, D., & Edgley, C. (Eds.). (2005). Life as Theater: A Dramaturgical Sourcebook. Transaction Publishers. Broadbent, E. (2017). Interactions with robots: The truths we reveal about ourselves. Annual review of psychology, 68, 627–652. Broadbent, E., Stafford, R., & MacDonald, B. (2009). Acceptance of Healthcare Robots for the Older Population: Review and Future Directions, International Journal of Social Robotics, 1(4), 319–330. Broadbent, E., Kuo, I. H., Lee, Y. I., Rabindran, J., Kerse, N., Stafford, R., & MacDonald, B. A. (2010). Attitudes and reactions to a healthcare robot. Telemedicine and e-Health, 16(5), 608–613. Brooks, R. (1986). A robust layered control system for a mobile robot. IEEE Journal on Robotics and Automation, 2(1), 14–23. Brooks, R. A. (2002). Robot: The future of flesh and machines. London: Penguin Books. Brooks, R. (2014). More robots won’t mean fewer jobs. Harvard Business Review. Retrieved August 2, 2019 from https://hbr.org/2014/06/more-robots-wontmean-fewer-jobs. Brown, A. M. (2005). A new software for carrying out one-way ANOVA post hoc tests. Computer Methods and Programs in Biomedicine, 79(1), 89–95. Brown, S. A., Venkatesh, V., & Goyal, S. (2012). Expectation confirmation in technology use. Information Systems Research, 23(2), 474–487. Brown, S. A., Venkatesh, V., & Goyal, S. (2014). Expectation confirmation in information systems research: A test of six competing models. MIS Quarterly, 38(3), 729–756. Bruns, S. B., & Ioannidis, J. P. (2016). P-curve and p-hacking in observational research. PloS one, 11(2), e0149144. Brynjolfsson, E., & Mitchell, T. (2017). What can machine learning do? Workforce implications. Science, 358(6370), 1530–1534. Buhrmester, M., Kwang, T., & Gosling, S. D. (2011). Amazon’s Mechanical Turk: A new source of cheap, yet high-quality, data?. Perspectives on Psychological Science, 6, 3–5. Bush, V. D., Rose, G. M., Gilbert, F., & Ingram, T. N. (2001). Managing culturally diverse buyer-seller relationships: The role of intercultural disposition and adaptive selling in developing intercultural communication competence. Journal of the Academy of Marketing Science, 29(4), 391–404. Cadotte, E., Woodruff, R., & Jenkins, R. (1987). Expectation and norms in models of consumer satisfaction. Journal of Marketing Research, 24(3), 305–314. ˇ c, M., Odekerken-Schröder, G., & Mahr, D. (2018). Service robots: value co-creation and Cai´ co-destruction in elderly care networks. Journal of Service Management, 29(2), 178–205. Cannon, J. P., & Perreault Jr, W. D. (1999). Buyer-seller relationships in business markets. Journal of Marketing Research, 36(4), 439–460. ˇ Capek, K. (2001). Rossum’s Universal Robots (P. Selver & N. Playfair, Trans.): Dover Publications. Carlson, L., & Carlson, R. (1984). Affect and psychological magnification: Derivations from Tomkins’ script theory. Journal of Personality, 52(1), 36–45.
206
References
Carpenter, J., Davis, J. M., Erwin-Stewart, N., Lee, T. R., Bransford, J. D., & Vye, N. (2009). Gender representation and humanoid robots designed for domestic use. International Journal of Social Robotics, 1(3), 261–265. Carrefour (2017). Customers Will be Able to Have Fun with Pepper the Robot, Website, http://www.carrefour.com/current-news/customers-will-be-able-to-have-fun-with-pep per-the-robot, accessed 30.05.2018. Carter, E. J., Mahler, M., & Hodgins, J. K. (2013). Unpleasantness of animated characters corresponds to increased viewer attention to faces. Proceedings of the ACM Symposium on Applied Perception, 35–40. Ceccarelli, M. (2004). Fundamentals of the mechanics of robots. In Fundamentals of Mechanics of Robotic Manipulation. Dordrecht: Springer. Chang, S., Gong, Y., & Shum, C. (2011). Promoting innovation in hospitality companies through human resource management practices. International Journal of Hospitality Management, 30(4), 812–818. Chatterjee, S., & Price B. (1991). Regression Analysis by Example. New York John Wiley. Chebat, J. C., & Kollias, P. (2000). The impact of empowerment on customer contact employees’ roles in service organizations. Journal of Service Research, 3(1), 66–81. Cheetham, M., Suter, P., & Jancke, L. (2014). Perceptual discrimination difficulty and familiarity in the uncanny valley: more like a “happy valley”. Frontiers in Psychology, 5, 1219. Chen, Y. Y., Huang, H. L., Hsu, Y. C., Tseng, H. C., & Lee, Y. C. (2010). Confirmation of expectations and satisfaction with the Internet shopping: The Role of Internet selfefficacy. Computer and Information Science, 3(3), 14–22. Choi, S., & Mattila, A. S. (2008). Perceived controllability and service expectations: Influences on customer reactions following service failure. Journal of Business Research, 61(1), 24–30. CHR Citadelle (2016). Rapport Annuel, Web-Document, 2016, https://www.chrcitadelle. be/CitadelleWebsite/media/Documents/Rapports%20annuels/Rapport-annuel-2016.pdf, accessed 30.05.2018. Christensen, K., Nørskov, S., Frederiksen, L., & Scholderer, J. (2017). In search of new product ideas: Identifying ideas in online communities by machine learning and text mining. Creativity and Innovation Management, 26(1), 17–30. Christov-Moore, L., Simpson, E. A., Coudé, G., Grigaityte, K., Iacoboni, M., & Ferrari, P. F. (2014). Empathy: Gender Effects in Brain and Behavior. Neuroscience & Biobehavioral Reviews, 46, 604–627. Churchill Jr, G. A. (1979). A paradigm for developing better measures of marketing constructs. Journal of Marketing Research 16(1), 64–73. Churchill Jr, G. A., & Peter, J. P. (1984). Research design effects on the reliability of rating scales: A meta-analysis. Journal of Marketing Research, 21(4), 360–375. Churchill Jr., G. A., & Surprenant, C. (1982). An investigation into the determinants of customer satisfaction. Journal of Marketing Research, 19(4), 491–504. Chuttur, M. Y. (2009). Overview of the technology acceptance model: Origins, developments and future directions. Working Papers on Information Systems, 9(37), 9–37. Cicchetti, D. V. (1994). Guidelines, criteria, and rules of thumb for evaluating normed and standardized assessment instruments in psychology. Psychological Assessment 6(4), 284– 290.
References
207
Coelho, F., & Augusto, M. (2010). Job characteristics and the creativity of frontline service employees. Journal of Service Research, 13(4), 426–438. Coelho, F., Augusto, M., & Lages, L. F. (2011). Contextual factors and the creativity of frontline employees: The mediating effects of role stress and intrinsic motivation. Journal of Retailing, 87(1), 31–45. Cohen, J., Cohen, P., West, S., & Aiken, L. (2003). Applied Multiple Regression/Correlation Analysis for the Behavioral Sciences, vol. 3, Mahwah. Collins, G. R. (2020). Improving human–robot interactions in hospitality settings. International Hospitality Review, 34(1), 61–79. Conti, D., Di Nuovo, S., Buono, S., & Di Nuovo, A. (2017). Robots in education and care of children with developmental disabilities: a study on acceptance by experienced and future professionals. International Journal of Social Robotics, 9(1), 51–62. Cortina, J. M. (1993). What is coefficient alpha? An examination of theory and applications. Journal of Applied Psychology 78(1), 98. Crnkovic-Friis, L., & Crnkovic-Friis, L. (2016). Generative choreography using deep learning. Proceedings of the 7th International Conference on Computational Creativity, 272–277. Cronbach, L. J. (1951). Coefficient alpha and the internal structure of tests. Psychometrika 16(3), 297–334. Csapo, A., Gilmartin, E., Grizou, J., Han, J., Meena, R., Anastasiou, D., ... & Wilcock, G. (2012). Multimodal conversational interaction with a humanoid robot. In 2012 IEEE 3rd International Conference on Cognitive Infocommunications, 667–672. Cummings, W. H., & Venkatesan, M. (1976). Cognitive dissonance and consumer behavior: a review of the evidence. Journal of Marketing Research, 13(3), 303–308. Dabholkar, P. A., & Spaid, B. I. (2012). Service failure and recovery in using technology-based self-service: effects on user attributions and satisfaction. The Service Industries Journal, 32(9), 1415–1432. Dahlbäck, N., Jönsson, A., & Ahrenberg, L. (1993). Wizard of Oz studies—why and how. Knowledge-Based Systems, 6(4), 258–266. Dautenhahn, K. (1999, August). Robots as social actors: Aurora and the case of autism. In Proc. CT99, The Third International Cognitive Technology Conference, August, San Francisco (Vol. 359, p. 374). Davidshofer, K. R., & Murphy, C. O. (2005). Psychological testing: principles and applications. Davis, M. H. (1983). Measuring individual differences in empathy: Evidence for a multidimensional approach. Journal of Personality and Social Psychology, 44, 113–126. Davis, F. D. (1989). Perceived usefulness, perceived ease of use, and user acceptance of information technology. MIS Quarterly, 13(3), 319–340. Davis, F. D., Bagozzi, R. P., & Warshaw, P. R. (1989). User acceptance of computer technology: a comparison of two theoretical models. Management Science, 35(8), 982–1003. Dawar, N., Parker, P. M., & Price, L. J. (1996). A cross-cultural study of interpersonal information exchange. Journal of International Business Studies, 27(3), 497–516. de Graaf, M. M., & Allouch, S. B. (2013). Exploring influencing variables for the acceptance of social robots. Robotics and Autonomous Systems, 61(12), 1476–1486. De Jong, J. P., & Kemp, R. (2003). Determinants of co-workers’ innovative behaviour: An investigation into knowledge intensive services. International Journal of Innovation Management, 7(2), 189–212.
208
References
DeVellis, R. F. (2016). Scale development: Theory and applications (vol. 26). Sage publications. de Wit, C. C., Siciliano, B., & Bastin, G. (2012). Theory of robot control. Springer Science & Business Media. Dhar, R. L. (2016). Ethical leadership and its impact on service innovative behavior: The role of LMX and job autonomy. Tourism Management, 57, 139–148. Diamantopoulos, A., & Winklhofer, H. M. (2001). Index construction with formative indicators: An alternative to scale development. Journal of Marketing Research 38(2), 269–277. Diamantopoulos, A., Sarstedt, M., Fuchs, C., Wilczynski, P., & Kaiser, S. (2012). Guidelines for choosing between multi-item and single-item scales for construct measurement: a predictive validity perspective. Journal of the Academy of Marketing Science, 40(3), 434– 449. DiSalvo, C., & Gemperle, F. (2003). From seduction to fulfillment: the use of anthropomorphic form in design. Proceedings of the 2003 International Conference on Designing Pleasurable Products and Interfaces, 67–72. Domingues, E., Lau, N., Pimentel, B., Shafii, N., Reis, L. P., & Neves, A. J. (2011). Humanoid behaviors: from simulation to a real robot. In Portuguese Conference on Artificial Intelligence, 352–364. Doney, P. M. & Cannon, J. P. (1997). An examination of the nature of trust in buyer–seller relationships. Journal of Marketing, 61, 35–51. Dubinsky, A. J., Howell, R. D., Ingram, T. N., & Bellenger, D. N. (1986). Salesforce socialization. Journal of Marketing, 50(4), 192–207. Duffy, D. L. (2003). Internal and external factors which affect customer loyalty. Journal of Consumer Marketing, 20(5), 480–485. Durand, V. M., & Barlow, D. H. (2012). Essentials of Abnormal Psychology. Cengage Learning. Edwards, J. R. (1994). The study of congruence in organizational behavior research: critique and a proposed alternative. Organizational Behavior and Human Decision Processes, 58(1), 51–100. Edwards, J. R. (2002). Alternatives to difference scores: Polynomial regression analysis and response surface methodology. In F. Drasgow & N. W. Schmitt (Eds.), Advances in Measurement and Data Analysis. Jossey-Bass, San Francisco, CA, 350–400. Edwards, A. (2018). Animals, humans, and machines: Interactive implications of ontological classification. Human-Machine Communication: Rethinking Communication, Technology, and Ourselves, 29–50. Edwards, J. R., & Harrison, R. V. (1993). Job demands and worker health: three-dimensional reexamination of the relationship between person-environment fit and strain. Journal of Applied Psychology, 78(4), 628–648 Edwards, J. R., & Parry, M. E. (1993). On the use of polynomial regression equations as an alternative to difference scores in organizational research. Academy of Management Journal, 36(6), 1577–1613. Edwards, A., Edwards, C., Spence, P. R., Harris, C., & Gambino, A. (2016). Robots in the classroom: differences in students’ perceptions of credibility and learning between “teacher as robot” and “robot as teacher”. Computers in Human Behavior, 65, 627–634.
References
209
Edwards, A., Edwards, C., & Gambino, A. (2020). The social pragmatics of communication with social robots: Effects of robot message design logic in a regulative context. International Journal of Social Robotics, 12(4), 945–957. Ehrlich, D., Guttman, I., Schönbach, P., & Mills, J. (1957). Postdecision exposure to relevant information. The Journal of Abnormal and Social Psychology, 54(1), 98. Engel, J. F. (1963). Are automobile purchasers dissonant consumers?. Journal of Marketing, 27(2), 55–58. Engelhardt, K. G. & Edwards, R. A. (1992). Human-robot integration for service robotics. In Mansour, R. & Karwowski, W. (eds.). Human-Robot Interaction, 315–326, London: Taylor & Francis Ltd. Epley, N., Waytz, A., & Cacioppo, J. T. (2007). On seeing human: A three-factor theory of anthropomorphism. Psychological Review, 114, 864–886. Erasmus, A. C., Bishoff, E., & Rousseau, G. G. (2002). The potential of using script theory in consumer behaviour research. Journal of Family Ecology and Consumer Sciences, 30(1), 1–9. Evans, J., Krishnamurthy, B., Pong, W., Croston, R., Weiman, C., & Engelberger, G. (1989). HelpMate™: A robotic materials transport system. Robotics and Autonomous Systems, 5(3), 251–256. Evers, V., Maldonado, H., Brodecki, T., & Hinds, P. (2008). Relational vs. group self-construal: Untangling the role of national culture in HRI. IEEE International Conference on HumanRobot Interaction, 255–262. Eyssel, F., & Hegel, F. (2012). (S)he’s got the look: Gender stereotyping of robots. Journal of Applied Social Psychology, 42(9), 2213–2230. Ezer, N., Fisk, A. D., & Rogers, W. A. (2009). Attitudinal and intentional acceptance of domestic robots by younger and older adults. International Conference on Universal Access in Human-Computer Interaction, 39–48. Falces, C., Sierra, B., Briñol, P., & Horcajo, J. (2002). Alteraciones del script y juicios afectivos: la satisfacción del consumidor. Psicothema, 14(3), 623–629. Fasola, J., & Mataric, M. (2011). Comparing physical and virtual embodiment in a socially assistive robot exercise coach for the elderly. Center for Robotics and Embedded Systems, Los Angeles, CA. Faul, F., Erdfelder, E., Lang, A. G., & Buchner, A. (2007). G* Power 3: A flexible statistical power analysis program for the social, behavioral, and biomedical sciences. Behavior Research Methods, 39(2), 175–191. Faul, F., Erdfelder, E., Buchner, A., & Lang, A. G. (2009). Statistical power analyses using G* Power 3.1: Tests for correlation and regression analyses. Behavior Research Methods, 41(4), 1149–1160. Fernández-Llamas, C., Conde, M. Á., Rodríguez-Sedano, F. J., Rodríguez-Lera, F. J., & Matellán-Olivera, V. (2017). Analysing the computational competences acquired by K-12 students when lectured by robotic and human teachers. International Journal of Social Robotics, 1–11. Fernando, Yudi & Mathath, Anas & Murshid, Mohsen. (2017). Improving Productivity. https:// doi.org/10.4018/978-1-5225-1759-7.ch107. Ferrucci, D., Levas, A., Bagchi, S., Gondek, D., & Mueller, E. T. (2013). Watson: beyond jeopardy!. Artificial Intelligence, 199, 93–105.
210
References
Ferrús, R. M., & Somonte, M. D. (2016). Design in robotics based in the voice of the customer of household robots. Robotics and Autonomous Systems, 79, 99–107. Festinger, L. A. (1957). A Theory of Cognitive Dissonance, Stanford, CA: Stanford University Press. Festinger, L. (1957). A theory of cognitive dissonance. Evanston, IL: Row,Peterson, & Co Festinger, L. (1962). Cognitive dissonance. Scientific American, 207(4), 93–106. Field, A. (2013). Discovering statistics using IBM SPSS statistics. Sage Publications. Finin, T., Murnane, W., Karandikar, A., Keller, N., Martineau, J., & Dredze, M. (2010). Annotating named entities in Twitter data with crowdsourcing. Proceedings of the NAACL HLT Workshop on Creating Speech and Language Data with Amazon’s Mechanical Turk, 80–88. Finn, A. (2005). Reassessing the foundations of customer delight. Journal of Service Research, 8(2), 103–116. Fishbein, M., & Ajzen, I. (1975). Belief, Attitude, Intention, and Behavior: An Introduction to Theory and Research. Reading, MA, Addison-Wesley. Fisher, R. A. (1935). The design of experiments, Oliver and Boyd, Edinburgh and London. Fisher, R. A. (1954). Statistical Methods for Research Workers, vol. 12, Oliver and Boyd, Edinburgh. Fisk, R. P., Brown, S. W., & Bitner, M. J. (1993). Tracking the evolution of the services marketing literature. Journal of Retailing, 69(1), 61–103. Fleenor, J. W., McCauley, C. D., & Brutus, S. (1996). Self-other rating agreement and leader effectiveness. The Leadership Quarterly, 7(4), 487–506. Fleming, P. (2019). Robots and organization studies: Why robots might not want to steal your job. Organization Studies, 40(1), 23–38. Fogg, B. J., & Nass, C. (1997a). How users reciprocate to computers: An experiment that demonstrates behavior change. CHI’97 Extended Abstracts on Human Factors in Computing Systems, 331–332. Fogg, B. J., & Nass, C. (1997b). Silicon sycophants: The effects of computers that flatter. International Journal of Human-Computer Studies, 46(5), 551–561. Folkes, V. S., Koletsky, S. & Graham, J. L. (1987). A field study of causal inferences and consumer reaction: The view from the airport. Journal of Consumer Research, 13(4), 534–539. Fong, T., Nourbakhsh, I., & Dautenhahn, K. (2003). A survey of socially interactive robots. Robotics and Autonomous Systems, 42(3–4), 143–166. Ford, C. M., & Gioia, D. A. (2000). Factors influencing creativity in the domain of managerial decision making. Journal of Management, 26(4), 705–732. Fornell, C. (1986). A Second Generation of Multivariate Analysis: Classification of Methods and Implications for Marketing Research, Working Paper, University of Michigan. Ann Arbor. Fornell, C., & Larcker, D. F. (1981). Evaluating structural equation models with unobservable variables and measurement error. Journal of Marketing Research 18(1), 39–50. Foster, M. E., Gaschler, A., & Giuliani, M. (2017). Automatically classifying user engagement for dynamic multi-party human–robot interaction. International Journal of Social Robotics, 9(5), 659–674. Fournier, S., & Mick, D. (1999). Rediscovering Satisfaction. Journal of Marketing, 63(4), 5–23.
References
211
Fox, J., Ahn, S. J., Janssen, J. H., Yeykelis, L., Segovia, K. Y., & Bailenson, J. N. (2015). Avatars versus agents: a meta-analysis quantifying the effect of agency on social influence. Human–Computer Interaction, 30(5), 401–432. Franzoi, S. L. (1996). Social psychology. Madison, WI: Brown and Benchmark. Frey, C. B., & Osborne, M. A. (2017). The future of employment: How susceptible are jobs to computerisation?. Technological Forecasting and Social Change, 114, 254–280. Friedman, V. J. (2002). The individual as agent of organizational learning. California Management Review, 44(2), 70–89. Friedrichs, J. (1990). Methoden Empirischer Sozialforschung. Springer-Verlag. Fullerton, G., & Taylor, S. (2002). Mediating, interactive, and non-linear effects in service quality and satisfaction with services research. Canadian Journal of Administrative Sciences/Revue Canadienne des Sciences de l’Administration, 19(2), 124–136. Gaitanides, M., & Stock, R. (2004). Interorganisationale Teams: Transaktionskostentheoretische Überlegungen und empirische Befunde zum Teamerfolg. Zeitschrift für betriebswirtschaftliche Forschung, 56(8), 436–451. Gai, S., Jung, E. J., & Yi, B. J. (2016). Multi-group localization problem of service robots based on hybrid external localization algorithm with application to shopping mall environment. Intelligent Service Robotics, 9(3), 257–275. Gambino, A., Fox, J., & Ratan, R. A. (2020). Building a stronger CASA: extending the computers are social actors paradigm. Human-Machine Communication, 1(1), 5. Garg, S., & Dhar, R. (2017). Employee service innovative behavior: the roles of leadermember exchange (LMX), work engagement, and job autonomy. International Journal of Manpower, 38(2), 242–258. Gaukroger, S. (2012). Objectivity: A very short introduction (no. 316). Oxford University Press. Gbadamosi, A. (2009), “Cognitive dissonance: the implicit explication in low-income consumers’ shopping behaviour for low-involvement grocery products”, International Journal of Retail and Distribution Management, Vol. 37 No. 12, pp. 1077–1095. Gelin, R., d’Alessandro, C., Le, Q. A., Deroo, O., Doukhan, D., Martin, J. C., ... & Rosset, S. (2010). Towards a storytelling humanoid robot. In 2010 AAAI Fall Symposium Series. Gerbing, D. W., & Anderson, J. C. (1988). An updated paradigm for scale development incorporating unidimensionality and its assessment. Journal of Marketing Research 25(2), 186–192. Gerhart, B., & Fang, M. (2005). National culture and human resource management: assumptions and evidence. The International Journal of Human Resource Management, 16(6), 971–986. Gerlach, J., Stock, R. M., & Buxmann, P. (2014). Never forget where you’re coming from: The role of existing products in adoptions of substituting technologies. Journal of Product Innovation Management, 31, 133–145. Ghazali, A. S., Ham, J., Barakova, E., & Markopoulos, P. (2020). Persuasive Robots Acceptance Model (PRAM): roles of social responses within the acceptance model of persuasive robots. International Journal of Social Robotics, 1–18. Gibson, C. B., Cooper, C. D., & Conger, J. A. (2009). Do you see what we see? The complex effects of perceptual distance between leaders and teams. Journal of Applied Psychology, 94(1), 62.
212
References
Ginzberg, M. J. (1981). Early diagnosis of MIS implementation failure: promising results and unanswered questions. Management Science, 27(4), 459–478. Gockley, R., Bruce, A., Forlizzi, J., Michalowski, M., Mundell, A., Rosenthal, S., ... & Wang, J. (2005). Designing robots for long-term social interaction. In 2005 IEEE/RSJ International Conference on Intelligent Robots and Systems, 1338–1343. Götz, S. (2019). Bitte entschuldige mein Unvermögen. Zeit Online, 12.06.2019, accessible at https://www.zeit.de/mobilitaet/2019-06/roboter-semmi-deutsche-bahn-kundenservicetest. Gould, J. D., Conti, J., & Hovanyecz, T. (1983). Composing letters with a simulated listening typewriter. Communications of the ACM, 26(4), 295–308. Graebner, M. E., & Eisenhardt, K. M. (2004). The seller’s side of the story: Acquisition as courtship and governance as syndicate in entrepreneurial firms. Administrative Science Quarterly, 49(3), 366–403. Graefe, A. (2016). Guide to automated journalism. New York: Tow Center for Digital Journalism, Columbia University. accessible at: http://towcenter.org/research/guide-to-automa ted-journalism/. Graf, B., Reiser, U., Hägele, M., Mauz, K., & Klein, P. (2009). Robotic home assistant CareO-bot® 3-product vision and innovation platform. IEEE Workshop on Advanced Robotics and its Social Impacts, 139–144. Green, R. D., MacDorman, K. F., Ho, C. C., & Vasudevan, S. (2008). Sensitivity to the proportions of faces that vary in human likeness. Computers in Human Behavior, 24(5), 2456–2474. Green, T., Hartley, N., & Gillespie, N. (2016). Service provider’s experiences of service separation: the case of telehealth. Journal of Service Research, 19(4), 477–494. Grewal, D., Levy, M., & Kumar, V. (2009). Customer experience management in retailing: An organizing framework. Journal of Retailing, 85(1), 1–14. Gupta, P., Tirth, V., & Srivastava, R. K. (2006, August). Futuristic humanoid robots: An overview. In First International Conference on Industrial and Information Systems, 247– 254. Hair, B., & Babin, A. T.(2006). Multivariate data analysis. Upper Saddle River, NJ. Hair, J. F., Anderson, R. E., Babin, B. J., & Black, W. C. (2010). Multivariate data analysis: A global perspective, vol. 7, Pearson Upper Saddle River. Hanson, D. (2006). Exploring the aesthetic range for humanoid robots. Proceedings of the ICCS/CogSci-2006 Long Symposium: Toward Social Mechanisms of Android Science, 39–42. Hanson, D., & Grimmer, M. (2007). The mix of qualitative and quantitative research in major marketing journals, 1993-2002. European Journal of Marketing, 41(1/2), 58–70. Hanson, D., Olney, A., Prilliman, S., Mathews, E., Zielke, M., Hammons, D., ... & Stephanou, H. (2005). Upending the uncanny valley. Proceedings of the American Association for Artificial Intelligence (AAAI), 5, 1728–1729. Han, M. J., Lin, C. H., & Song, K. T. (2012). Robotic emotional expression generation based on mood transition and personality model. IEEE Transactions on Cybernetics, 43(4), 1290–1303. Han, J., Campbell, N., Jokinen, K., & Wilcock, G. (2012). Investigating the use of nonverbal cues in human-robot interaction with a Nao robot. In 2012 IEEE 3rd International Conference on Cognitive Infocommunications, 679–683.
References
213
Haring, K. S., Watanabe, K., & Mougenot, C. (2013). The influence of robot appearance on assessment. ACM/IEEE International Conference on Human-Robot Interaction (HRI), 131–132. Harmon-Jones, E., & Mills, J. (1999). An introduction to cognitive dissonance theory and an overview of current perspectives on the theory, in Eddie Harmon-Jones and Judson Mills (eds), Cognitive dissonance: progress on a pivotal theory in social psychology, American Psychological Association, Washington, DC. Harris, K. E., Mohr, L. A., & Bernhardt, K. L. (2006). Online service failure—consumer attributions and expectations. Journal of Services Marketing, 20(7), 453–458. Harris, M. M., Anseel, F., & Lievens, F. (2008). Keeping up with the Joneses: A field study of the relationships among upward, lateral, and downward comparisons and pay level satisfaction. Journal of Applied Psychology 93(3), 665. Heine, S. J., & Lehman, D. R. (1997). Culture, dissonance, and self-affirmation. Personality and Social Psychology Bulletin, 23(4), 389–400. Hennig-Thurau, T., Groth, M., Paul, M., & Gremler, D. (2006). Are all smiles created equal? How emotional contagion and emotional labor affect service relationships. Journal of Marketing, 70(3), 58–73. Herrmann, A., & Johnson, M. D. (1999). Die Kundenzufriedenheit als Bestimmungsfaktor der Kundenbindung. Zeitschrift für betriebswirtschaftliche Forschung, 51(6), 576–598. Herrmann, A., & Landwehr, J.R. (2008). Varianzanalyse, in: Herrmann, A., Homburg, Ch., Klarmann, M. (Hrsg.), Handbuch Marktforschung: Methoden—Anwendungen—Praxisbeispiele, vol. 3, Wiesbaden, 579–606. Herrmann, P. N., Kundisch, D. O., & Rahman, M. S. (2015). Beating irrationality: does delegating to IT alleviate the sunk cost effect?. Management Science, 61(4), 831–850. Hess, R. L., Ganesan, S., & Klein, N. M. (2003). Service failure and recovery: The impact of relationship factors on customer satisfaction. Journal of the Academy of Marketing Science, 31(2), 127–145. Hildebrandt, L. (1998). Kausalanalytische Validierung in der Marketingforschung, in: Hildebrandt, L., & Homburg, C., Die Kausalanalyse: Ein Instrument der empirischen betriebswirtschaftlichen Forschung, Stuttgart, 85–110. Himme, A. (2009). Gütekriterien der Messung: Reliabilität, Validität und Generalisierbarkeit. Methodik der empirischen Forschung, Gabler Verlag, Wiesbaden, 485–500. Hinkle D. E., Wiersma, W., Jurs S. G. (1982). One-way analysis of variance, Houghton Miffin Companz, Boston, 251–279. Hockstein, N. G., Gourin, C. G., Faust, R. A., & Terris, D. J. (2007). A history of robots: from science fiction to surgical robotics. Journal of Robotic Surgery, 1(2), 113–118. Hoffman, M. L. (1977). Sex differences in empathy and related behaviors. Psychological Bulletin, 84(4), 712–722. Hoffman, D. L., & Novak, T. P. (2018). Consumer and object experience in the internet of things: An assemblage theory approach. Journal of Consumer Research, 44(6), 1178– 1204. Hofstede, G. (1980). Culture and organizations. International Studies of Management & Organization, 10(4), 15–41. Hofstede, G. (1984). Culture’s Consequences: International Differences in Work-Related Values, vol. 5, Sage Publications.
214
References
Hofstede, G. (2003). Culture’s Consequences: Comparing Values, Behaviors, Institutions and Organizations Across Nations, vol. 2, Sage Publications. Hofstede, G., & Hofstede, G. J. (2005). Cultures and Organizations Software of the Mind, New York, McGraw-Hill Publishing. Hofstede, G., & McCrae, R. R. (2004). Personality and culture revisited: Linking traits and dimensions of culture. Cross-Cultural Research, 38(1), 52–88. Hogan, J., & Hogan, R. (1984). How to Measure service orientation. Journal of Applied Psychology, 69, 167–173. Hogan, R. and Hogan, J., (1995). Hogan Personality Inventory Manual. Tulsa: Hogan Assessment Systems. Ho, V. T., & Gupta, N. (2012). Testing an empathy model of guest-directed citizenship and counterproductive behaviours in the hospitality industry: Findings from three hotels. Journal of Occupational and Organizational Psychology, 85(3), 433–453. Holloway, B. B., & Beatty, S. E. (2003). Service failure in online retailing: A recovery opportunity. Journal of Service Research, 6(1), 92–105. Holzwarth, M., Janiszewski, C., & Neumann, M. M. (2006). The influence of avatars on online consumer shopping behavior. Journal of Marketing, 70(4), 19–36. Homack, S. R. (2001). Understanding what ANOVA post hoc tests are, really, Annual Meeting of the Southwest Educational Research Association, New Orleans, LA. Homburg, C. (2000). Kundennähe von industriegüterunternehmen. Aufl., Wiesbaden, 13. Homburg, C. (2016). Marketingmanagement: Strategie-Instrumente-UmsetzungUnternehmensführung. Springer-Verlag. Homburg, N. (2018, January). How to include humanoid robots into experimental research: A multi-step approach. Proceedings of the 51st Hawaii International Conference on System Sciences, 4423–4432. Homburg, C., & Garbe, B. (1999). Towards an improved understanding of industrial services: quality dimensions and their impact on buyer-seller relationships. Journal of Business-toBusiness Marketing, 6(2), 39–71. Homburg, C., & Giering, A. (1996). Konzeptualisierung und Operationalisierung komplexer Konstrukte. Ein Leitfaden für die Marketingforschung. Marketing ZfP, 18(1), 5–24. Homburg, C., & Krohmer, H. (2006). Marketingmanagement–Strategie–Instrumente–Umsetzung–Unternehmensführung, Wiesbaden. Franz-Rudolf Esch, Elisabeth von Einem, Vanessa Rühl, 89. Homburg, C. & Stock, R. (2001). Der Zusammenhang zwischen Mitarbeiter- und Kundenzufriedenheit. Bestandsaufnahme und Entwicklung eines theoretischen Untersuchungsrahmens. Die Unternehmung, 55 (6), 377–400. Homburg, C., & Stock, R. M. (2004). The link between salespeople’s job satisfaction and customer satisfaction in a business-to-business context: A dyadic analysis. Journal of the Academy of Marketing Science, 32(2), 144–158. Homburg, C., & Stock, R. M. (2005). Exploring the conditions under which salesperson work satisfaction can lead to customer satisfaction. Psychology & Marketing, 22(5), 393–420. Homburg, C., Sieben, F., & Stock, R. (2004). Einflussgrößen des Kundenrückgewinnungserfolgs: Theoretische Betrachtung und Empirische Befunde. Marketing—Zeitschrift für Forschung und Praxis, 26(1), 25–41.
References
215
Homburg, C., Klarmann, M., & Krohmer, H. (2008). Statistische Grundlagen der Datenanalyse, in: Herrmann, A., Homburg, C., Klarmann, M. (eds.), Handbuch Marktforschung: Methoden—Anwendungen—Praxisbeispiele, vol. 3, Wiesbaden, 213–239. Homburg, C., Wieseke, J., & Hoyer, W. D. (2009). Social identity and the service-profit chain. Journal of Marketing, 73(2), 38–54. Honig, S.S. and Oron-Gilad, T. (2018). Understanding and resolving failures in human–robot interaction: literature review and model development. Frontiers in Psychology, 9, 1–21. Horton, J. J., Rand, D. G., & Zeckhauser, R. J. (2011). The online laboratory: Conducting experiments in a real labor market. Experimental Economics, 14, 399–425. House, R. J., Hanges, P. J., Javidan, M., Dorfman, P. W., & Gupta, V. (2004). Culture, Leadership, and Organizations: The GLOBE Study of 62 Societies, Sage Publications. Howard, John A. and Jagdish N. Sheth (1969). The Theory of Buyer Behavior. New York: John Wiley. Hsu, M. H., Yen, C. H., Chiu, C. M., & Chang, C. M. (2006). A longitudinal investigation of continued online shopping behavior: An extension of the theory of planned behavior. International Journal of Human-Computer Studies, 64(9), 889–904. Hsu, H. M., Hsu, J. S. C., Wang, S. Y., & Chang, I. C. (2016). Exploring the effects of unexpected outcome on satisfaction and continuance intention. Journal of Electronic Commerce Research, 17(3), 239–255. Hu, M. L. M. (2009). Knowledge sharing and innovative service behavior relationship: Guanxi as mediator. Social Behavior and Personality: an International Journal, 37(7), 977–992. Huang, M. H., & Rust, R. T. (2017). Technology-driven service strategy. Journal of the Academy of Marketing Science, 45(6), 906–924. Huang, M. H., & Rust, R. T. (2018). Artificial intelligence in service. Journal of Service Research, 21(2), 155–172. Hudson, J., Orviska, M., & Hunady, J. (2017). People’s attitudes to robots in caring for the elderly. International Journal of Social Robotics, 9(2), 199–210. Hu, H., & Gu, D. (2000). A multi-agent system for cooperative quadruped walking robots. Proceedings of the IASTED International Conference Robotics and Applications, 1–5. Hu, M. L. M., Horng, J. S., & Sun, Y. H. C. (2009). Hospitality teams: Knowledge sharing and service innovation performance. Tourism Management, 30(1), 41–50. Hüttner, M., & Schwarting, U. (2008). Exploratorische Faktorenanalyse, in: Herrmann, A., Homburg, C., & Klarmann, M. (eds.), Handbuch Marktforschung: Methoden, Anwendungen, Praxisbeispiele, vol. 3, Wiesbaden, 241–270. Hyytinen, A., & Toivanen, O. (2005). Do financial constraints hold back innovation and growth? Evidence on the role of public policy. Research Policy, 34(9), 1385–1403. International Federation of Robotics, IFR. (2017). https://ifr.org/downloads/press/Executive_ Summary_WR_Service_Robots_2017.pdf, accessed 01.09.2018. International Federation of Robotics (2019). World Robotics Service Robots, Report, [accessed 28.08.2020], https://ifr.org/worldrobotics/. Io, H. N., & Lee, C. B. (2020). Social Media Comments about Hotel Robots. Journal of China Tourism Research, 1–20. Ipeirotis, R. G., Provost, F., & Wang, J. (2010). Quality Management on Amazon Mechanical Turk. Proceedings of the ACM SIGKDD Workshop on Human Computation, 64–67. ISO 8373 (1994). Manipulating Industrial Robots—Vocabulary (EN/ISO 8373 1996).
216
References
ISO 8373 (2012). Robots and Robotic Devices. International Organization for Standardization, Geneva, Switzerland [https://www.iso.org/standard/55890.html]. Ivanov, S. H., Webster, C., & Berezina, K. (2017). Adoption of robots and service automation by tourism and hospitality companies. Revista Turismo & Desenvolvimento, 27(28), 1501– 1517. Ivanov, S., Webster, C., & Berezina, K. (2017). Adoption of robots and service automation by tourism and hospitality companies. INVTUR Conference, Aveiro, Portugal, 17–19. Ivanov, S., Gretzel, U., Berezina, K., Sigala, M. and Webster, C. (2019), “Progress on robotics in hospitality and tourism: a review of the literature”, Journal of Hospitality and Tourism Technology, 10(4), 489–421. Jackson, J. (1998). Contemporary criticisms of role theory. Journal of Occupational Science, 5(2), 49–55. Jacobs, T., & Graf, B. (2012). Practical evaluation of service robots for support and routine tasks in an elderly care facility. IEEE Workshop on Advanced Robotics and its Social Impacts (ARSO), 46–49. Janssen, O. (2000). Job demands, perceptions of effort-reward fairness and innovative work behaviour. Journal of Occupational and Organizational Psychology, 73(3), 287–302. Janssen, O., & Van Yperen, N. W. (2004). Employees’ goal orientations, the quality of leadermember exchange, and the outcomes of job performance and job satisfaction. Academy of Management Journal, 47(3), 368–384. Jarvis, C. B., MacKenzie, S. B., & Podsakoff, P. M. (2003). A critical review of construct indicators and measurement model misspecification in marketing and consumer research. Journal of Consumer Research 30(2), 199–218. Jayawardena, C., Kuo, I. H., Unger, U., Igic, A., Wong, R., Watson, C. I., ... & Sohn, J. (2010). Deployment of a service robot to help older people. IEEE/RSJ International Conference on Intelligent Robots and Systems, 5990–5995. Jeffrey J. Stoltman, Shelley R. Tapp, Richard S. Lapidus. 1989. An Examination of Shopping Scripts, in: H. Keith Hunt (Ed.), Advances in Consumer Research, vol. XVI, Association for Consumer Research, Ann Arbor, MI. Jehn, K. (1997). A qualitative analysis of conflict types and dimensions in organizational groups. Administrative Science Quarterly, 42(3), 530–557. Johnson, D., Gardner, J., & Wiles, J. (2004). Experience as a moderator of the media equation: the impact of flattery and praise. International Journal of Human-Computer Studies, 61(3), 237–258. Jörling, M., Böhm, R., & Paluch, S. (2019). Service robots: Drivers of perceived responsibility for service outcomes. Journal of Service Research, 22(4), 404–420. Kaartemo, V. and Helkkula, A. (2018), “A systematic review of artificial intelligence and robots in value co-creation: current status and future research avenues”, Journal of Creating Value, 4(2), 211–228. Kale, S. H., & Barnes, J. W. (1992). Understanding the domain of cross-national buyer-seller interactions. Journal of International Business Studies, 23(1), 101–132. Kanda, T., Sato, R., Saiwaki, N., & Ishiguro, H. (2007). A two-month field trial in an elementary school for long-term human–robot interaction. IEEE Transactions on Robotics, 23(5), 962–971. Kanda, T., Shiomi, M., Miyashita, Z., Ishiguro, H., & Hagita, N. (2010). A communication robot in a shopping mall. IEEE Transactions on Robotics, 26(5), 897–913.
References
217
Kaplan, F. (2004). Who is afraid of the humanoid? Investigating cultural differences in the acceptance of robots. International Journal of Humanoid Robotics, 1(3), 465–480. Karr-Wisniewski, P., & Prietula, M. (2010). CASA, WASA, and the dimensions of us. Computers in Human Behavior, 26(6), 1761–1771. Katagiri, Y., Nass, C., & Takeuchi, Y. (2001). Cross-cultural studies of the computers are social actors paradigm: The case of reciprocity. Usability Evaluation and Interface Design: Cognitive Engineering, Intelligent Agents, and Virtual Reality, 1558–1562. Kätsyri, J., Förger, K., Mäkäräinen, M., & Takala, T. (2015). A review of empirical evidence on different uncanny valley hypotheses: support for perceptual mismatch as one road to the valley of eeriness. Frontiers in Psychology, 6, 390. Keiningham, T. L., & Vavra, T. G. (2001). The customer delight principle: Exceeding customers’ expectations for bottom-line success. McGraw-Hill. Kelley, J. F. (1984). An iterative design methodology for user-friendly natural language office information applications. ACM Transactions on Information Systems, 2(1), 26–41. Kelly, F. and Huston, R. (1980). Recent advances in robotics research. Society of Automotive Engineers, Inc. Paper no. 800383. Kim, T. T., & Lee, G. (2013). Hospitality employee knowledge-sharing behaviors in the relationship between goal orientations and service innovative behavior. International Journal of Hospitality Management, 34, 324–337. Kim, S., & McGill, A. L. (2011). Gaming with Mr. Slot or gaming the slot machine? Power, anthropomorphism, and risk perception. Journal of Consumer Research, 38(1), 94–107. Kim, K. J., Park, E., & Sundar, S. S. (2013). Caregiving role in human–robot interaction: A study of the mediating effects of perceived benefit and social presence. Computers in Human Behavior, 29(4), 1799–1806. Kim, S., Chen, R. P., & Zhang, K. (2016). Anthropomorphized helpers undermine autonomy and enjoyment in computer games. Journal of Consumer Research, 43(2), 282–302. King, W. R., & He, J. (2006). A meta-analysis of the technology acceptance model. Information & Management, 43(6), 740–755. Kirby, R., Forlizzi, J., & Simmons, R. (2010). Affective social robots. Robotics and Autonomous Systems, 58(3), 322–332. Koo, T. K., & Li, M. Y. (2016). A guideline of selecting and reporting intraclass correlation coefficients for reliability research. Journal of Chiropractic Medicine 15(2), 155–163. Koschate, N. (2002). Kundenzufriedenheit und Preisverhalten: Theoretische und empirisch experimentelle Analysen, Wiesbaden. Krämer, N. C., Eimler, S., Von Der Pütten, A., & Payr, S. (2011). Theory of companions: what can theoretical models contribute to applications and understanding of human-robot interaction?. Applied Artificial Intelligence, 25(6), 474–502. Kumar, N., Scheer, L. K., & Steenkamp, J.-B. (1995). The effects of supplier fairness on vulnerable resellers. Journal of Marketing Research, 32(1), 54–65. Kuno, Y., Sadazuka, K., Kawashima, M., Yamazaki, K., Yamazaki, A., & Kuzuoka, H. (2007). Museum guide robot based on sociological interaction analysis. Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, 1191–1194. Kuo, I. H., Rabindran, J. M., Broadbent, E., Lee, Y. I., Kerse, N., Stafford, R. M. G., & MacDonald, B. A. (2009). Age and gender factors in user acceptance of healthcare robots. IEEE International Symposium on Robot and Human Interactive Communication, 214– 219.
218
References
Kvale, S. (1996). Interviews: An Introduction to Qualitative Research Interviewing, Thousand Oaks, California, SAGE Publications. Lankton, N. K., McKnight, D. H., Wright, R. T., & Thatcher, J. B. (2016). Research note— Using expectation disconfirmation theory and polynomial modeling to understand trust in technology. Information Systems Research, 27(1), 197–213. Lapré, M. A. (2011). Reducing customer dissatisfaction: How important is learning to reduce service failure?. Production and Operations Management, 20(4), 491–507. LaTour, S., & Peat, N. (1979). Conceptual and methodological issues in consumer satisfaction on research. Advances in Consumer Research, 6, 431–437. Lee, K., & Ashton, M. C. (2004). Psychometric properties of the HEXACO personality inventory. Multivariate Behavioral Research, 39(2), 329–358. Lee, K., & Ashton, M. C. (2009). The HEXACO personality inventory–revised. A Measure of the Six Major Dimensions of Personality. Pobrane z: http://hexaco.org/scaledescriptions. Lee, K. M., & Nass, C. (2003). Designing social presence of social actors in human computer interaction. Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, 289–296. Lee, H. R., Sung, J., Šabanovi´c, S., & Han, J. (2012). Cultural design of domestic robots: A study of user expectations in Korea and the United States. In 2012 IEEE RO-MAN: The 21st IEEE International Symposium on Robot and Human Interactive Communication, 803–808. Lee, C. S., Wang, M. H., Yen, S. J., Wei, T. H., Wu, I. C., Chou, P. C., ... & Yan, T. H. (2016). Human vs. computer go: Review and prospect. IEEE Computational Intelligence Magazine, 11(3), 67–72. Lee, N., Kim, J., Kim, E., & Kwon, O. (2017). The influence of politeness behavior on user compliance with social robots in a healthcare service setting. International Journal of Social Robotics, 9(5), 727–743. Leeflang, P. S. H., & Koerts, J. (1973). Modelling and marketing: two important concepts and the connection between them. European Journal of Marketing, 7(3), 203–217. Lewis, B. R., & P. McCann, P. (2004). Service failure and recovery: Evidence from the hotel industry. International Journal of Contemporary Hospitality Management, 16(1), 6–17. Li, M., & Hsu, C. H. (2016). A review of employee innovative behavior in services. International Journal of Contemporary Hospitality Management, 28(12), 2820–2841. Lindsey-Mullikin, J. (2003), “Beyond reference pricing: understanding consumers’ encounters with unexpected prices”, Journal of Product & Brand Management, 12(3), 140-53. Linnhoff-Popien, C., & Hofbauer, F. (2018). AIDA Cruises–Kreuzfahrt von morgen. Digitale Welt, 2(2), 18–24. Lin, H. H., Wang, Y. S., & Chang, L. K. (2011). Consumer responses to online retailer’s service recovery after a service failure: A perspective of justice theory. Managing Service Quality: An International Journal, 21(5), 511–534. Lin, H., Chi, O. H., & Gursoy, D. (2019). Improving human–robot interactions in hospitality settings. International Hospitality Review, 34(1), 1–20. Little, T. D., Lindenberger, U., & Nesselroade, J. R. (1999). On selecting indicators for multivariate measurement and modeling with latent variables: When “good” indicators are bad and “bad” indicators are good. Psychological Methods 4(2), 192. Li, D., Rau, P. P., & Li, Y. (2010). A cross-cultural study: Effect of robot appearance and task. International Journal of Social Robotics, 2(2), 175–186.
References
219
Lohse, M., Hegel, F., Swadzba, A., Rohlfing, K., Wachsmuth, S., & Wrede, B. (2007). What can I do for you? Appearance and application of robots. Proceedings of AISB, 7, 121–126. Looije, R., Neerincx, M. A., Peters, J. K., & Henkemans, O. A. B. (2016). Integrating robot support functions into varied activities at returning hospital visits. International Journal of Social Robotics, 8(4), 483–497. Looser, C. E., & Wheatley, T. (2010). The tipping point of animacy: how, when, and where we perceive life in a face. Psychological Science, 21(12), 1854–1862. Louloudi, A., Mosallam, A., Marturi, N., Janse, P., & Hernandez, V. (2010). Integration of the humanoid robot Nao inside a smart home: A case study. Proceedings of the Swedish AI Society Workshop, 48, 35–44. Louloudi, A., Mosallam, A., Marturi, N., Janse, P., & Hernandez, V. (2010). Integration of the humanoid robot Nao inside a smart home: A case study. In The Swedish AI Society Workshop at Uppsala University, 35–44. Lovelock, C., & Gummesson, E. (2004). Whither services marketing? In search of a new paradigm and fresh perspectives. Journal of Service Research, 7(1), 20–41. MacDorman, K. F. (2005). Androids as an experimental apparatus: Why is there an uncanny valley and can we exploit it. CogSci-2005 Workshop: Toward Social Mechanisms of Android Science, 3, 106–118. MacDorman, K. F., & Ishiguro, H. (2006). The uncanny advantage of using androids in cognitive and social science research. Interaction Studies, 7(3), 297–337. MacDorman, K. F., Green, R. D., Ho, C. C., & Koch, C. T. (2009). Too real for comfort? Uncanny responses to computer generated faces. Computers in Human Behavior, 25(3), 695–710. MacDorman, K. F., Vasudevan, S. K., & Ho, C. C. (2009). Does Japan really have robot mania? Comparing attitudes by implicit and explicit measures. AI & Society, 23(4), 485–510. MacKenzie, S. B., Podsakoff, P. M., & Jarvis, C. B. (2005). The problem of measurement model misspecification in behavioral and organizational research and some recommended solutions. Journal of Applied Psychology, 90(4), 710–730. Malhotra, Y., Galletta, D. F., & Kirsch, L. J. (2008). How endogenous motivations influence user intentions: Beyond the dichotomy of extrinsic and intrinsic user motivations. Journal of Management Information Systems, 25(1), 267–299. Manfrè, A., Infantino, I., Augello, A., Pilato, G., & Vella, F. (2017). Learning by demonstration for a dancing robot within a computational creativity framework. First IEEE International Conference on Robotic Computing, 434–439. Marinova, D., de Ruyter, K., Huang, M.-H., Meuter, M.L. and Challagalla, G. (2017), “Getting smart: learning from technology-empowered frontline interactions”, Journal of Service Research, 20(1), 29–42. Mason, M. T. (2001). Mechanics of robotic manipulation. MIT press. Mathur, M. B., & Reichling, D.B. (2016). Navigating a social world with robot partners: A qualitative cartography of the uncanny valley. Cognition, 146, 22–32. Mayr, S., & Kläsgen, M. (2017). Spaß mit Pepper. Süddeutsche Zeitung, 15.10.2017, accessible at https://www.sueddeutsche.de/wirtschaft/roboter-spass-mit-pepper-1.3709030. McCollough, M. A., Berry, L. L., & Yadav, M. S. (2000). An empirical investigation of customer satisfaction after service failure and recovery. Journal of Service Research, 3(2), 121–137.
220
References
McKinney, V., Yoon, K., & Zahedi, F. M. (2002). The measurement of web-customer satisfaction: An expectation and disconfirmation approach. Information Systems Research, 13(3), 296–315. McKinney, V., Kanghyun, Y., & Zahedi, F. M. (2002). The measurement of web-customer satisfaction: An expectation and disconfirmation approach. Information Systems Research, 13(3), 296–315. McKinney, V., Yoon, K., & Zahedi, F. M. (2002). The measurement of web-customer satisfaction: An expectation and disconfirmation approach. Information Systems Research, 13(3), 296–315. McMurry, R. (1981). The mystique of supersalesmanship. Harvard Business Review, 39(2), 113–122. McSweeney, B. (2002). Hofstede’s model of national cultural differences and their consequences: A triumph of faith-a failure of analysis. Human Relations, 55(1), 89–118. Mead, G. H. (1934). Mind, self and society (vol. 111). University of Chicago Press.: Chicago. Mende, M., Scott, M. L., van Doorn, J., Grewal, D., & Shanks, I. (2019). Service robots rising: How humanoid robots influence service experiences and elicit compensatory consumer responses. Journal of Marketing Research, 56(4), 535–556. Meng, H. (2008). Social script theory and cross-cultural communication. Intercultural Communication Studies, 17(1), 132–138. Merkle, M. (2019). Customer responses to service robots—comparing human-robot interaction with human-human interaction, Proceedings of the 52nd Hawaii International Conference on System Sciences, 1396–1405. Meuter, M.L., Ostrom, A. L., Roundtree, R. I., & Bitner, M. J. (2000). Self-service technologies: Understanding customer satisfaction with technology-based service encounters. Journal of Marketing, 64(3), 50–64. Michael, L. H., Hou, S. T., & Fan, H. L. (2011). Creative self-efficacy and innovative behavior in a service setting: Optimism as a moderator. The Journal of Creative Behavior, 45(4), 258–272. Michel, S., Bowen, D. E., & Johnston, R. (2009). Why service recovery fails: Tensions among the customer, employee and process perspectives. Journal of Service Management, 20(3), 253–273. Mills, P. K., Chase, R. B., & Margulies, N. (1983). Motivating the client/employee system as a service production strategy. Academy of Management Review, 8(2), 301–310. Mirheydar, H. S., & Parsons, J. K. (2013). Diffusion of robotics into clinical practice in the United States: process, patient safety, learning curves, and the public health. World Journal of Urology, 31(3), 455–461. Miskam, M. A., Masnin, N. F. S., Jamhuri, M. H., Shamsuddin, S., Omar, A. R., & Yussof, H. (2014). Encouraging children with autism to improve social and communication skills through the game-based approach. Procedia Computer Science, 42, 93–98. Mittal, V., & Kamakura, W. A. (2001). Satisfaction, repurchase intent, and repurchase behavior: Investigating the moderating effect of customer characteristics. Journal of Marketing Research, 38(1), 131–142. Moeller, S. (2010). Characteristics of services–a new approach uncovers their value. Journal of Services Marketing, 24(5), 359–368. Mohr, L. A., & Bitner, M. J. (1991). Mutual understanding between customers and employees in service encounters. Advances in Consumer Research, 18, 611–617.
References
221
Molinari, G. (1964). Latest developments in automatic retailing in Europe. Journal of Marketing, 28(4), 5–9. Montero, I., & León, O. G. (2007). A guide for naming research studies in Psychology. International Journal of clinical and Health Psychology, 7(3), 847–862. Moorman, C., Deshpande, R., & Zaltman, G. (1993). Factors affecting trust in market research relationships. Journal of Marketing, 57(1), 81–101. Moosa, N., & Panurach, P. (2008). Encouraging front-line employees to rise to the innovation challenge. Strategy & Leadership, 36(4), 4–9. Moosa, M. M., & Ud-Dean, S. M. (2010). Danger avoidance: An evolutionary explanation of uncanny valley. Biological Theory, 5(1), 12–14. Mori, M. (1970). The Uncanny Valley. Energy, 7, 33–35. Mori, M. (1970/2005). The uncanny valley. (K. F. MacDorman, & T. Minato, Trans.). Energy, 7, 33–35. Mori, M., MacDorman, K. F., & Kageki, N. (2012). The uncanny valley [from the field]. IEEE Robotics & Automation Magazine, 19(2), 98–100. Morse, J. M., & Richards, L. (2002). Readme First: For a User’s Guide to Qualitative Methods, Thousand Oaks, California, SAGE Publications. Moshkina, L., Trickett, S., & Trafton, J. G. (2014). Social engagement in public places: a tale of one robot. ACM/IEEE International Conference on Human-Robot Interaction, 382–389. Moshkina, L., Trickett, S., & Trafton, J. G. (2014). Social engagement in public places: a tale of one robot. In Proceedings of the 2014 ACM/IEEE International Conference on Human-Robot Interaction, 382–389. Nass, C., & Lee, K. M. (2001). Does computer-synthesized speech manifest personality? Experimental tests of recognition, similarity-attraction, and consistency-attraction. Journal of Experimental Psychology: Applied, 7(3), 171–181. Nass, C., & Moon, Y. (2000). Machines and mindlessness: Social responses to computers. Journal of Social Issues, 56(1), 81–103. Nass, C., & Steuer, J. (1993). Voices, boxes, and sources of messages: computers and social actors. Human Communication Research, 19(4), 504–527. Nass, C., Steuer, J., & Tauber, E. R. (1994). Computers are social actors. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, 72–78. Nass, C., Moon, Y., & Green, N. (1997). Are machines gender neutral? Gender-stereotypic responses to computers with voices. Journal of Applied Social Psychology, 27(10), 864– 876. Neil, D. (2001). Introducing LISREL: A guide for the uninitiated. International Journal of Market Research 43(4), 455–465. Nestlé (2016). Nestlé to Use Humanoid Robot to Sell Nescafé in Japan. [http://www.nestle. com/media/news/nestle-humanoid-robot-nescafe-japan]. Nishio, S., Ishiguro, H., & Hagita, N. (2007). Geminoid: Teleoperated android of an existing person. Humanoid Robots: New Developments, 14, 343–352. Nomura, T., Kanda, T., Suzuki, T., & Kato, K. (2004). Psychology in Human-Robot Communication: An Attempt through Investigation of Negative Attitudes and Anxiety toward Robots. IEEE International Workshop on Robot and Human Interactive Communication. Nomura, T., Suzuki, T., Kanda, T., & Kato, K. (2006). Altered Attitudes of People toward Robots: Investigation through the Negative Attitudes toward Robots Scale. AAAI-06 Workshop on Human Implications of Human-Robot Interaction, 29–35.
222
References
Nomura, T., Suzuki, T., Kanda, T., & Kato, K. (2006). Measurement of Anxiety Toward Robots. IEEE International Symposium on Robot and Human Interactive Communication, 372–377. Nomura, T., Kanda, T., Suzuki, T., & Kato, K. (2008). Prediction of human behavior in human—robot interaction using psychological scales for anxiety and negative attitudes toward robots. IEEE Transactions on Robotics, 24(2), 442–451. Nomura, S., Hasegawa-Ohira, M., Kurosawa, Y., Hanasaka, Y., Yajima, K., & Fukumura, Y. (2012). Skin Tempereture as a Possible Indicator of Student’s Involvement in E-Learning Sessions. International Journal of Electronic Commerce Studies, 3(1), 101–110. Northfield, R. (2015). Robot hotel. Engineering & Technology, 10(6), 50–51. Nottenburg, G., & Shoben, E. J. (1980). Scripts as linear orders. Journal of Experimental Social Psychology, 16(4), 329–347. Nunnally, J. C. (1978). Psychometric Theory, vol. 2, New York. O’Neill, M., & Palmer, A. (2004). Cognitive dissonance and the stability of service quality perceptions. Journal of Services Marketing, 433–449. Ogata, T., & Sugano, S. (2000). Emotional communication between humans and the autonomous robot wamoeba-2 (waseda amoeba) which has the emotion model. JSME International Journal Series C Mechanical Systems, Machine Elements and Manufacturing, 43(3), 568–574. Olazabal, A. Morales, Anita Cava, and Rene Sacasas (2005). Marketing and the Law. Journal of the Academy of Marketing Science, 33 (1), 116–118. Oldham, G. R., & Cummings, A. (1996). Employee creativity: Personal and contextual factors at work. Academy of Management Journal, 39(3), 607–634. Oliver, R. L. (1977). Effect of expectation and disconfirmation on postexposure product evaluations: An alternative interpretation. Journal of Applied Psychology, 62(4), 480. Oliver, R. L. (1980). A cognitive model of the antecedents and consequences of satisfaction decisions. Journal of Marketing Research, 17(4), 460–469. Oliver, R. L. (2014). Satisfaction: A behavioral perspective on the consumer: A behavioral perspective on the consumer, Routledge. Oliver, R. L., and DeSarbo, W. S. (1988). Response determination in satisfaction judgements. Journal of Consumer Research, 10, 250–255. Oliver, R. L., & DeSarbo, W. S. (1988). Response determinants in satisfaction judgments. Journal of Consumer Research, 14(4), 495–507. Oliver, R. L., Balakrishnan, P. S., & Barry, B. (1994). Outcome satisfaction in negotiation: A test of expectancy disconfirmation. Organizational Behavior and Human Decision Processes, 60(2), 252–275. Oliver, R. L., Rust, R. T., & Varki, S. (1997). Customer delight: foundations, findings, and managerial insight. Journal of Retailing, 73(3), 311. Oshikawa, S. (1969). Can cognitive dissonance theory explain consumer behavior?. Journal of Marketing, 33(4), 44–49. Osipyan, H., Vedadi, S., & Cheok, A. D. (2017). Machines as an assistants for humans’ creativity: A conceptual model. In IECON 2017–43rd Annual Conference of the IEEE Industrial Electronics Society, 3316–3321. Ottenbacher, M. C., & Harrington, R. J. (2009). The product innovation process of quick-service restaurant chains. International Journal of Contemporary Hospitality Management, 21(5), 523–541.
References
223
Ouwehand, A. N. (2017). The role of culture in the aceptance of elderly towards ‘social assertive robots: How do cultural factors influence the acceptance of elderly people towards social assertive robotics in the Netherlands and Japan?. Bachelor’s Thesis at University of Twente. Pandey, A. K., & Gelin, R. (2018). A mass-produced sociable humanoid robot: pepper: the first machine of its kind. IEEE Robotics & Automation Magazine, 25(3), 40–48. Pan, Y., Okada, H., Uchiyama, T., & Suzuki, K. (2015). On the reaction to robot’s speech in a hotel public space. International Journal of Social Robotics, 7(5), 911–920. Parasuraman, A., Zeithaml, V. A., & Berry, L. L. (1988). Servqual: A multiple-item scale for measuring consumer perc. Journal of Retailing 64(1), 12. Parasuraman, A., Zeithaml, V. A., & Berry, L. L. (1988). SERVQUAL: A Multiple-item scale for measuring consumer perceptions of service quality. Journal of Retailing, 64, 12–39. Parasuraman, A., Zeithaml, V. A., & Berry, L. L. (1991). Refinement and Reassessment of the SERVQUAL Scale. Journal of Retailing, 67, 421–450. Parasuraman, A., Zeithaml, V. A., & Berry, L. L. (1994). Reassessment of Expectations as a Comparison Standard in Measuring Service Quality: Implications for Further Research. Journal of Marketing, 58(1), 111–124. Paré, G., Trudel, M. C., Jaana, M., & Kitsiou, S. (2015). Synthesizing information systems knowledge: A typology of literature reviews. Information & Management, 52(2), 183–199. Park, J. W., Kim, W. H., Lee, W. H., & Chung, M. J. (2010). Artificial emotion generation based on personality, mood, and emotion for life-like facial expressions of robots. IFIP Human-Computer Interaction Symposium, 223–233. Paul, J. (2000). Are you delighting your customers? Nonforprofit World, 18(5), 34–36. Payne, A. F., Storbacka, K., & Frow, P. (2008). Managing the co-creation of value. Journal of the Academy of Marketing Science, 36(1), 83–96. Perez, J. A., Deligianni, F., Ravi, D., & Yang, G. Z. (2018). Artificial intelligence and robotics. arXiv preprint arXiv:1803.10813. Perry-Smith, J. E., & Shalley, C. E. (2014). A social composition view of team creativity: The role of member nationality-heterogeneous ties outside of the team. Organization Science, 25(5), 1434–1452. Peter, J. P. (1979). Reliability: A review of psychometric basics and recent marketing practices. Journal of Marketing Research, 16(1), 6–17. Peter, J. P., & Churchill Jr, G. A. (1986). Relationships among research design choices and psychometric properties of rating scales: A meta-analysis. Journal of Marketing Research 23(1), 1–10. Peterson, C., & Seligman, M. E. (2004). Character strengths and virtues: A handbook and classification, Washington, DC: American Psychological Association. Phillips, E., Zhao, X., Ullman, D., & Malle, B. F. (2018). What is human-like?: Decomposing robots’ human-like appearance using the anthropomorphic robot (abot) database. Proceedings of the 2018 ACM/IEEE International Conference on Human-Robot Interaction, 105–113. Pierce, C., Block, R., & Aguinis, H. (2004). Cautionary note on reporting eta-squared values from multifactor ANOVA designs. Educational and Psychological Measurement 64(6), 916–924. Piezzo, C., & Suzuki, K. (2017). Feasibility study of a socially assistive humanoid robot for guiding elderly individuals during walking. Future Internet, 9(3), 30–46.
224
References
Pindyck, R. S., & Rubinfeld, D. L. (1991). Econometric Models and Economic Forecasts. McGraw-Hill, USA. Pinillos, R., Marcos, S., Feliz, R., Zalama, E., & Gómez-García-Bermejo, J. (2016). Longterm assessment of a service robot in a hotel environment. Robotics and Autonomous Systems, 79, 40–57. Plambeck, E. L., & Taylor, T. A. (2005). Sell the plant? The impact of contract manufacturing on innovation, capacity, and profitability. Management Science, 51(1), 133–150. Podsakoff, N. P. (2003). Common method biases in behavioral research: a critical review of the literature and recommended remedies. Journal of Applied Psychology 885(879), 10–1037. Podsakoff, P. M., MacKenzie, S. B., & Podsakoff, N. P. (2012). Sources of method bias in social science research and recommendations on how to control it. Annual Review of Psychology, 63, 539–569. Pot, E., Monceaux, J., Gelin, R., & Maisonnier, B. (2009). Choregraphe: a graphical tool for humanoid robot programming. RO-MAN 2009-The 18th IEEE International Symposium on Robot and Human Interactive Communication, 46–51. Pulido, J. C., González, J. C., Suárez-Mejías, C., Bandera, A., Bustos, P., & Fernández, F. (2017). Evaluating the child–robot interaction of the NAOTherapist platform in pediatric rehabilitation. International Journal of Social Robotics, 9(3), 343–358. Qazi, A., Tamjidyamcholo, A., Raj, R. G., Hardaker, G., & Standing, C. (2017). Assessing consumers’ satisfaction and expectations through online opinions: Expectation and disconfirmation approach. Computers in Human Behavior, 75, 450–460. Rae, I., Takayama, L., & Mutlu, B. (2013). In-body experiences: embodiment, control, and trust in robot-mediated communication. Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, 1921–1930. Rajesh, M. (2015). Inside Japan’s First Robot-Staffed Hotel. Japan Holidays [http://www.the guardian.com/travel/2015/aug/14/japan-henn-na-hotel-staffed-by-robots]. Rand, D. G. (2012). The promise of Mechanical Turk: How online labor markets can help theorists run behavioral experiments. Journal of Theoretical Biology, 299(4), 172–179. Rau, P. P., Li, Y., & Li, D. (2009). Effects of communication style and culture on ability to accept recommendations from robots. Computers in Human Behavior, 25(2), 587–595. Reed II, A., Forehand, M. R., Puntoni, S., & Warlop, L. (2012). Identity-based consumer behavior. International Journal of Research in Marketing, 29(4), 310–321. Reeves, B., and Nass, C. I. (1996). The media equation: How people treat computers, television, and new media like real people and places. Stanford, CA: CSLI Publications. Rego, A., Sousa, F., Marques, C., & e Cunha, M. P. (2014). Hope and positive affect mediating the authentic leadership and creativity relationship. Journal of Business Research, 67(2), 200–210. Rehm, M., & André, E. (2005). Where do they look? Gaze behaviors of multiple users interacting with an embodied conversational agent. International Workshop on Intelligent Virtual Agents, 241–252. Reich, N., & Eyssel, F. (2013). Attitudes towards service robots in domestic environments: The role of personality characteristics, individual interests, and demographic variables. Paladyn, Journal of Behavioral Robotics, 4(2), 123–130. Reichheld, F. F., & Sasser, W. E. (1990). Zero defeofions: Quoliiy comes to services. Harvard Business Review, 68(5), 105–111.
References
225
Richardson, K., Coeckelbergh, M., Wakunuma, K., Billing, E., Ziemke, T., Gomez, P., ... & Belpaeme, T. (2018). Robot enhanced therapy for children with autism (dream): A social model of autism. IEEE TEchnology and SocIETy MagazInE, 37(1), 30–39. Riek, L. D., Paul, P. C., & Robinson, P. (2010). When my robot smiles at me: Enabling humanrobot rapport via real-time head gesture mimicry. Journal on Multimodal User Interfaces, 3(1-2), 99–108. Rijsdijk, S. A., Hultink, E. J., & Diamantopoulos, A. (2007). Product intelligence: its conceptualization, measurement and impact on consumer satisfaction. Journal of the Academy of Marketing Science, 35(3), 340–356. Rodriguez-Lizundia, E., Marcos, S., Zalama, E., Gómez-García-Bermejo, J., & Gordaliza, A. (2015). A bellboy robot: Study of the effects of robot behaviour on user engagement and comfort. International Journal of Human-Computer Studies, 82, 83–95. Rosenbaum, P. 2005. Observational studies, Everitt, B. S. & Howell, D. C. (eds.). Encyclopedia of Statistics in Behavioral Sciences, Chichester: Wiley & Sons. Rossiter, J. R. (2002). The C-OAR-SE procedure for scale development in marketing. International Journal of Research in Marketing, 19(4), 305–335. Russell, S. J., Norvig, P., Canny, J. F., Malik, J. M., and Edwards, D. D. (1995). Artificial Intelligence: a Modern Approach, volume 2. Englewood Cliffs: Prentice Hall. Rust, R. T., & Oliver, R. L. (2000). Should we delight the customer?. Journal of the Academy of Marketing Science, 28(1), 86. Rust, R. T., Inman, J. J., Jia, J., & Zahorik, A. (1999). What you don’t know about customerperceived quality: The role of customer expectation distributions. Marketing Science, 18(1), 77–92. Ruxton, G. D., & Beauchamp, G. (2008). Time for some a priori thinking about post hoc testing. Behavioral Ecology, 19(3), 690–693. Sabanovic, S., Michalowski, M. P., & Simmons, R. (2006). Robots in the wild: Observing human-robot social interaction outside the lab. IEEE International Workshop on Advanced Motion Control, 596–601. Sabelli, A. M., and T. Kanda 2016. Robovie as a mascot: a qualitative study for long-term presence of robots in a shopping mall. International Journal of Social Robotics 8(2), 211–221. Salem, M., Ziadee, M., & Sakr, M. (2014). Marhaba, how may I help you? Effects of politeness and culture on robot acceptance and anthropomorphization. ACM/IEEE International Conference on Human-Robot Interaction, 74–81. Sanborn, L. (2015). The future of academic librarianship: MOOCs and the robot revolution. Reference & User Services Quarterly, 55(2), 97–101. Scheffé, H. (1953). A method for judging all contrasts in the analzsis of variance, Biometrika 40, 87–104. Schneider, B., & Bowen, D. E. (1999). Understanding customer delight and outrage. Sloan Management Review, 41(1), 35–45. Schraft, R. D., Degenhart, E., & Hagele, M. (1993). Service robots: the appropriate level of automation and the role of users/operators in the task execution. Proceedings of IEEE Systems Man and Cybernetics Conference-SMC, 4, 163–169. Schuchard-Ficher, C., Backhaus, K., Humme, U., Lohrberg, W., & Plinke, W. (2013). Multivariate Analysemethoden: Eine anwendungsorientierte Einführung. Springer-Verlag.
226
References
Searleman, A., & Herrmann, D. J. (1994). Memory from a broader perspective. New York: McGraw-Hill. Sebastian, J., Tai, C. Y., Lindholm, K., & Hsu, Y. L. (2015). Development of caricature robots for interaction with older adults. International Conference on Human Aspects of IT for the Aged Population, 324–332. Sénécal, S., Léger, P.-M., Fredette, M., & Riedl, R. (2012). Consumers’ online cognitive scripts: A neurophysiological approach. Proceedings of the 33rd International Conference on Information Systems, Orlando, FL, 1–10. Setiawan, H., & Sayuti, A. J. (2017). Effects of service quality, customer trust and corporate image on customer satisfaction and loyalty: an assessment of travel agencies customer in South Sumatra Indonesia. IOSR Journal of Business and Management, 19(5), 31–40. Seyama, J. I., & Nagayama, R. S. (2007). The uncanny valley: effect of realism on the impression of artificial human faces. Presence: Teleoperators and virtual environments, 16(4), 337–351. Sgorbissa, A., Papadopoulos, I., Bruno, B., Koulouglioti, C., & Recchiuto, C. (2018). Encoding guidelines for a culturally competent robot for elderly care. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 1988–1995. Shackle, G. L. S. (2012). Expectation in economics. Cambridge University Press. Shamsuddin, S., Yussof, H., Ismail, L. I., Mohamed, S., Hanapiah, F. A., & Zahari, N. I. (2012). Initial response in HRI-a case study on evaluation of child with autism spectrum disorders interacting with a humanoid robot Nao. Procedia Engineering, 41, 1448–1455. Shankar, Venkatesh (2018). How artificial intelligence (AI) is reshaping retailing. Journal of Retailing, 94(4), 6–11. Shankar, V. (2018). How artificial intelligence (AI) is reshaping retailing. Journal of Retailing, 94(4), vi–xi. Shankarmahesh, M. N., Ford, J. B., & LaTour, M. S. (2003). Cultural dimensions of switching behavior in importer-exporter relationships. Academy of Marketing Science Review, 3(6), 1–17. Shanock, L., Baran, B., Gentry, W., Pattison, S., & Heggestad, E. (2010). Polynomial regression with response surface analysis: a powerful approach for examining moderation and overcoming limitations of difference scores. Journal of Business and Psychology 25(4), 543–554. Shanock, L. R., Baran, B. E., Gentry, W. A., Pattison, S. C., & Heggestad, E. D. (2010). Polynomial regression with response surface analysis: A powerful approach for examining moderation and overcoming limitations of difference scores. Journal of Business and Psychology, 25(4), 543–554. Sheth, J. N. (1967). A review of buyer behavior. Management Science, 13(12), B-718. Shiomi, M., Shinozawa, K., Nakagawa, Y., Miyashita, T., Sakamoto, T., Terakubo, T., ... & Hagita, N. (2013). Recommendation effects of a social robot for advertisement-use context in a shopping mall. International Journal of Social Robotics, 5(2), 251–262. Shrout, P., & Fleiss, J.L. (1979). Intraclass correlation: Uses in assessing rater reliability. Psychological Bulletin 86(2), 420–428. Siciliano, B., Sciavicco, L., Villani, L., & Oriolo, G. (2010). Robotics: modelling, planning and control. Springer Science & Business Media.
References
227
Sidner, C. L., Kidd, C. D., Lee, C., & Lesh, N. (2004). Where to look: a study of humanrobot engagement. Proceedings of the 9th International Conference on Intelligent User Interfaces, 78–84. Silver, D., Hubert, T., Schrittwieser, J., Antonoglou, I., Lai, M., Guez, A., ... & Lillicrap, T. (2018). A general reinforcement learning algorithm that masters chess, shogi, and Go through self-play. Science, 362(6419), 1140–1144. Simmons, J. P., Nelson, L. D., & Simonsohn, U. (2011). False-positive psychology: undisclosed flexibility in data collection and analysis allows presenting anything as significant. Psychological science, 22(11), 1359–1366. Sivakumar, K., & Nakata, C. (2001). The stampede toward Hofstede’s framework: Avoiding the sample design pit in cross-cultural research. Journal of International Business Studies, 32(3), 555–574. Skiera, B., & Albers, S. (2008). Regressionsanalyse, in: Herrmann, A., Homburg, Ch., Klarmann, M. (Hrsg.), Marktforschung: Methoden—Anwendungen—Praxisbeispiele, vol. 3, Wiesbaden, 467–497. Slåtten, T., & Mehmetoglu, M. (2011). Antecedents and effects of engaged frontline employees: A study from the hospitality industry. Managing Service Quality, 21(1), 88–107. Smith, M. D. (2002). The impact of shopbots on electronic markets. Journal of the Academy of Marketing Science, 30(4), 446. Soares, A. M., Farhangmehr, M., & Shoham, A. (2007). Hofstede’s dimensions of culture in international marketing studies. Journal of Business Research, 60(3), 277–284. Söderlund, M. (2002). Customer familiarity and its effects on satisfaction and behavioral intentions. Psychology & Marketing, 19(10), 861–879. Softbank, Website (2018). https://www.softbankrobotics.com, accessed on 30.05.2018. Solomon, M. R., Surprenant, C., Czepiel, J. A., & Gutman, E. G. (1985). A role theory perspective on dyadic interactions: the service encounter. Journal of Marketing, 49(1), 99–111. Sparks, B., & Fredline, L. (2007). Providing an explanation for service failure: Context, content, and customer responses. Journal of Hospitality & Tourism Research, 31(2), 241– 260. Spence, P. R., Westerman, D., Edwards, C., & Edwards, A. (2014). Welcoming our robot overlords: initial expectations about interaction with a robot. Communication Research Reports, 31(3), 272–280. Spreng, R. A., MacKenzie, S. B., & Olshavsky, R. W. (1996). A reexamination of the determinants of consumer satisfaction. Journal of Marketing, 60(3), 15–32. Steenkamp, J. B. E. (2001). The role of national culture in international marketing research. International Marketing Review, 18(1), 30–44. Steenkamp, J. B. E., Ter Hofstede, F., & Wedel, M. (1999). A cross-national investigation into the individual and national cultural antecedents of consumer innovativeness. Journal of Marketing, 63(2), 55–69. Stewart, D. M. (2003). Piecing together service quality: A framework for robust service. Production and Operations Management, 12(2), 246–265. Stock, R. (2003). Der Einfluss der Kundenzufriedenheit auf die Preissensitivität von Firmenkunden. Die Betriebswirtschaft, 63, 333–348.
228
References
Stock, R. (2004). Erfolgsauswirkungen der marktorientierten Gestaltung des Personalmanagements. Zeitschrift für betriebswirtschaftliche Forschung, 56(5), 237–258. Stock, R. (2013). Teams an der Schnittstelle zwischen Anbieter-und Kunden-Unternehmen: eine integrative Betrachtung. In: Neue Betriebswirtschaftliche Forschung (Vol. 317), Springer-Verlag, Wiesbaden. Stock, R. (2016). Emotion transfer from frontline social robots to human customers during service encounters: Testing an artificial emotional contagion model. International Conference on Information Systems, Dublin, Ireland. Stock, R. M. (2015). Is boreout a threat to frontline employees’ innovative work behavior?. Journal of Product Innovation Management, 32(4), 574–592. Stock, R. M. (2016). Understanding the relationship between frontline employee boreout and customer orientation. Journal of Business Research, 69(10), 4259–4268. Stock, R. M. (2018). Can service robots hamper customer anger and aggression after a service failure?, International Conference on Information Systems, San Francisco, USA. Stock, R., & Merkle, M. (2018). Can humanoid service robots perform better than service employees? Hawaii International Conference on System Sciences, Waikoloa, USA. Stock, R. M., & Bednarek, M. (2014). As they sow, so shall they reap: Customers’ influence on customer satisfaction at the customer interface. Journal of the Academy of Marketing Science, 42(4), 400–414. Stock, R., & Genisyürek, N. (2012). A taxonomy of expatriate leaders’ cross-cultural uncertainty: Insights into the leader–employee dyad. International Journal of Human Resource Management, 23(15), 3258–3286. Stock-Homburg, R., Peters, J., Schneider, K., Prasad, V., & Nukovic, L. (2020). Evaluation of the handshake turing test for anthropomorphic robots, ACM/IEEE International Conference on Human Robot Interaction (HRI), Cambridge, UK. Stock-Homburg, R. M., Heald, S. L., Holthaus, C., Gillert, N. L., & von Hippel, E. (2020). Need-solution pair recognition by household sector individuals: Evidence, and a cognitive mechanism explanation. Research Policy, in press. Stock, R., & Merkle, M. (2017). A service robot acceptance model: User acceptance of humanoid robots during service encounters. Proceedings of the IEEE International Conference on Pervasive Computing and Communications, 339–344. Stock, R., & Merkle, M. (2018). Customer responses to robotic innovative behavior cues during the service encounter. Proceedings of the International Conference on Information Systems, San Francisco, CA 1–17. Stock, R., & Nguyen, M. A. (2019, January). Robotic psychology. What do we know about human-robot interaction and what do we still need to learn?. In Proceedings of the 52nd Hawaii international conference on system sciences, 1936–1945. Stock, R., & Özbek-Potthoff, G. (2014). Implicit leadership in an intercultural context: Theory extension and empirical investigation. International Journal of Human Resource Management, 25(12), 1651–1668. Stock, R. M., Jong, A. D., & Zacharias, N. A. (2017). Frontline employees’ innovative service behavior as key to customer loyalty: Insights into FLEs’ resource gain spiral. Journal of Product Innovation Management, 34(2), 223–245. Stock, R., Merkle, M., Eidens, D., Hannig, M., Heineck, P., Nguyen, M. A., & Völker, J. (2019). When Robots Enter Our Workplace: Understanding Employee Trust in Assistive
References
229
Robots. Proceedings of the 40th International Conference on Information Systems, ICIS, Munich. Sun, Q., Wang, C., & Cao, H. (2009). An extended TAM for analyzing adoption behavior of mobile commerce. Eighth International Conference on Mobile Business, 52–56. Super, N. (2002). Who will be there to care? The growing gap between caregiver supply and demand. National Health Policy Forum, George Washington University. Suri, S., S., & Watts, D. J. (2011). Cooperation and contagion in webbased, networked public goods experiments. PLoS One, 6(3). Swanson, S. R., & Kelley, S. W. (2001). Attributions and outcomes of the service recovery process. Journal of Marketing Theory and Practice, 9(4), 50–65. Sweeney, J. C., Hausknecht, D., & Soutar, G. N. (2000). Cognitive dissonance after purchase: a multidimensional scale. Psychology & Marketing, 17(5), 369–385. Szajna, B., & Scamell, R. W. (1993). The effects of information system user expectations on their performance and perceptions. MIS Quarterly, 17(4), 493–516. Szymanski, D. M., & Henard, D. H. (2001). Customer satisfaction: A meta-analysis of the empirical evidence. Journal of the Academy of Marketing Science, 29(1), 16–35. Tanaka, F., Isshiki, K., Takahashi, F., Uekusa, M., Sei, R., & Hayashi, K. (2015). Pepper learns together with children: Development of an educational application. IEEE International Conference on Humanoid Robots, 270–275. Tax, S. S., & Brown, S. W. (1998). Recovering and learning from service failure. MIT Sloan Management Review, 40(1), 75–88. Tax, S. S., Brown, S. W., & Chandrashekaran, M. (1998). Customer evaluations of service complaint experiences: Implications for relationship marketing. Journal of Marketing, 62(2), 60–76. Tay, B., Jung, Y., & Park, T. (2014). When stereotypes meet robots: the double-edge sword of robot gender and personality in human–robot interaction. Computers in Human Behavior, 38, 75–84. Taylor, S. A., Cronin Jr., J. J., & Hansen, R. S. (1991). Schema and script theory in channels research. Marketing Theory and Applications, 2, 15–24. Telci, E. E., Maden, C., & Kantur, D. (2011). The theory of cognitive dissonance: a marketing and management perspective. Procedia-Social and Behavioral Sciences, 24, 378–386. Thompson, J. C., Trafton, J. G., & McKnight, P. (2011). The perception of humanness from the movements of synthetic agents. Perception, 40(6), 695–704. Thornhill, R. & Gangestad, S.W. (1993). Human facial beauty—Averageness, symmetry, and parasite resistance. Human Nature 4(3), 237–269. Tomkins, S. S. (1978). Script theory: Differential magnification of affects. Nebraska Symposium on Motivation, 26, 201–236. Toothaker, L. E. (1993). Multiple Comparison Procedures (no. 89). Sage Publications. Touré-Tillery, M., & McGill, A. L. (2015). Who or what to believe: Trust and the differential persuasiveness of human and anthropomorphized messengers. Journal of Marketing, 79(4), 94–110. Trompenaars F., & Woolliams, P. (2002). A new framework for managing change across cultures. Journal of Change Management, 3(4), 361–375. Trovato, G., Ramos, J. G., Azevedo, H., Moroni, A., Magossi, S., Simmons, R., ... & Takanishi, A. (2017). A receptionist robot for Brazilian people: study on interaction involving illiterates. Paladyn, Journal of Behavioral Robotics, 8(1), 1–17.
230
References
Tse, D. K., & Wilton, P. C. (1988). Models of consumer satisfaction formation: An extension. Journal of Marketing Research, 25(2), 204–212. Tukey, J. W. (1953). The problem of multiple comparisons, Princeton University, unpublished notes. Van Birgelen, M., de Ruyter, K., de Jong, A., & Wetzels, M. (2002). Customer evaluations of after-sales service contact modes: An empirical analysis of national culture’s consequences. International Journal of Research in Marketing, 19(1), 43–64. van Doorn, J., Mende, M., Noble, S.M., Hulland, J., Ostrom, A.L., Grewal, D. and Petersen, J.A. (2017). Domo arigato Mr. Roboto: emergence of automated social presence in organizational frontlines and customers’ service experiences. Journal of Service Research, 20(1), 43–58. Van Everdingen, Y. M., & Waarts, E. (2003). The effect of national culture on the adoption of innovations. Marketing Letters, 14(3), 217–232. Van Nort, D., & Hogeveen, E. (2017). A systematic review of computational creativity practices across disciplines. SSHRC Knowledge Synthesis Report, York University. Vargo, S. L., & Lusch, R. F. (2004). The four service marketing myths: remnants of a goodsbased, manufacturing model. Journal of Service Research, 6(4), 324–335. Vaussard, F., Fink, J., Bauwens, V., Rétornaz, P., Hamel, D., Dillenbourg, P., & Mondada, F. (2014). Lessons learned from robotic vacuum cleaners entering the home ecosystem. Robotics and Autonomous Systems, 62(3), 376–391. Veloso, M. M. (2002). Entertainment robotics. Communications of the ACM, 45(3), 59–63. Venaik, S., & Brewer, P. (2016). National culture dimensions: the perpetuation of cultural ignorance. Management Learning, 47(5), 563–589. Venkatesh, V., & Davis, F. D. (2000). A theoretical extension of the technology acceptance model: Four longitudinal field studies. Management Science, 46(2), 186–204. Venkatesh, V., & Goyal, S. (2010). Expectation disconfirmation and technology adoption: polynomial modeling and response surface analysis. MIS Quarterly, 34(2), 281–303. Venkatesh, V., Morris, M. G., Davis, G. B., & Davis, F. D. (2003). User acceptance of information technology: Toward a unified view. MIS Quarterly, 27(3), 425–478. Wang, X. (2011). The effect of unrelated supporting service quality on consumer delight, satisfaction, and repurchase intentions. Journal of Service Research, 14(2), 149–163. Wang, S., Lilienfeld, S. O., & Rochat, P. (2015). The uncanny valley: Existence and explanations. Review of General Psychology, 19(4), 393–407. Wang, F. Y., Zhang, J. J., Zheng, X., Wang, X., Yuan, Y., Dai, X., ... & Yang, L. (2016). Where does AlphaGo go: From church-turing thesis to AlphaGo thesis and beyond. IEEE/CAA Journal of Automatica Sinica, 3(2), 113–120. Warburton, N. (2017). What does a portrait of Erica the android tell us about being human? The Guardian,https://www.theguardian.com/technology/2017/sep/09/robot-human-artifi cial-intelligence-philosophy, accessed 31.05.2018. Weun, S., Beatty, S. E., & Jones, M. A. (2004). The impact of service failure severity on service recovery evaluations and post-recovery relationships. Journal of Services Marketing, 18(2), 133–146. Wilson, D. T., & Bozinoff, L. (1980). Role theory and buying-selling negotiations: A critical overview. American Marketing Association Educators’ Conference Proceedings, 118– 121.
References
231
Wirtz, J., & Zeithaml, V. (2018). Cost-effective service excellence. Journal of the Academy of Marketing Science, 46(1), 59–80. Wirtz, J., Patterson, P. G., Kunz, W. H., Gruber, T., Lu, V. N., Paluch, S., & Martins, A. (2018). Brave new world: service robots in the frontline. Journal of Service Management, 29(5), 907–931. Witte, A. E. (2012). Making the case for a post-national cultural analysis of organizations. Journal of Management Inquiry, 21(2), 141–159. World Bank (2016). World Development Report 2016: Digital Dividends. Wuyts, S., Dutta, S., & Stremersch, S. (2004). Portfolios of interfirm agreements in technology-intensive markets: Consequences for innovation and profitability. Journal of Marketing, 68(2), 88–100. Wyckham, R. G., Fitzroy, P. T., & Mandry, G. D. (1993). Marketing of services an evaluation of the theory. European Journal of Marketing, 9(1), 59–67. Xiao, L., & Kumar, V. (2019). Robotics for customer service: A useful complement or an ultimate substitute?. Journal of Service Research, 1094670519878881. Xu, K., & Lombard, M. (2016). Media are social actors: expanding the CASA paradigm in the 21st Century. Annual Conference of the International Communication Association. Xu, J., Broekens, J., Hindriks, K., & Neerincx, M. A. (2014). Robot Mood is Contagious: Effects of Robot Body Language in the Imitation Game. Proceedings of the 2014 International Conference on Autonomous Agents and Multi-agent Systems, 973–980. Yamada, Y., Kawabe, T., & Ihaya, K. (2013). Categorization difficulty is associated with negative evaluation in the “uncanny valley” phenomenon. Japanese Psychological Research, 55(1), 20–32. Yamazaki, A., Yamazaki, K., Burdelski, M., Kuno, Y., & Fukushima, M. (2010). Coordination of verbal and non-verbal actions in human–robot interaction at museums and exhibitions. Journal of Pragmatics, 42(9), 2398–2414. Yeniyurt, S., & Townsend, J. D. (2003). Does culture explain acceptance of new products in a country? An empirical investigation. International Marketing Review, 20, 377–395. Yi, Y. (1988). A Critical Review of Customer Satisfaction. Review of Marketing, Zeithaml, V. A. (ed.), American Marketing Association, Chicago, Illinois, 68–123. Yi, Y. (1990). A critical review of consumer satisfaction. Review of Marketing, 4(1), 68–123. Yuan, F., & Woodman, R. W. (2010). Innovative behavior in the workplace: The role of performance and image outcome expectations. Academy of Management Journal, 53(2), 323–342. Yüksel, A., & Yüksel, F. (2001). The expectancy-disconfirmation paradigm: a critique. Journal of Hospitality & Tourism Research, 25(2), 107–131. Zalama, E., García-Bermejo, J. G., Marcos, S., Domínguez, S., Feliz, R., Pinillos, R., & López, J. (2014). Sacarino, a service robot in a hotel environment. ROBOT2013: First Iberian Robotics Conference, 3–14. Zeithaml, V. A. (1981). How Consumer Evaluation Processes Differ between Goods and Services. Marketing of Services, 186–190. Zeithaml, V. A., Berry, L. L., & Parasuraman, A. (1993). The nature and determinants of customer expectations of service. Journal of the Academy of Marketing Science, 21(1), 1–12.
232
References
Zhang, T., Zhu, B., Lee, L., & Kaber, D. (2008). Service robot anthropomorphism and interface design for emotion in human-robot interaction. IEEE International Conference on Automation Science and Engineering, 674–679. Zimina, A., Rimer, D., Sokolova, E., Shandarova, O., & Shandarov, E. (2016). The humanoid robot assistant for a preschool children. International Conference on Interactive Collaborative Robotics, 219–224. Złotowski, J., Proudfoot, D., Yogeeswaran, K., & Bartneck, C. (2015). Anthropomorphism: opportunities and challenges in human–robot interaction. International Journal of Social Robotics, 7(3), 347–360.