Facul t y of Appl i ed and Cr eat i ve Ar t s
Louis Laja Uggah
The Uncanny Val l ey Ef f ect s i n Di gi t al Charact ers Based on Human Physi cal Qual i t i es
Doctor of Philosophy 2016
The Uncanny Valley Effects in Digital Characters Based on Human Physical Qualities
Louis Laja Uggah
A thesis submitted
in fulfillment of the requirements for the degree of Doctor of Philosophy
Faculty of Applied and Creative Arts UNIVERSITY MALAYSIA SARAWAK
2016
UNIVERSITI MALAYSIA SARAWAK
Grade: _ _ _ __
Please tick (v)
Final Year Project Report
D
Masters
D
PhD ~
DECLARATION OF ORIGINAL WORK
This dcclaratIOn IS made on the .. \ S ...day of..."'l~ ...201~.
Student's Declaration:
I LOUIS LAJ UGGAH 11011553 hereby declare that the work entitled THE UNCANNY VALLEY EF FECTS IN DIGITAL CHARACTERS BASED ON HUMAN PHYSICAL QUALITIES is my original work. I have not copied from any other students' work or from any other sources except where due reference or acknowledgement is made explicitly in the text, nOr has any part been written for me by another person.
I S!S f .2.0l~
Date subml tted Name of the student (Matnc No.)
Supervisor's Declaration:
I DR AHMAD AZAINI BIN ABDUL MANAF hereby certifies that the work entitled THE UNCAN ;ry VALLEY EFFECTS IN DIGITAL CHARACTERS BASED ON HUMAN PHYSICAL QUALITIES was prepared by the above named student, and was submitted to the
"FACULTY" as a * partlaUfuli fulfillment for the conferment of DOCTOR OF PHILOSOPHY (pH.D) and the aforementioned work, to the best of my knowledge, is the said student's work.
(Name of the supervisor)
•
Received for examInation by: Date:_ S_ f_ !>_I'_ .L__o_l_S<>_ _ .oil.. /I \<"""D AZAIN IQ (f.! A.l'>Oll.\.- 'I\/>.,,;,F
.• eclare that Project/Thesis is classifie.d as (Please tick (0):
o
CONFIDENTIAL (Contains confidential Information under the Official Secret Act 1972)*D RESTRICTED (Contains restricted information as speciJied by the organisatIOn where research was done)*
i2]OPE:-I ACCESS
Validation of Project/Thesis
[ therefore duly affirm with free consent and willingly declare that this said Project/Thesis shall be placed officially 10 t.he Centre for Academic Information Services with the abiding interest and rights as follows:
• This ProjectlThesis is the sole legal property ofUniversiti Malaysia Sarawak (UNlMAS).
• The Centre for Academic Information Services has the lawful fight to make copies for the purpose of academic and research only and not for other purpose.
• The Centre for Academic Information Services has the lawful right to digitalise Lhe content for the Local Content Database.
• The Centre for Academic Information Services has the lawful right to make ~'Opies of t.he Project/Thesis for academic exchange between Higher Learning Institute.
• No dispute or any claim shall an se from the student itself neither third party on this ProJect/Thesis once it becomes the sole property ofUNlMAS.
• This Project/Thesis or any material, data and information related to it shall not be distributed, published or disclosed to any party by the student except with UNIMAS permISSIon.
r L~ f}
Student signa tui'e _ __ -1
r-_v)+ _ _ ___
Supervisor sIgnature:~
(Da te)
~e)
Current Address 442 LORONG B4 TAMAN BDC STAMPIN 93350 RUCHING SARAWAlC
No\c" * If the Project/ThesIs is CONFIDENTIAL or RESTRICTED, please attach together as annexure a letter from the organisatIOn with the period and reasons of confidentlaJity and restric.;tion.
[The instrument is dtlly prepared by The Centre for Academic Information ServIces]
•
ii
AUTHOR'S DECLARATION
I declare that the work in this thesis was carried out in accordance with the regulations of Universiti Malaysia Sarawak. It is original and is the result of my work, unless otherwise indicated or acknowledged as referenced work. This thesis has not been submitted at
Universiti Malaysia Sarawak or to any other academic institution or non-academic institution for any other degree or qualification.
Name of Student : Louis Laja Uggah
Student ID No : 11011553
Programme Degree : Doctor of Philosophy
Faculty : Creative and Applied Arts
Thesis Title : The Uncanny Valley Effects in Digital Characters Based on Human Physical Qualities
Signature of Student :
Date : 19th May 2016
iii
DECLARATION
This thesis contains my published works. The bibliographical details of the work and where it appears in the thesis are outlined below:
§ Uggah, L.L. & Manaf, A.A. (2015). Evaluating the Uncanny Valley Theory Based on Human Attitudes. Korean Society of Design Science, 28(2), 27-41. Retrieved 2 June, 2015, from http://dx.doi.org/10.15187/adr.2015.05.28.2.27 - (Status:
PUBLISHED), (Appears in: Chapter 2, 4, 5 and 6)
§ Uggah, L.L. & Manaf, A.A. (2015). Overcoming the Uncanny Valley Theory Based on Human Attitudes. Pertanika Journal of Social Science and Humanities, 23(S), 13-22 (Status: PUBLISHED), (Appears in: Chapter 2, 4, 5, and 6)
§ Uggah, L.L. & Manaf (2016) Moderating Effects of Stylized and Realistic Stimuli on The Relationship Between the Uncanny Valley Theory and Digital Characters' Features. 2nd International Colloquium of Art and Design Education
Research (i-CADER 2015). (Status: PUBLISHED) (Appears in: Chapter 2, 4,5 and 6)
§ Uggah, L.L. & Manaf, A.A. (2014). Evaluating the Uncanny Valley Theory Based on Human Attitudes, Global Conference on Business and Social Science 2014, PROCEEDING: ISBN 978-967013147-0-8 (Status: PUBLISHED) (Appears in: Chapter 4, 5 and 6)
Signature of Student:
Date: 15th May 2016
iv
ACKNOWLEDGEMENT
I would like to thank my current thesis supervisor Dr Ahmad Azaini bin Abdul Manaf, Lecturer for Faculty of Creative and Applied Arts, University Malaysia Sarawak for his valuable guidance in terms of thesis writting, research methodology, research analysis and journal publication. Besides that, I would like to thank him as well for his effort to develop the animation industry in my hometown Sarawak and we are hopefull that one day Sarawak will be able to produce top quality animations in the near future.
I am also grateful to other lecturers from the Faculty of Creative and Applied Arts, UNIMAS such as Dr Siti Shukhaila bt Shaharuddin, Associate Professor Dr June Ngo, Dr Musdi bin Hj, Shanat, Dr Khairul Adil and Mr Asadul Islam for their constructive comments have made a positive significant effect on my thesis work.
Finally, my sincere thanks go to my parents Douglas Uggah Embas and Doreen Mayang for their encouragement and motivation efforts for me to complete this thesis in time. My heartiest thanks also goes to my wife Marleni Thomas who always provided encouragement during tough times.
v ABSTRACT
The uncanny valley theory is an idea pioneered by Masahiro Mori in 1970 regarding the psychological effects of lifelike robotics (Mori, 1970). The uncanny valley is a phenomenon that occurs in animation and robotics, wherein things that look extremely similar to the human face, but which have slight differences from the natural appearance or the natural movements and expressions of humans, are found to appear disturbing, uncanny, and revolting (Mewes &
Heloir , 2009). This study aims to accomplish three goals: 1) analyze the participants’
attitudes towards digital characters based on a series of validated semantic differential questionnaires; 2) develop a conceptual model focusing on overcoming the uncanny valley theory in computer generated digital characters based key animation features; and 3) validate the conceptual model through structural equation model (SEM) analysis and provide guidelines and recommendations on overcoming the uncanny valley theory by avoiding negative human attitude responses. Based on results from 229 participants, this study examines the key factors of digital characters from games and movies which caused an uncanny response from participants based on their attitudes. The structural model indicated that digital characters’ facial expressions have the strongest influence on participants’
perceived humanness, followed by the stimulus’s physical movements. The digital characters’
animated hair has the next strongest influence on participants’ familiarity, followed by its facial expression.
Keywords: 3d, animation, stylized animation, realistic animation, uncanny valley
vi
Kesan Uncanny Valley Terhadap Watak-Watak Digital Berdasarkan Kualiti Fizikal Manusia
ABSTRAK
Teori "Uncanny Valley" adalah idea yang diterajui oleh Masahiro Mori pada tahun 1970 mengenai kesan psikologi robotik yang realistik (Mori, 1970). Uncanny Valley adalah satu fenomena yang berlaku dalam animasi dan robotik, di mana sesuatu perkara kelihatan amat serupa dengan muka manusia, tetapi mempunyai sedikit perbezaan dari penampilan atau pergerakan semula jadi serta air muka manusia serta didapati kelihatan terganggu, luar biasa, dan tidak sesuai (Mewes & Heloir, 2009). Kajian ini bertujuan untuk mencapai tiga matlamat: 1) menganalisis sikap peserta terhadap watak digital berdasarkan satu siri soal selidik semantik yang berlainan dan yang telah disahkan; 2) membangunkan model berkonsep yang memberi tumpuan kepada teori mengatasi lembah luar biasa dalam watak digital berkomputer berdasarkan penemuan empirik; dan 3) mengesahkan model teori dengan menyediakan garis panduan khusus dalam mengatasi teori lembah luar biasa dengan mengelakkan sikap negatif manusia. Berdasarkan keputusan daripada 229 peserta, kajian ini meneliti faktor-faktor utama watak digital dari permainan dan filem yang menyebabkan sambutan yang luar biasa daripada peserta berdasarkan sikap mereka. Model struktur menunjukkan bahawa ekspresi wajah watak digital lebih mempengaruhi sifat kemanusiaan dalam peserta, diikuti oleh pergerakan fizikal rangsangan itu. Selain itu, rambut animasi watak digital mempunyai pengaruh yang kuat atas kebiasaan peserta, diikuti dengan ekspresi wajah mereka.
Kata kunci: 3d, animasi, animasi digayakan, animasi realistik, lembah luar biasa
vii
TABLE OF CONTENTS
AUTHOR'SDECLARATION ... II
DECLARATION ... III
ACKNOWLEDGEMENT ... IV
ABSTRACT ... V
ABSTRAK ... VI
LISTOFTABLES ... XI
LIST OF FIGURES ... XII LISTOFSYMBOLS&ABBREVIATION ... XIV
CHAPTER1INTRODUCTION ... 1
1.1 Overview ... 1
1.2 Research objectives ... 1
1.3 Research problems ... 2
1.4 Scope and limitations ... 3
1.5 Significance of study ... 4
CHAPTER2LITERATUREREVIEW ... 5
2.1 What is the uncanny valley? ... 8
2.2 Causes of the uncanny valley ... 10
2.3 The uncanny valley in empirical studies ... 15
2.4 The uncanny valley effects on CGI & robotics ... 25
2.5 Criticisms of the uncanny valley theory ... 32
2.6 The uncanny valley theory & familiarity ... 42
2.7 Measuring the uncanny valley ... 49
viii
2.8 Summary ... 51
CHAPTER3CONCEPTUALFRAMEWORK ... 53
3.1 Texture, lighting & rendering ... 54
3.2 Skeletal animation & rigging (Motion) ... 56
3.3 Digital eyes ... 58
3.4 Facial expression ... 59
3.5 Hair animation ... 61
3.6 Lip sync ... 62
CHAPTER4METHODOLOGY ... 65
4.1 Research design ... 65
4.2 Sampling ... 66
4.3 Data collection ... 67
4.4 Pilot test ... 68
4.5 Data analysis method ... 70
4.6 Stimuli ... 71
4.61 Clementine ... 72
4.62 ElastiGirl ... 74
4.63 Digital Emily (Image Metrics) ... 76
4.64 Madison Paige (Heavy Rain) ... 78
4.65 Human (Controlled stimulus) ... 79
4.7 Measurement model ... 80
4.71 Exploratory factor analysis ... 80
4.72 Construct validity test ... 81
4.73 Reliability test ... 82
ix
4.8 Confirmatory factor analysis ... 83
CHAPTER5RESULTSANDFINDINGS ... 85
5.1 Analysis of variance (ANOVA) ... 86
5.11 Digital eyes (Fake-lifelike) ... 86
5.13 Digital hair (Fake-lifelike) ... 89
5.14 Lip sync (Fake-lifelike) ... 90
5.15 Motion (Fake-lifelike) ... 91
5.16 Lighting (Fake-lifelike) ... 92
5.17 Digital eyes (Mechanical-organic) ... 93
5.18 Facial expression (Mechanical-organic) ... 94
5.19 Digital hair (Mechanical-organic) ... 96
5.2 Lip sync (Mechanical-organic) ... 97
5.21 Motion (Mechanical-organic) ... 98
5.22 Lighting (Mechanical-organic) ... 99
5.23 Digital eyes (Eerie-familiar) ... 100
5.25 Digital hair (Eerie-familiar) ... 102
5.26 Lip Sync (Eerie-familiar) ... 103
5.27 Motion (Eerie-familiar) ... 104
5.28 Lighting (Eerie-familiar) ... 105
5.29 Familiarity (Eerie-reassuring) ... 106
5.3 Familiarity (Freaky-numbing) ... 107
5.31 Familiarity (Boring-shocking) ... 108
5.32 Perceived humanness (Synthetic-real) ... 109
5.33 Perceived humanness (Inanimate-living) ... 110
x
5.34 Perceived humanness (Humanmade-humanlike) ... 111
5.35 Scatterplot data ... 112
5.4 Assessment of model fit ... 113
5.5 Assessment of structural model ... 114
5.6 Moderating effects ... 116
CHAPTER6DISCUSSION,IMPLICATIONS&CONCLUSION ... 120
6.1 Hypotheses testing ... 121
6.2 Summary ... 127
6.3 Implication of the research ... 131
6.31 Theoretical implications ... 131
6.32 Implications on guidelines ... 132
6.4 Overall contributions ... 135
6.5 Future researches ... 137
6.6 Conclusion ... 138
REFERENCES ... 140
xi List of Tables
Table 1: Profiles of the survey participants ... 67
Table 2: Measurement model items ... 69
Table 4: Moderation effects of stimuli ... 117
Table 5: Moderation effects of stylised and realistic stimuli ... 118
Table 6: Hypothesis testing for both types of stimuli (Realistic & Stylised) ... 121
Table 7: Hypothesis testing for stylised stimuli ... 122
Table 8: Hypothesis testing for realistic stimuli ... 123
Table 9: Hypothesis testing for familiarity and perceived humanness ... 124
xii List of Figures
Figure 1: 3D-Generated face featured in Futureworld (Golder, 2011) ... 6
Figure 2: Star Wars: A New Hope (Golder, 2011) ... 6
Figure 4: Proposed conceptual framework ... 54
Figure 10: The Final measurement model ... 84
Figure 11: Digital eyes (Fake-lifelike) ... 87
Figure 12: Facial expression (Fake-lifelike) ... 88
Figure 24: Facial expression (Eerie-familiar) ... 101
Figure 25: Digital hair (Eerie-familiar) ... 102
Figure 26: Lip sync (Eerie-familiar) ... 103
Figure 28 : Lighting (Eerie-familiar) ... 105
Figure 30: Familiarity (Freaky-numbing) ... 108
Figure 31 : Familiarity (Boring-shocking) ... 109
Figure 32: Perceived humanness (Synthetic-real) ... 110
Figure 33 : Perceived humanness (Inanimate-living) ... 111
Figure 34: Perceived humanness (Humanmade-humanlike) ... 112
Figure 36: Structural equation model ... 114
Figure 37 : Stylised digital character ... 128
xiii
Figure 38 : Realistic humanlike digital character ... 129 Figure 39: Clementine & Madison Paige ... 130
xiv
List of Symbols and Abbreviation
3D Three Dimensional
CGI Computer Generated Imagery
FIFA Football International Federation Association
EA Electronic Arts
CFA Confirmatory Factor Analysis
SPSS Statistical Package for the Social Sciences
CFI Comparative Fit Index
AGFI Goodnesss of Fit Index
RMSEA Mean Square Error of Approximation AVE Average Variance Extracted
xv
List of Appendices
Appendix A : KMO and Bartlett's test ... 150
Appendix B : Communalities table ... 151
Appendix C : Pattern matrix ... 153
Appendix D : ANOVA results (Group 1) ... 155
Appendix E : Post-hoc Test (EYE1) ... 157
Appendix F : Post-hoc test (FACIAL1) ... 158
Appendix G : Post-hoc test (HAIR1) ... 160
Appendix: Post-hoc test (LIP1) ... 161
Appendix I :Post-hoc test (MOVE1) ... 162
Appendix J :Post-hoc test (TXT1) ... 163
Appendix K : ANOVA results (Group 2) ... 164
Appendix L :Post-hoc test (EYES2) ... 165
Appendix M :Post-hoc test (FACIAL2) ... 166
Appendix N :Post-hoc test (HAIR2) ... 167
Appendix O :Post-hoc test (LIP2) ... 168
Appendix P :Post-hoc test (MOVE2) ... 169
Appendix Q :Post-hoc test (TXT2) ... 170
xvi
Appendix: ANOVA results (Group 3) ... 172
Appendix S :Post-hoc test (EYES3) ... 173
Appendix T:Post-hoc test (FACIAL3) ... 174
Appendix U:Post-hoc test (HAIR3) ... 175
Appendix V:Post-hoc test (LIP3) ... 176
Appendix W :Post-hoc test (MOVE3) ... 177
Appendix:Post-hoc test (TXT3) ... 178
Appendix: ANOVA results (Humaness & familiarity groups) ... 179
Appendix Z:Regression weights (Realistic model) ... 181
Appendix AA Regression weights: (Stylized) ... 183
Appendix BB Questionnaires...185
CHAPTER 1 INTRODUCTION
1.1 Overview
The ‘uncanny valley’ theory is an idea pioneered by Masahiro Mori in 1970 with regards to the psychological effects of lifelike robotics (Mori, 1970). Uncanny valley refers to a graph which compares human likeness and familiarity; it indicates that familiarity will increase as human likeness increases to a certain point before the familiarity takes a swift dive into negativity as the likeness is either very similar or not yet similar enough. Familiarity then rises again as likeness increases and approaches the appearances of a natural human face (Mori, 1970). However, currently, there has been a lack of research on the movement and speech of realistic digital and stylised characters which may prompt the studios, designers and animators to opt for stylised digital characters to avoid negative emotional responses from viewers. This study seeks to develop a conceptual model which focuses on overcoming the uncanny valley theory in computer-generated digital characters based on empirical findings.
Finally, this study also aims to validate the conceptual model by providing a structural equation model and moderation effect analyses of six animation features based on the audiences’ perceptions of humanness and familiarity.
1.2 Research objectives
➢ To examine both the existing theories and empirical findings related to the uncanny valley theory in computer-generated digital characters.
2
➢ To analyse participants’ attitudes towards realistic digital characters using the analysis of variance (ANOVA) method based on a series of validated semantic differential questionnaires.
➢ To develop a conceptual model which focuses on overcoming the uncanny valley theory in computer-generated digital characters based on empirical findings as guidelines for animators.
➢ To validate the conceptual model through the analysis of structural equation model (SEM) and moderation effects to provide guidelines for animators to overcome the uncanny valley effect in stylised and realistic digital characters.
1.3 Research problems
Overcoming the uncanny valley theory effects is still an area of long standing confusion. For decades, animation companies such as Pixar prefer to opt for stylised animations instead of realistic animations to avoid negative audience reaction (Wolchover, 2011). This could be due to Masahiro’s uncanny valley theory; however, there are inadequate empirical data to support and validate Masahiro's recommendation for avoiding realistic animation altogether (Segaya & Nagayama, 2009; Scheidel, 2009). Furthermore, even though there are multiple studies conducted with robots and still images of digital characters, the number of research which focused on moving and talking digital characters is still limited.
For example, negative responses from the audiences on realistic but emotionless and lifeless digital characters in films such as The Final Fantasy, Beowulf and Polar Express have created an issue for production companies (Segaya & Nakashima, 2007; Cheetham, Suter, Jancke,
3
2011). Therefore, as hypothesised by Mori (1970), the solution to the uncanny valley effects is not to attempt to bridge it, but to avoid it altogether by allowing artificial things to remain looking artificial. Consequently, many animation studios have taken Mori’s advice to their hearts, regardless of the empirical evidences for or against the existence of uncanny valley.
This is because uncanny valley has firmly entrenched sufficient concept in animations and robotics that it has simply been accepted. Studios such as Pixar, for example, are famous for consistently creating characters that are noticeably non-human in an attempt to avoid uncanny valley. On the other hand, vague and conflicting recommendations on how to overcome the uncanny valley theory have been debated by researchers such as Hanson, D., Olney, A., Pereira, I. A., & Zielke, M (2005), Tinwell, Grimshaw, & Williams (2011) and Karl Macdorman. This unresolved problem has led to confusion among designers and animators and thus, should be addressed.
1.4 Scope and limitations
The main objective of this study is to analyse human attitudes' towards various digital characters as stimuli in order to examine the uncanny valley effects. However, the results of this study are limited in certain ways; thus, in order to generalise the findings, this should be verified. One of the limitations of this study is that fact data collection was only conducted in a single country, Malaysia, and all the participants were Malaysians. The results were also based on a limited number of stimuli which consisted of 2 stylised, 2 realistic digital characters and a control stimulus due to time to collect the data from 229 participants.
4 1.5 Significance of study
The significance of this study can be divided into two parts. The first significant is that the conceptual framework developed during this research will be able to guide novice animators, studios and students to identify the key factors which are responsible for the occurrence of uncanny valley effects in digital characters. This is because although several previous research have pointed out that digital eyes and facial expressions as the main factors that caused the uncanny valley effect to occur, this claim remains a speculation as it was not supported by any empirical evidences. In addition, other features such as hair animation, motion movements and rendering techniques should also be considered. By identifying the factors that are more sensitive towards the uncanny valley effect, this will make animators to be more aware of these factors when modelling and animating their models in the development of stylised and realistic digital characters which requires different techniques as both characters are different from each other. Currently, there has not been any research which examines the moderation effects of these types of stimuli between animation techniques and human attitudes such as familiarity and perceived humanness.
The second significance of this research is that it disagrees with Mori’s (1970) recommendations to completely avoid moving realistic characters to prevent the uncanny effects. This research hypothesize that realistic humanlike digital characters such as Digital Emily is not fallen in the uncanny valley category as suggested by Mori (1970).
5
CHAPTER 2
LITERATURE REVIEW
Since the early 90s, computer-generated imagery (CGI) technology has changed the visual effects industry in filmmaking. CGI provides filmmakers the freedom to develop and create photorealistic digital characters and environments in their films. The visual effects in film-making can be divided into four categories: matte paintings, live-action effects, digital animation and digital effects. The 3-dimensional (3D) animation emerged and surpassed other leading animation methods such as stop-motion animations and drawn animation. The first major feature film to adapt 3D animation technique was Futureworld which was released in 1976. Prior to this, two graduate students from the American International University of Utah, Edwin Catmull and Fred Parke used the 3D animation concepts to develop computer- generated hands and faces which initially appeared in an experimental short film, A Computer Animated Hand in 1971. The next feature film to apply the 3D animation technique was Star Wars: A New Hope in 1977 which was developed by Lucasfilm and directed by George Lucas. The 3D animation techniques were also applied to develop wireframe imageries and the targeting computers in the X-Wings, Tie fighters and the Millennium Falcon. The film went on to be one of the most financially successful films of all time and received praises for its spectacular visual effects.
