Ruttkay / Pelachaud From Brows to Trust

Evaluating Embodied Conversational Agents
1. Auflage 2006
ISBN: 978-1-4020-2730-7
Verlag: Springer Netherlands
Format: PDF
Kopierschutz: 1 - PDF Watermark

E-Book, Englisch, Band 7, 372 Seiten

Reihe: Human-Computer Interaction Series

Embodied conversational agents (ECAs) are autonomous software entities with human-like appearance and communication skills. These agents can take on a number of different roles, for example, as an assistant, tutor, information provider, or customer service agent. They may also simply represent or entertain a user. The precise nature and benefits of different characteristics of ECAs requires careful investigation. Questions range from the function of an eyebrow raise to mechanisms for assessing and improving ECA trustworthiness. This book will help experts and designers in the specification and development of applications incorporating ECAs. Part 1 provides guidelines for evaluation methodologies and the identification of design and evaluation parameters. Part 2 demonstrates the importance of considering the user's perspective and interaction experience. Part 3 addresses issues in fine-tuning design parameters of ECAs and verifying the perceived effect. Finally, in Part 4 lessons learned from a number of application case studies are presented. The book is intended for both ECA researchers in academia and industry, and developers and designers interested in applying the technology.

Autoren/Hrsg.

Ruttkay, Zsófia

Pelachaud, Catherine

Inhaltsverzeichnis

1;Contents;6
2;Contributing Authors;8
3;Preface;16
4;I EVALUATION METHODOLOGY;21
4.1;Chapter 1 THE BLIND MEN AND THE ELEPHANT REVISITED;23
4.1.1;1. Introduction;24
4.1.2;2. The Taxonomy;26
4.1.2.1;2.1 Believability;32
4.1.2.2;2.2 Sociability;34
4.1.2.3;2.3 Task and Application Domains;36
4.1.2.4;2.4 Agency and Computational Issues;37
4.1.3;3. Applying the Taxonomy;38
4.1.4;4. On Production Values;40
4.1.5;5. Further Classi.cation;41
4.1.6;6. Conclusions;42
4.1.7;Notes;43
4.1.8;References;43
4.2;Chapter 2 EMBODIED CONVERSATIONAL AGENTS ON A COMMON GROUND;47
4.2.1;1. Introduction;48
4.2.1.1;1.1 Motivations and Problems;49
4.2.2;2. ECAs from a Design Perspective;50
4.2.2.1;2.1 The Embodiment;52
4.2.2.2;2.2 The Mental Aspects;56
4.2.2.3;2.3 Implementation Aspects;59
4.2.2.4;2.4 Range of Applicability;60
4.2.3;3. On Evaluation Methodology;60
4.2.3.1;3.1 Why to Evaluate?;61
4.2.3.2;3.2 How to De.ne the Evaluation Variables?;66
4.2.3.3;3.3 Testing by what Users?;68
4.2.3.4;3.4 How to Collect and Evaluate Data?;71
4.2.4;4. Dimensions of Evaluation;73
4.2.4.1;4.1 Usability;74
4.2.4.2;4.2 Evaluation of User Perception of ECAs;76
4.2.5;5. Conclusions;81
4.2.6;Notes;82
4.2.7;References;83
4.3;Chapter 3 EMPIRICAL EVALUATION METHODOLOGY FOR EMBODIED CONVERSATIONAL AGENTS;87
4.3.1;1. Introduction;88
4.3.2;2. Types of Evaluation Research;88
4.3.2.1;2.1 Star Life Cycle;89
4.3.2.2;2.2 Formative and Summative Evaluation;91
4.3.3;3. Research Model, Research Question and Construct;92
4.3.3.1;3.1 Research Model;92
4.3.3.2;3.2 Research Questions;94
4.3.3.3;3.3 Psychological Construct;95
4.3.4;4. Research Strategy;96
4.3.4.1;4.1 Survey;97
4.3.4.2;4.2 Experiment;98
4.3.4.3;4.3 Case Study;99
4.3.5;5. Data Collection Methods;100
4.3.5.1;5.1 Qualitative Methods;101
4.3.5.2;5.2 Quantitative Methods;102
4.3.5.3;5.3 More Data Collection Methods;106
4.3.6;6. Samples, Reliability and Validity;106
4.3.6.1;6.1 Samples;106
4.3.6.2;6.2 Reliability;108
4.3.6.3;6.3 Validity;110
4.3.7;7. Data Analysis;112
4.3.8;8. Concluding: Guidelines for Evaluating ECAs;114
4.3.9;Acknowledgments;116
4.3.10;Notes;116
4.3.11;References;116
4.4;Chapter 4 EVALUATING USERS’ REACTIONS TO HUMAN-LIKE INTERFACES;121
4.4.1;1. Evaluation of Dialogue Systems;122
4.4.1.1;1.1 Evaluation of User’s Satisfaction;122
4.4.1.2;1.2 Evaluation Criteria for Multimodal Dialogue Systems;123
4.4.2;2. Prosodic Cues and Non-verbal Behaviour as new Evaluation Measures;124
4.4.2.1;2.1 Prosodic and Visual Cues of Emotions;125
4.4.2.2;2.2 Prosodic and Visual Cues of Emotions and Evaluation;125
4.4.3;3. The Investigation;126
4.4.3.1;3.1 Material;126
4.4.3.2;3.2 Analysis of Prosodic Cues;130
4.4.3.3;3.3 Analysis of Non-verbal Communicative Behaviour;135
4.4.4;4. Conclusions and Further Investigation;140
4.4.5;Acknowledgments;141
4.4.6;Notes;141
4.4.7;References;141
5;II THE USER IN FOCUS;145
5.1;Chapter 5 USER-CENTRED DESIGN AND EVALUATION OF AFFECTIVE INTERFACES;147
5.1.1;1. Introduction;148
5.1.2;2. Underlying Philosophy and Method;149
5.1.2.1;2.1 Our Philosophy;151
5.1.2.2;2.2 Our Method;152
5.1.3;3. Studies of Three Affective Interfaces;157
5.1.4;4. Agneta & Frida;158
5.1.4.1;4.1 Non-correlation of Measurements;159
5.1.4.2;4.2 Narrative Experience;161
5.1.4.3;4.3 Implications for Design Method;162
5.1.5;5. The Influencing Machine;163
5.1.5.1;5.1 Study Method;164
5.1.5.2;5.2 The First In.uencing Machine Study;166
5.1.5.3;5.3 The Second Influencing Machine Study;167
5.1.5.4;5.4 Implications for Design Method;169
5.1.6;6. SenToy and FantasyA;170
5.1.6.1;6.1 Wizard of Oz;171
5.1.6.2;6.2 Second Study of SenToy Used in FantasyA;173
5.1.6.3;6.3 Implications for Design Method;174
5.1.7;7. Discussion;175
5.1.8;Acknowledgments;177
5.1.9;Notes;177
5.1.10;References;177
5.2;Chapter 6 ‘USER AS ASSESSOR’ APPROACH TO EMBODIED CONVERSATIONAL AGENTS;181
5.2.1;1. Introduction;182
5.2.2;2. Attention as an Example of the ‘User as Assessor’ Approach;186
5.2.2.1;2.1 How Do Humans Attend?;186
5.2.2.2;2.2 Conceptualizing Apparent Attention;188
5.2.3;3. Creating Apparently Attentive ECAs;190
5.2.3.1;3.1 Apparent Selectivity in ECAs;190
5.2.3.2;3.2 Apparent Breadth in ECAs;195
5.2.4;4. Conclusion;199
5.2.5;Notes;201
5.2.6;References;201
6;III EVALUATION OF ECAS;209
6.1;Chapter 7 MORE ABOUT BROWS;211
6.1.1;1. Introduction;212
6.1.2;2. About Brows;213
6.1.3;3. Materials;216
6.1.3.1;3.1 Speech;216
6.1.3.2;3.2 Animations;218
6.1.4;4. Experiment 1: Subjective Preference;219
6.1.4.1;4.1 Method;219
6.1.4.2;4.2 Results (Dutch);220
6.1.4.3;4.3 Results (Italian);220
6.1.4.4;4.4 Discussion;221
6.1.5;5. Experiment 2: Perceived Prominence;221
6.1.5.1;5.1 Method;221
6.1.5.2;5.2 Results (Dutch);222
6.1.5.3;5.3 Results (Italian);224
6.1.5.4;5.4 Discussion;224
6.1.6;6. Experiment 3: Functional Analysis;225
6.1.6.1;6.1 Method;225
6.1.6.2;6.2 Results (Dutch);226
6.1.6.3;6.3 Results (Italian);227
6.1.6.4;6.4 Discussion;227
6.1.7;7. General Discussion;228
6.1.7.1;7.1 Eyebrows in Dutch and Italian;228
6.1.7.2;7.2 About Analysis-by-Synthesis;230
6.1.7.3;7.3 Analysis-by-observation;231
6.1.8;Acknowledgments;233
6.1.9;Notes;233
6.1.10;References;234
6.2;Chapter 8 EVALUATION OF MULTIMODAL BEHAVIOUR OF EMBODIED AGENTS;237
6.2.1;1. Introduction;238
6.2.2;2. Experimental Setting;240
6.2.2.1;2.1 Participants;240
6.2.2.2;2.2 Apparatus;241
6.2.2.3;2.3 Scenarios;241
6.2.2.4;2.4 Independent Variables;242
6.2.2.5;2.5 Generation of Multimodal Behaviour;245
6.2.2.6;2.6 Dependent Variables;246
6.2.2.7;2.7 Data Analysis;247
6.2.3;3. Results;247
6.2.3.1;3.1 Subjective Variables;247
6.2.3.2;3.2 Recall Performance;250
6.2.4;4. Discussion;251
6.2.4.1;4.1 Effects of Multimodal Strategies;252
6.2.4.2;4.2 Effects of ECAs’ Appearance;253
6.2.4.3;4.3 Additional Results;254
6.2.5;5. Conclusions and Future Directions;254
6.2.6;Acknowledgments;256
6.2.7;Notes;256
6.2.8;References;256
6.3;Chapter 9 ECA AS USER INTERFACE PARADIGM;259
6.3.1;1. Introduction;260
6.3.1.1;1.1 Our Research Objectives;262
6.3.2;2. Research Framework;264
6.3.2.1;2.1 Features of the User;265
6.3.2.2;2.2 Features of the ECA;267
6.3.2.3;2.3 Features of the Task;270
6.3.2.4;2.4 Interaction of Variables;273
6.3.2.5;2.5 Approaches to Assessing ECAs;274
6.3.3;3. An Experimental Study;274
6.3.3.1;3.1 Goal and Hypotheses;275
6.3.3.2;3.2 Participants, Materials, and Procedure;276
6.3.3.3;3.3 Results;279
6.3.3.4;3.4 Additional Observations from the Experiment;282
6.3.4;4. Conclusions;283
6.3.5;Notes;284
6.3.6;References;284
7;IV EVALUATION OF APPLICATIONS;289
7.1;Chapter 10 TALKING TO DIGITAL FISH;291
7.1.1;1. Introduction;292
7.1.1.1;1.1 Evaluating Animated Characters in Educational Software;293
7.1.1.2;1.2 Goals of the Study;294
7.1.2;2. Methods;294
7.1.2.1;2.1 Participants, Task, and Procedure;294
7.1.2.2;2.2 Simulation Environment;296
7.1.2.3;2.3 Text to Speech Manipulation;297
7.1.2.4;2.4 Research Design and Analyses;298
7.1.2.5;2.5 Data Coding and Dependent Measures;299
7.1.3;3. Results;302
7.1.3.1;3.1 Engagement in Interface and Ease of Use;302
7.1.3.2;3.2 Distribution of Question Types;304
7.1.3.3;3.3 Impact of TTS Voice Type on Child Queries;304
7.1.4;4. Discussion;305
7.1.4.1;4.1 Acoustic Characteristics of Animated Character Design;305
7.1.4.2;4.2 Conversational Interfaces as Educational Interfaces;308
7.1.4.3;4.3 Conclusion;309
7.1.5;Acknowledgments;310
7.1.6;Notes;310
7.1.7;References;310
7.2;Chapter 11 EXPERIMENTAL EVALUATION OF THE USE OF ECAS IN ECOMMERCE APPLICATIONS;313
7.2.1;1. Introduction;314
7.2.2;2. Experimental Approach;315
7.2.3;3. The Phoebe Experiment;317
7.2.3.1;3.1 Experimental Procedure;319
7.2.3.2;3.2 The Usability Questionnaire;321
7.2.3.3;3.3 Results;321
7.2.3.4;3.4 Summary;324
7.2.4;4. The Voice Personae Experiment;324
7.2.4.1;4.1 Voice Personae;327
7.2.4.2;4.2 Experimental Procedure;328
7.2.4.3;4.3 Results;329
7.2.4.4;4.4 Summary;332
7.2.5;5. The Trust Experiment;332
7.2.5.1;5.1 Experimental Procedure;335
7.2.5.2;5.2 Results;335
7.2.5.3;5.3 Summary;338
7.2.6;6. Conclusion;339
7.2.7;References;340
7.3;Chapter 12 WHAT WE CAN LEARN FROM AVATAR-DRIVEN INTERNET COMMUNITIES;343
7.3.1;1. Introduction;344
7.3.2;2. The sysis NetLife;346
7.3.2.1;2.1 The NetLife Platform;346
7.3.2.2;2.2 The Flirtboat Application;347
7.3.2.3;2.3 The derSpittelberg Application;350
7.3.3;3. User Data Analysis;351
7.3.3.1;3.1 Goals for Data Collection;352
7.3.3.2;3.2 Methodology;353
7.3.3.3;3.3 Generalized Avatar Pro.le;355
7.3.3.4;3.4 Lifestyle Analysis;360
7.3.4;4. Data Evaluation;361
7.3.4.1;4.1 Evaluation Results;362
7.3.5;5. Conclusion;370
7.3.6;Acknowledgments;371
7.3.7;Notes;371
7.3.8;References;372
8;More eBooks at www.ciando.com;0

Leseproben

2. About Brows (p.193-194)

In a seminal paper, Ekman (1979) describes the role of eyebrow movements as emotional and conversational signals. Sometimes the distinction between these two kinds of signals is di.cult to make (for instance because both often occur during conversation). Still clear di.erences between the two exist: conversational signals typically do not occur when a person believes (s)he is unobserved, while emotional signals do. Moreover, emotional but not conversational signals are believed to be universal.

While the use of eyebrows as emotional signals has been addressed in many studies (already in Darwin (1872)), the conversational use is still relatively understudied and most of the work that has been done in this area is based on intuitions and impressionistic observations. This is surprising, since eyebrow movements are according to Ekman (1979:183) "probably among the most frequent facial actions employed as conversational signals". Various authors have suggested that eyebrow movements can be used to emphasize important pieces of information (see e.g., Birdwhistell (1970); Eibl-Eibesfeldt (1972); Condon (1976); Ekman (1979)). Ekman observes that eyebrows can play this accentuation role in two di.erent ways: they can function as a baton (in the terminology of Efron (1941)), which may be used to accentuate a particular word as it is spoken, or they can function as an underliner (in Ekman’s own terminology), where the emphasis stretches out over more than one word.

It is well-known that speakers may use auditory speech signals to emphasize words as well. For instance, speakers of Germanic languages (such as Dutch, English and German) can use pitch accents to indicate the information status of words: accents tend to distinguish information that is in focus (since it is new or contrastive) from information which is given from the prior discourse context (see e.g., Chafe (1974); Terken (1984); Hirschberg (1993)). That both eyebrow movements and pitch accents can be used to signal focus, suggests that there is close correspondence between the two. This correspondence has indeed been noted by Morgan (1953) and Bolinger (1985:202.).

The latter formulated his Metaphor of Up and Down which implies, among other things, that when the pitch rises or falls, eyebrows tend to follow the same pattern. As an illustration of this metaphor, it is instructive to try and utter a two-word phrase, say "blue square," with a pitch accent (and no corresponding eyebrow movement) on the word "blue" and an eyebrow movement (but no pitch accent) on the word "square".

This implies that eyebrow movements often co-occur with pitch accents. It is important to realize that the opposite is not the case. Ekman (1979:184): "There are many occasions when people mark emphasis in their speech without either a baton or an underliner." People do more with their pitch than with their eyebrows, as the reader can easily verify by looking at an arbitrary speaker. If not all emphasized words are accompanied by an eyebrow movement, which words are? This is still an open question. Ekman (1979:184) is "not optimistic about being able to predict when a baton or underliner will be used and when emphasis will be carried just by voice, although perhaps there might be some weak relationship with overall involvement in what is said."