ResearchPDF Available

The Uncanny Valley Effect: Implications on Robotics and A.I. Development


Abstract and Figures

This paper aims to explore the uncanny valley hypothesis and its implications on future development of humanoid robots and artificial intelligence. The original hypothesis by Mori (1970) will be introduced in regards to current research focus and criticism received. The succeeding sections present theoretical approaches in evolutionary psychology, as well as empirical studies within cognitive neuroscience; validity and representativeness of research findings will then be reviewed. To conclude, implications of findings on the uncanny valley will be discussed in relation to general directions of future robotics and A.I. design.
Content may be subject to copyright.
University of CopenhagenMA Cognition and Communication
Introduction to Cognitive Science Fall 2016 Exam
The Uncanny Valley Effect and its Implications on Robotics and A.I. Development Yum-Mei Simone Wong (stl350)
The Uncanny Valley Effect:
Implications on Robotics and
A.I. Development
Date: Jan 2, 2017
Name: Yum-Mei Simone Wong
Student ID: stl350
Number of standard pages: 13.43
Number of keystrokes: 32221
University of CopenhagenMA Cognition and Communication
Introduction to Cognitive Science Fall 2016 Exam
The Uncanny Valley Effect and its Implications on Robotics and A.I. Development Yum-Mei Simone Wong (stl350)
1. Introduction
The rapid development in technology and computer science in recent years acknowledges
the significance of building human-like robots and artificial intelligence. For example, an
apparatus that is indistinguishable from humans may be used in controlled experiments as a
‘participant’ to test for cognitive theories (MacDorman, 2005). In addition to experimental
purposes within academic research, there is an increasing trend in building robots that
simulate humans physically and virtually to provide aid in every day situations or even to
replace human labour.
Modern androids are increasingly penetrating the social realm of humans on everyday
interactions and activities. Within humanoid robotics and service robotics community, one of
the biggest aims is the ability to design robots that interact with people in social environments.
For example, providing assistance to the disabled and the elderly, or as teaching assistants
in educational settings (Kawamura et al., 1996; Nourbakhsh et al., 1999; as cited in Breazeal,
2003). Humanoid robots have also been used clinically to encourage social interaction skills
of autistic children; they have been suggested to provide a ‘simplified, safe, predictable and
reliable environment’ for patients with impaired social skills as a starting point for therapeutic
intervention (Robins et al., 2005, p.107).
Sociable robots that simulate human appearance, facial expressions and gestures are
capable of eliciting emotional responses on humans during face-to-face interactions that
mechanical robots could not induce (Breazeal, 2003). It is therefore important to consider
how imposing human-like features can possibly affect human-robot interactions. For
example, does realism of anthropomorphic appearances influence trustworthiness or
friendliness of robots? To what extent should we exploit the familiarity effect of implementing
anthropomorphic features to machines, such that designers could control whether the
product is treated more as humanistic agents or, simply, an interactive computer? The
answers to the above questions would vary depending on the circumstances and purpose of
the machine – nonetheless, they raise the concerns of using robots beyond mechanical and
industrial purposes. As technical advances have allowed the construction of extremely
realistic robots, their capability of inducing emotional responses on humans is therefore of
high importance for the development in robotics and A.I. design.
University of CopenhagenMA Cognition and Communication
Introduction to Cognitive Science Fall 2016 Exam
The Uncanny Valley Effect and its Implications on Robotics and A.I. Development Yum-Mei Simone Wong (stl350)
A well-known and influential concept in the field of robotics is the ‘uncanny valley’, which has
been a critical topic in exploring human reactions towards anthropomorphic robots. The term
describes the negative responses of humans to agents with slightly imperfect human-
likeness, which induced feelings of eeriness and discomfort (Mathur and Reichling, 2009).
By understanding possible explanations and causes of the uncanny valley effect, we can gain
insight to whether robotics and AI should be implemented with human-like features, and to
what extent this will aid future research and application to everyday communicative uses.
This paper aims to explore the uncanny valley hypothesis and its implications on future
development of humanoid robots and artificial intelligence. The original hypothesis by Mori
(1970) will be introduced in regards to current research focus and criticism received. The
succeeding sections present theoretical approaches in evolutionary psychology, as well as
empirical studies within cognitive neuroscience; validity and representativeness of research
findings will then be reviewed. To conclude, implications of findings on the uncanny valley
will be discussed in relation to general directions of future robotics and A.I. design.
2. Background of the Uncanny Valley Hypothesis
2.1 The original uncanny valley hypothesis (Mori, 1970).
The Uncanny Valley hypothesis originated from Japanese roboticist Masahiro Mori’s
research in 1970. Mori observed that as human-likeness of robots increased, familiarity
ratings by observers also increased. However, this positive correlation inverted just when the
robots reached a degree of realism that they gave an almost, but not perfectly, human-like
appearance - familiarity was rated exceptionally negative by the observers. Mori (1970)
graphed this relationship between human realism and familiarity ratings, and the sudden ‘dip’
into negative peak at the ‘almost human level’ was named the ‘uncanny valley’ (bukimi no
tani in Japanese), an analogy of the graph’s shape as a mountain (see Figure 1).
The uncanny valley phenomenon was identified when robots appeared too close to being
human that it became strange than familiar to individuals. In the original study, Mori (1970)
also included stimuli such as corpses, zombies and lifelike prosthetic hands that showed the
same effect (Tinwell et al., 2011). This suggested that the phenomenon is not restricted to
only human-like robots, but also objects that resembles human figures or limbs. For example,
a prosthetic hand can simulate all the physical features of a real limb: muscles, tendons,
University of CopenhagenMA Cognition and Communication
Introduction to Cognitive Science Fall 2016 Exam
The Uncanny Valley Effect and its Implications on Robotics and A.I. Development Yum-Mei Simone Wong (stl350)
veins, skin pigmentation, finger prints… to a degree where it is indistinguishable from real
hands at first glance. Due to the lack of body warmth and soft tissues, individuals may
experience the sensation of uneasiness if the prosthetic hand were to be held or touched.
This uncanny effect, Mori argued, would potentially be increased with the addition of
automatic movement (MacDorman, 2005).
Figure 1. Mori (1970)'s graph displaying the relationship between human likeness of
various humanoid objects and familiarity ratings. Familiarity increased along with human
likeness until it reached the 'uncanny valley', where the subtle differences produced an
uncomfortable effect on observers (MacDorman, 2005).
2.2 Criticism on Mori’s theory.
The validity of uncanny valley hypothesis has been questioned. For example, interpretations
of the original concepts of ‘uncanny’, ‘familiarity’ and ‘affinity’ vary between translations from
the Japanese study; this slight level of ambiguity and variance in wording may cause
problems when comparing results for dependent variables. Researchers have recently
attempted to refine the measurements for levels of human-likeness, eeriness and other
indices with the aim to produce more representative ratings (Ho and MacDorman, 2016).
Research has also reported on failure to replicate the uncanny effect, such that modifications
of human-likeness on visual stimuli did not produce an inverted rating of affinity or familiarity
on participants as the original study found (Bartneck et al., 2009). The simplicity of Mori’s
University of CopenhagenMA Cognition and Communication
Introduction to Cognitive Science Fall 2016 Exam
The Uncanny Valley Effect and its Implications on Robotics and A.I. Development Yum-Mei Simone Wong (stl350)
graph has also been criticised for being misleading, as it is unlikely that only two factors
(human-likeness and perceived familiarity) mediate our uncanny feeling. Other factors, such
as attractiveness, has also been found to be correlated with human-likeness (Hanson, 2006).
In addition, environmental factors have also been found to affect the uncanny valley; some
suggested that individual differences on uncanny valley sensitivity may be moderated by both
sociocultural constructions and biological adaptations (MacDorman and Entezari, 2015).
Whilst there is a possibility of defining a multi-dimensional model to the uncanny effect, further
investigation is still required.
2.3 Current research on the uncanny valley.
Despite studies that go against the theory, Mori (1970)’s hypothesis has been extensively
studied across a broad range of academic and industrial fields, including cognitive
neuroscience, social psychology, robotics and digital graphic design. The theory’s intuitive
approach to understanding human-computer interaction also raised appeal in popular culture,
such that it was used to explain negative reactions to past animated films with realistic
computer generated graphics, for example the poor responses to the film The Polar Express
(2004), due to reported ‘animation eerie’ (Geller, 2008). Majority of empirical research on the
uncanny valley effect has focused on providing a theoretical explanation to the phenomenon,
as well as exploring what factors influence the arousal of negative affinity and eeriness
(Misselhorn, 2009). Many have proposed possible explanations of the uncanny valley that
can be divided into two main approaches: those that suggest it to be stimulus-driven, early
perceptual processing that holds an evolutionary purpose, and those that explain the effect
with a more general range of late cognitive processing (MacDorman et al., 2009).
University of CopenhagenMA Cognition and Communication
Introduction to Cognitive Science Fall 2016 Exam
The Uncanny Valley Effect and its Implications on Robotics and A.I. Development Yum-Mei Simone Wong (stl350)
3. Evolutionary psychology approach on the uncanny valley
Explanations that emphasise on the role of early perceptual processing often raise the
importance of individuals’ ability to detect human-likeness as a form of self-protection and
understanding of others’ intentions. Researchers have hereby attempted to explain the
uncanny valley from an evolutionary psychology approach, suggesting that it serves an
evolutionary purpose to enhance our fitness and to solve certain adaptive problems. This
section explores the uncanny valley in relation to proposed explanations that mainly concern
the following factors: pathogen avoidance, threat avoidance, and as indicators of fertility and
3.1 Primate research on the uncanny valley.
Research has found that other primates also exhibit the uncanny valley effect. Steckenfinger
and Ghazanfar (2009) tested monkeys with unrealistic and realistic synthetic monkey faces,
as well as images of real monkey faces. Preferential looking time consistently showed the
classic uncanny valley effect: the primates spent more time looking at the real facial images
and the unrealistic synthetic faces, and avoided looking at realistic synthetic faces. The
researchers also found more robust results with dynamic face stimuli. Steckenfinger and
Ghazanfar suggested that realistic synthetic agents do not meet evolved standards of facial
aesthetics, and that computer generated avatars can appear to be anemic (thus unhealthy)
and unattractive; their study demonstrated that this aversion is also exhibited by primates,
hereby suggesting an evolutionary origin to the uncanny valley effect.
3.2 Pathogen avoidance, threat detection and ‘necrophobia’.
An evolutionary approach of the uncanny valley suggested is that it is part of our mechanism
for indicating health of others (Rozin & Fallon, 1987). The examples include disgust as
mechanism for pathogen avoidance, as well as evaluating human faces and bodies as an
indicator of fertility (MacDorman & Ishiguro, 2006). As an evolutionary outcome of evoking
disgust responses toward diseased-looking humans, when we see highly humanistic
synthetic agents, their defects would stand out as an indicator of disease, as well as the
possibility of contracting bacteria, viruses and other parasites; for example, leprosy evokes
the feeling of disgust, yet leaf spots do not have the same effect. Similarly, Moosa and Ud-
Dean (2010) also suggested that instead of specific pathogen avoidance, the uncanny valley
University of CopenhagenMA Cognition and Communication
Introduction to Cognitive Science Fall 2016 Exam
The Uncanny Valley Effect and its Implications on Robotics and A.I. Development Yum-Mei Simone Wong (stl350)
is a general mechanism of danger avoidance. They referred back to Mori (1970)’s example
of our avoidance of corpses and attributed it to our evolutionary roots of necrophobia’. A
human corpse elicits strong negative emotions as it suggested danger; with the addition of
movements (i.e. a ‘fresh’ corpse), this indicates an even greater degree of danger as it
signifies close proximity of a potential attacker. Nonetheless, studies have found cultural
differences on perception of corpses and dead bodies, and more importantly, the concept of
death (Field et al., 1997, as cited in Moosa and Ud-Dean, 2010.) This raises the question of
the extent to which we can attribute the uncanny valley to a generalised, evolved defense
3.3 Fertility, health and attractiveness.
As mentioned, attractiveness also serves as evolved indicators of fertility and health. Some
suggested that this automatic, stimulus-driven processing has a biological basis that may
contribute to the uncanny valley effect. Factors of attractiveness such as youth, skin quality,
and familiarity have been found to correlate with resistance to disease and parasites (Jones
et al., 2004); aversion to unattractive features that are often enhanced when viewing
humanistic androids may therefore serve an evolutionary purpose as an inherited perceptual
mechanism to aid mate selection. However, MacDorman and colleagues (2009) criticised
this viewpoint by stating out that Homo sapiens did not evolve with robots or cartoons; it is
therefore difficult to make the bold claim that our inherited aesthetic preferences, or
‘detectors’ of fitness may contribute to our perception of modern, artificial agents like androids
or virtual avatars.
3.4 Summary of findings and discussion.
In general, studies that attempted to explain the uncanny valley with an evolutionary
approach often focus on its function in detecting human-likeness for self-protection, as well
as perceiving human-likeness as an indication of fertility, attractiveness and health. Therefore
the discomfort and eeriness we experience when perceiving an human-like agent may be
related to our innate, evolved mechanisms in detecting threat and preferences for aesthetics,
which are stimulus-driven and automatic. Nonetheless, one should consider the following
limitations in explaining the uncanny valley effect with an evolutionary approach.
University of CopenhagenMA Cognition and Communication
Introduction to Cognitive Science Fall 2016 Exam
The Uncanny Valley Effect and its Implications on Robotics and A.I. Development Yum-Mei Simone Wong (stl350)
Firstly, most research in evolutionary psychology are based on the biological processes and
adaptive problems that took place during the Pleistocene as the environment of evolutionary
adaptedness (EEA). One should question whether any proposed evolved mechanisms is
representative enough to explain our aversion toward realistic androids, which only emerged
in very recent decades.
Moreover, a main presumption of most evolutionary explanations on the uncanny valley is
that our aversion stems from the expectation of actions from the humanlike robots – such as
diseases spreading, or danger. However, Gray and Wegner (2012) questioned the idea that
agency is the main factor during an uncanny experience; they found that whilst participants
attribute similar level of agency to a humanlike and mechanical robot, they attribute more
experience (ratings of statements such as ‘this computer can feel pain’) to the realistic
machine; this was also found to mediate their ratings of uncanniness, even in an experimental
condition where the computers have no actual physical appearances (Gray and Wegner,
2012). It is therefore probable that machines become unnerving to us not because we ascribe
to them agency, but mainly due to their potential capacity to feel and sense. Instead of limiting
to any innate, self-defense mechanisms, it is more likely that a higher level of cognitive
processing is involved.
Whilst Mori’s original hypothesis included the examples of corpses and prosthetics, they also
differ fundamentally from realistic androids, in a sense that robots suggest a much higher
level of realism and human-likeness as a fully functioning, moving object. Experimental
stimuli in the studies discussed, however, focused largely on visual realism, whereas factors
of voice and motion were not explored. Considering that they are distinct elements that
dissociate a generic realistic object (e.g. synthetic skin) from an anthropomorphic agent (e.g.
robots), research should focus more on exploring the influence of motion and voice on the
uncanny valley.
In general, considering the mentioned research studies on uncanny valley from an
evolutionary approach, it is perhaps overly simplistic to explain our aversion toward realistic
looking androids with an automatic mechanism that is innate and evolved.
University of CopenhagenMA Cognition and Communication
Introduction to Cognitive Science Fall 2016 Exam
The Uncanny Valley Effect and its Implications on Robotics and A.I. Development Yum-Mei Simone Wong (stl350)
4. Cognitive neuroscience approach on the uncanny valley
The uncanny valley hypothesis has also been explored extensively from a cognitive
neuroscience approach. Some suggested that an element of higher cognition is involved that
concerns prominently our explicit definitions of human mind (Gray and Wegner, 2012).
Instead of explaining the uncanny valley with a stimulus-driven mechanism that largely
emphasise on its evolutionary purposes, researchers have suggested that more complex,
late cognitive processing may be involved. This section explores the uncanny valley by
reviewing empirical research on facial brain networks, the effects of mismatched visual,
auditory and motion elements, as well as proposed theories related to cognitive dissonance
and terror management.
4.1 Facial brain network, human-likeness and realism.
The uncanny valley effect concerns our ability to relate to robots. Individuals have the
tendency to attribute human characteristics to both inanimate and moving objects - for
example, attributing emotional significance to motions of shapes, or to recognise faces in
non-living objects like cars and houses (Castellano et al., 2007). Our sensitivity to
anthropomorphism in turn activates different areas in the brain. Core regions of the face
processing network in our brain include the fusiform face area (FFA), responsible for face
recognition and processing of invariant aspects (e.g. identity); and the superior temporal
sulcus (STS) that processes variant aspects of faces like expressions.
Measurement of brain activity can give an indication of how human-like factors on robots
(such as facial features) are perceived by individuals. James and colleagues (2015) used
fMRI to look at how individuals process and respond to stimuli (avatar) that are of difference
levels of human-likeness and realism. Participants were shown images of human and animal
faces, as well as animated human and animal figures (cartoons). BOLD fMRI found a
sensitivity preference for real faces throughout the face network, and also beyond the core
regions; this preference by participants were specific only to human facial stimuli, but not
animal faces or cartoons; they also suggested that the brain responds differently to cartoon
human faces than to real human faces. Our distinct responses towards highly realistic
human-like agents is therefore something specific to realism of human faces, instead of
realism of any class of objects (James et al., 2015). Instead of any living organisms, we often
University of CopenhagenMA Cognition and Communication
Introduction to Cognitive Science Fall 2016 Exam
The Uncanny Valley Effect and its Implications on Robotics and A.I. Development Yum-Mei Simone Wong (stl350)
experience uncanny feelings towards realistic robots because they specifically remind us of
living humans.
4.2 Mismatch in realism of human-looking and less human-looking elements.
Studies have looked into the role of visual discrepancy between more and less realistic
elements on face stimuli. Seyama and Nagayama (2007) investigated the phenomenon by
measuring participants’ impressions on facial stimuli that was manipulated by morphing
artificial and real human faces (see Figure 2). Whilst they did not initially find the classic UV
effect, they discovered, by a modification to the experimental apparatus, that mismatch
between the degree of realism of stimulus head and eyes resulted in the uncanny valley. We
are highly sensitive to the eyes as they provide social cues and emotional information; as a
result we tend to rely on them for communication of intent (Donath, 2011, p.62). Failure to
recognise ‘life’ within eye gazes of unrealistic robots can therefore promote fear, as it also
signifies unpredictable, or incomprehensible behavior (Brenton et al., 2005).
Figure 2. Visual stimuli used in Seyama and Nagayama (2007)'s experiment. Morphed
images of doll and human faces differing in eye size resulted in the uncanny valley effect.
Furthermore, as mentioned in the evolutionary approaches to the uncanny valley, auditory
and motor elements are often neglected by research that focus on the role of early visual
processing. Mitchell and colleagues (2011) therefore explored the effect of auditory
characteristics on individuals’ attitudes toward human and robot stimuli. They proposed that
visual-auditory mismatch should correlate with uncanniness. In their study participants were
shown a series of videos with either a human or a robot figure, and either a matching or
mismatching voice (real or synthetic). When asked to rate on the humanness, eeriness and
warmth of each character, they found that videos with mismatched conditions (human figure
with synthetic voice and robot figure with human voice) were rated the highest in eeriness by
University of CopenhagenMA Cognition and Communication
Introduction to Cognitive Science Fall 2016 Exam
The Uncanny Valley Effect and its Implications on Robotics and A.I. Development Yum-Mei Simone Wong (stl350)
participants, meaning that a human voice heightened the eeriness of the robot, whilst a
synthetic voice heightened the eeriness of the human. Moreover, whilst robots with a
synthetic voice received the highest warmth ratings, human figures with synthetic voices were
rated the lowest. Their results indicated that incongruence in realism of face and voice may
be a critical element in relation to induced uncanniness. More importantly, their study also
suggested that humanness was accentuated by the low realism of the robot face and voices;
whilst the oddity of a synthetic voice overpowers the realism of a anthropomorphic facial
features, and resulted in perception of low human-likeness.
4.3 Cognitive dissonance.
MacDorman and colleagues (2009) discussed the possible explanations of uncanny valley
related to cognitive dissonance. As mentioned, it is unlikely that we could attribute our
preferences for robots to a purely evolved mechanism as they were not present during the
EEA, which most established evolutionary theories were temporally based on. Instead, our
understanding of robots or animated characters should be considered as largely socially
constructed (MacDorman et al., 2009). They suggested that the discomfort caused by
realistic, artificial human forms is not a result of how they look, but what they signify - which
is ‘a challenge to their maker’s uniqueness’. As these human-like robots give a vague sense
of identity that lie between boundaries of human and non-human, the perception of one
hereby causes uncanny experiences related to cognitive dissonance; they further suggested
that such cognitive processing very likely involves brain regions that engage in emotion-
biased, motivated reasoning, such as the ventromedial prefrontal and anterior cingulate
(Westen et al., 2006, as cited in MacDorman et al., 2009, p.5). This is also related to the
previously discussed study by Gray and Wegner (2012), which suggested that the capacity
for robots to experience (feel and sense) plays a critical role in the uncanny valley effect.
4.4 Terror management and the challenge to human uniqueness.
Furthermore, MacDorman explained the phenomenon with the help of terror management
research, and suggested that highly realistic robots elicit our innate fear of mortality (Solomon
et al., 1998, as cited in MacDorman et al., 2009, p.6). He argued that cultural worldviews
(such as family, nations, or God) provides a meaning and sense of permanence to our lives
by offering a ‘symbolic transcendence of death to those who live up to their standards’, hence
reducing the anxiety we may experience during awareness of our inevitable death. Unlike the
University of CopenhagenMA Cognition and Communication
Introduction to Cognitive Science Fall 2016 Exam
The Uncanny Valley Effect and its Implications on Robotics and A.I. Development Yum-Mei Simone Wong (stl350)
evolutionary explanation of threat avoidance that can also be found on primates, cultural
worldview and awareness of mortality are uniquely human. MacDorman suggested that as
presence of android robots or highly realistic animated characters challenge this sense of
human uniqueness, and undermines our personal identity, the uncanny feeling we
experience is therefore caused by us triggering a similar psychological defense as a
subliminal reminder of death. However this theory has been criticised for being overly
simplistic, as a reminder of death is necessary yet not sufficient to cause a sense of eeriness
(Misslehorn, 2009); nonetheless, it gives an insight to how our unique identity as humans can
potentially be triggered by robots and cause aversion or terror this should therefore be
carefully considered during the design of artificial intelligence.
4.5 Summary of findings and discussion.
In general, empirical research with a cognitive neuroscience approach have provided a
number of possible explanations to the uncanny valley – a higher level, cognitive processing
in the face brain networks is most likely involved, in which mismatch in voice or visual realism
(especially the eyes) can also contribute to an experience of cognitive dissonance. The
uniquely human understanding of agency and, more importantly, the capacity to feel or sense,
has also been suggested to trigger an uncanny and eerie feeling. Nonetheless, one should
consider the experimental setup, as well as the external validity of most cognitive
neuroscience research on the uncanny valley. Whilst dependent variables often vary
between studies (e.g. ratings of affinity, warmth, ‘human-likeness’), realism of robot face
stimuli also vary between a lot of studies; this can potentially result in a floor or ceiling effect
- for example due to low realism of the robot faces, stimuli that morph both kinds of faces
together would always be rated low on human-likeness. Moreover, most studies in the field
look at participants’ ratings toward facial stimuli, or video tapes of humans and robots. Yet
even when incorporating elements of audio and movement, there is still a fundamental
difference than viewing a robot in real life – which is the anticipation that they will respond or
interact with us. In fact, affordances of the stimuli are not reflected in most experimental setup.
It is therefore unknown whether the uncanny valley effect would be amplified or even
diminished if participants are faced with the stimuli in real life. In order to gain a more in-depth
understanding of the uncanny valley, stimuli used should be representative of how or what
actions we perform on them. Future research on the uncanny valley should therefore focus
on implementing physical stimuli and face-to-face interaction.
University of CopenhagenMA Cognition and Communication
Introduction to Cognitive Science Fall 2016 Exam
The Uncanny Valley Effect and its Implications on Robotics and A.I. Development Yum-Mei Simone Wong (stl350)
5. Discussion
This essay explored the uncanny valley by evaluating some of its theoretical explanations
from an evolutionary approach, as well as cognitive neuroscience. Mori (1970)’s uncanny
valley hypothesis demonstrates the sense of unease and discomfort when we look at
increasingly realistic agents. This section discusses implications of research findings, the
possibility of habituation in future attitudes towards humanoid androids, and additional
insights from marketing use of anthropomorphism. Concluding remarks of the discussion will
address general directions in future robotics and A.I. design.
5.1 Implications of current research findings.
The idea that realistic robots hinders our concept and awareness of death has been
suggested in both evolutionary psychology research (Moosa and Ud-Dean, 2010), and in
relations to terror management research by MacDorman. We are more uncomfortable with
the fact that robots can experience and sense, than the actions they may perform. Related
to MacDorman’s idea that robots challenge the maker’s uniqueness, this perhaps have the
most relevant implication on A.I. and robot designs for clinical uses, which machines will have
direct contact with patients that may have had traumatic experiences, or heightened
sensitivity. Considering the rate of development in current graphics and voice-related
technology, it is likely that in near future we may eventually be able to produce robots that
are completely indistinguishable from humans. However, it is also worth questioning whether
this is the best solution in creating ‘likable’ and ‘trustworthy’ robots for domestic or social
James and colleagues’ findings suggested that brain regions of the facial network are
activated similarly during perception of both real and synthetic human faces, but not during
perception of real and cartoon animal faces. We evaluate realism of human and animal
figures differently - meaning that the uncanny valley involves category representation of
specifically humans, but not any living organisms. This is crucial in a sense that this
specialised process is what allows us to expect from robots a higher form of cognition.
Implementing elements that diverge our attention away from seeing it as a human can likely
eliminate the uncanny feeling. For example, realistic animations of fairies and elves are often
not considered as ‘humans’ as first glance, albeit being highly anthropomorphic; similarly,
University of CopenhagenMA Cognition and Communication
Introduction to Cognitive Science Fall 2016 Exam
The Uncanny Valley Effect and its Implications on Robotics and A.I. Development Yum-Mei Simone Wong (stl350)
fictional characters with non-human features (animal ears, paws…) are often seen as likeable
but not uncanny. Similar to socially constructed representations of fable or fairy tale
characters, we may implement ‘robot’ features to enhance a visually appealing concept of
humanoid robots (e.g. exposing batteries, circuits or chips); thereby avoiding the uncanny
valley whilst focusing on development of realistic skin texture, movements and voice.
Mitchell and colleagues (2011) suggested that synthetic features tend to heighten sensitivity
to humanistic cues. Therefore one approach would be to deliberately design robots and other
appliances that do not resemble humans at all, yet incorporating humanistic features that
accentuates its liveliness. A present example of A.I. would be the personal assistant function
on iOS devices (Siri), which features a natural language interface that interacts in a human
voice. Siri is likable in a sense that its humanistic element is accentuated when applied to the
mechanical appearance of a smartphone, even though one might find the voice to be
unrealistic or ‘slightly off’ when the software is given an anthropomorphic appearance.
5.2 The effect of habituation and adaptation.
Confidence in robot trustworthiness is a crucial factor when considering its application human
life and welfare (Marthur and Reichling, 2009). The rapid development of robotics and A.I. in
recent years has allowed us to be more exposed, and more knowledgeable about the utility
of androids in every day lives. Whilst we can foresee this change in society, it is likely that
our attitudes toward robots as a socially constructed concept may also be habituated
gradually. Brenton and colleagues (2005) suggested that the uncanny valley is an emotional
reaction that is subject to change over time. Increased human-computer interactions and
positive media descriptions of robots (e.g. movies, computer games) may therefore alter our
responses to realistic avatars in the future. Apart from development of domestic robots and
A.I., this potential ‘shift’ in attitude should also be considered in future research of the
uncanny valley – for example, individual differences on level of acceptance towards robotics,
or other mediators such as cultural differences in exposure to A.I. technology.
5.3 Additional insights from anthropomorphism and uncanny valley in marketing.
The hypothetical graph by Mori (1970) also suggested that stimuli with slight anthropomorphic
features are rated as likeable. In fact, this effect has been widely utilised within marketing
research. Japan has a long history of implementing anthropomorphic elements to products
University of CopenhagenMA Cognition and Communication
Introduction to Cognitive Science Fall 2016 Exam
The Uncanny Valley Effect and its Implications on Robotics and A.I. Development Yum-Mei Simone Wong (stl350)
to induce likeability and familiarity for example implementing human voice guidance to
home electronic appliances (Fukuda, 2011).
Figure 3. Examples of Japanese anthropomorphic mascots. They are a popular marketing
tactic to represent regions, companies, or authorities. (Harrison, 2011)
In addition, ‘mascotisation’ is a popular marketing technique in Japan that ‘exploits’ Mori’s
hypothetical graph. These mascots are often anthropomorphic – enough to remind the
audience that it is an interactive cartoon figure, yet artificial enough to prevent the uncanny
valley (see figure 3). This can be seen from numerous examples beyond commercial
marketing but also in tourism, social services, and even Japan Self Defense Force (Walters,
2014). The sense of ‘playfulness’ and ‘cuteness’ in mascots helps to release tension between
corporations or authorities and the general public. Carefully considering what factors people
find likeable’ in mascot characters, we can apply the same logic to the development of
robotics that act as companions, and communicate with users in a sentient way.
5.4 Concluding remarks: general directions of future robots and A.I. development.
Mori (1970)’s hypothetical ‘uncanny valley’ curve suggests that our affinity rating begins to
drop when a stimulus is highly realistic, and become ‘likable’ again when they are completely
indistinguishable from humans. Whilst more empirical research is needed that consider the
representation of robots’ affordances, our attitudes toward realistic computer agents appear
to be subject to constant change. To conclude, and by combining insights from research
findings on the uncanny valley, I hereby propose a number of directions that may be
appraised during the development of A.I. and robotics for every day uses. 1) In general, it
should not be the primary aim to design robots that look exactly like humans – not only does
University of CopenhagenMA Cognition and Communication
Introduction to Cognitive Science Fall 2016 Exam
The Uncanny Valley Effect and its Implications on Robotics and A.I. Development Yum-Mei Simone Wong (stl350)
it require highly precise and advanced technology that is currently unattainable, it increases
the risk of triggering a challenge to our deep understanding of human uniqueness; 2)
development of voice-related technology but not visual human-likeness should be the main
focus of A.I. for everyday interaction; 3) whilst keeping realistic textures, face ratios and voice,
visually incorporating distinct ‘robot’ features will reinforce a likable representation of androids
and avoid the mental categorisation of humans; and 4) learning from the Japanese marketing
technique of ‘mascotisation’, creating robots with low level of human-likeness but with
adequate anthropomorphic features can likely enhance their likeability and perceived affinity.
University of CopenhagenMA Cognition and Communication
Introduction to Cognitive Science Fall 2016 Exam
The Uncanny Valley Effect and its Implications on Robotics and A.I. Development Yum-Mei Simone Wong (stl350)
Breazeal, C. (2003). Emotion and sociable humanoid robots. International Journal of
Human-Computer Studies, 59(1), 119-155.
Brenton, H., Gillies, M., Ballin, D., & Chatting, D. (2005, September). The uncanny valley:
does it exist. In Proceedings of conference of human computer interaction, workshop on human
animated character interaction.
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.
Castellano, G., Villalba, S. D., & Camurri, A. (2007, September). Recognising human
emotions from body movement and gesture dynamics. In International Conference on Affective
Computing and Intelligent Interaction (pp. 71-82). Springer Berlin Heidelberg.
Donath, C. (2012). Persuasion's empire: French imperial reformism, 1763-1801.
Fukuda, R. (2011). Gerontechnology for a super-aged society. In The silver market
phenomenon (pp. 79-89). Springer Berlin Heidelberg.
Geller, T. (2008). Overcoming the uncanny valley. IEEE computer graphics and
applications, 28(4), 11-17.
Gray, K., & Wegner, D. M. (2012). Feeling robots and human zombies: Mind perception and
the uncanny valley. Cognition, 125(1), 125-130.
Hanson, D. (2006, July). Exploring the aesthetic range for humanoid robots. In Proceedings
of the ICCS/CogSci-2006 long symposium: Toward social mechanisms of android science (pp. 39-
Harrison, E. (2011). Fuzz and Fur: Japan's Costumed Characters. [Book]. Retrieved from
Ho, C. C., & MacDorman, K. F. (2016). Measuring the Uncanny Valley Effect. International
Journal of Social Robotics, 1-11.
James, T. W., Potter, R. F., Lee, S., Kim, S., Stevenson, R. A., & Lang, A. (2015). How
Realistic Should Avatars Be?. Journal of Media Psychology
Jones, B. C., Little, A. C., Feinberg, D. R., Penton-Voak, I. S., Tiddeman, B. P., & Perrett, D.
I. (2004). The relationship between shape symmetry and perceived skin condition in male facial
attractiveness. Evolution and Human Behavior, 25(1), 24-30.
MacDorman, K. F. (2005, July). Androids as an experimental apparatus: Why is there an
uncanny valley and can we exploit it. In CogSci-2005 workshop: toward social mechanisms of
android science (pp. 106-118).
University of CopenhagenMA Cognition and Communication
Introduction to Cognitive Science Fall 2016 Exam
The Uncanny Valley Effect and its Implications on Robotics and A.I. Development Yum-Mei Simone Wong (stl350)
MacDorman, K. F., & Ishiguro, H. (2006). The uncanny advantage of using androids in
cognitive and social science research. Interaction Studies, 7(3), 297-3
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., & Entezari, S. O. (2015). Individual differences predict sensitivity to the
uncanny valley. Interaction Studies, 16(2), 141-172.
Mathur, M. B., & Reichling, D. B. (2009, March). An uncanny game of trust: social
trustworthiness of robots inferred from subtle anthropomorphic facial cues. In Proceedings of the 4th
ACM/IEEE international conference on Human robot interaction (pp. 313-314). ACM.
Mitchell, W. J., Szerszen, K. A., Lu, A. S., Schermerhorn, P. W., Scheutz, M., &
MacDorman, K. F. (2011). A mismatch in the human realism of face and voice produces an
uncanny valley. i-Perception, 2(1), 10-12.
Misselhorn, C. (2009). Empathy with inanimate objects and the uncanny valley. Minds and
Machines, 19(3), 345-359.
Moosa, M. M., & Ud-Dean, S. M. (2010). Danger avoidance: an evolutionary explanation of
the uncanny valley. Biological Theory, 5(1), 12-14.
Mori, S., & Leblond, C. P. (1970). Electron microscopic identification of three classes of
oligodendrocytes and a preliminary study of their proliferative activity in the corpus callosum of
young rats. Journal of Comparative Neurology, 139(1), 1-29.
Seyama, J. I., & Nagayama, R. S. (2007). The uncanny valley: Effect of realism on the
impression of artificial human faces. Presence, 16(4), 337-351.
Steckenfinger, S. A., & Ghazanfar, A. A. (2009). Monkey visual behavior falls into the
uncanny valley. Proceedings of the National Academy of Sciences, 106(43), 18362-18366.
Robins, B., Dautenhahn, K., Te Boekhorst, R., & Billard, A. (2005). Robotic assistants in
therapy and education of children with autism: can a small humanoid robot help encourage social
interaction skills?. Universal Access in the Information Society, 4(2), 105-120.
Rozin, P., & Fallon, A. E. (1987). A perspective on disgust. Psychological review, 94(1), 23.
Tinwell, A., Grimshaw, M., & Williams, A. (2011). The uncanny wall. International journal of
arts and technology, 4(3), 326-341.
Walters, D. (2014, February 4). Japanese mascotization, marketing, and imagined
communities [Blog post]. Retrieved from
... El tecnolecto de la robótica nos entrega un concepto específico que puede aplicarse al ámbito tecnológico y de la cibercultura: uncanny valley. Este concepto, acuñado por Masahiro Mori en 1970 (Wong, 2017) y título además de uno de los filmes que analizamos en este trabajo, refiere a la respuesta cognitiva humana frente a una réplica, cuya apariencia es tan próxima a la forma humana que, paradójicamente, en lugar de provocar simpatía o afecto, induce rechazo, repulsión y temor: una inquietante familiaridad frente a un androide, robot o avatar digital, por ejemplo. Ya Freud (1986) en "Lo ominoso" indica que un motivo clave de esta categoría es el doble, figura clave de las narrativas góticas. ...
Full-text available
This work analyzes the monster’s notion as an uncanny representation of the posthuman subjectivities in the digital era, in two Latin-American films of the recent years: the Peruvian director Juan Daniel F. Molero’s Videofilia: y otros síndromes virales (2015), and the Argentinian Federico Heller’s short film Uncanny Valley (2015). If, as it has been previously suggested, the monster is a cultural dispositive (Moraña, 2017) that shapes the anxieties of a time; and if it is also the readable body on which the cultural fears are inscripted (Cohen, 1996), in the digital era, the monster will appear as a fundamental sign to embody new fears and become itself the interface of horror.
Full-text available
En este trabajo se pretende realizar un estudio comparativo de lo que se entiende por robot desde el punto de vista jurídico en Europa y en Japón. Este debe ser el punto de partida de cualquier trabajo de investigación sobre las propuestas jurídicas que deben presentarse ante la aparición de los robots en el mundo laboral, económico, industrial, sanitario, cultural, etc. y los retos que tal aparición conlleva. Las dos zonas consideradas, Europa y Japón, lo son por tratarse de las más robotizadas del mundo, junto con China y Estados Unidos. Cualquier posible respuesta jurídica al reto que supone la entrada de los robots en la sociedad actual implica necesariamente un acuerdo en cuanto a lo que debe entenderse por robot, pues así podremos identificar claramente en qué situaciones o a qué objetos/sujetos podremos aplicar estas nuevas normas jurídicas. Hay que considerar que lo que se entiende por robot es diferente dependiendo de la disciplina desde la que nos queramos aproximar, por lo que la aproximación jurídica tendrá en cuenta, lógicamente, unos condicionantes distintos de otras disciplinas. Acompañando a la definición de robot propondremos también diversas clasificaciones que ya han ido surgiendo y comprobaremos su validez en el plano jurídico. De la comparación de los avances que se han ido produciendo en Europa y en Japón podremos extraer conclusiones útiles para analizar el impacto de los robots y su consideración en las distintas ramas del Derecho.
Full-text available
It can be creepy to notice that something human-looking is not real. But can sensitivity to this phenomenon, known as the uncanny valley, be predicted from superficially unrelated traits? Based on results from at least 489 participants, this study examines the relation between nine theoretically motivated trait indices and uncanny valley sensitivity, operationalized as increased eerie ratings and decreased warmth ratings for androids presented in videos. Animal Reminder Sensitivity, Neuroticism, its Anxiety facet, and Religious Fundamentalism significantly predicted uncanny valley sensitivity. In addition, Concern over Mistakes and Personal Distress significantly predicted android eerie ratings but not warmth. The structural equation model indicated that Religious Fundamentalism operates indirectly, through robot-related attitudes, to heighten uncanny valley sensitivity, while Animal Reminder Sensitivity increases eerie ratings directly. These results suggest that the uncanny valley phenomenon may operate through both sociocultural constructions and biological adaptations for threat avoidance, such as the fear and disgust systems. Trait indices that predict uncanny valley sensitivity warrant investigation by experimental methods to explicate the processes underlying the uncanny valley phenomenon.
Full-text available
Increased interaction with characters in games and online necessitates a better understanding of how different characteristics of these agents impact media users. This paper investigates a possible neurological underpinning for a common research finding; namely, that animated characters designed to be comparatively more human, more real, and more similar to the people they represent elicit more positive self-reported evaluations. The goal of this study was to examine the extent to which these results might be due to differential processing of character features in brain networks recruited for face recognition. There is some evidence that parts of the face network may be specifically tuned for real human faces. An experiment was conducted where subjects viewed photographs of faces of actual agents (humans and animals) or colored drawings of matched agents (cartoon humans and animals). Using fMRI to measure BOLD activation in the whole brain and specifically in the face network, we investigated the variation in patterns of activation with human and animal faces that were more or less real. The results were consistent with previous reports that the core regions of the face network were sensitive to the humanness of faces. However, our results extend previous work by showing that regions of the core and extended regions of the face network – and some regions outside the network – were sensitive to realism, but only realism of human faces.
Full-text available
This paper proposes that increasing technological sophistication in the creation of realism for human-like, virtual characters is matched by increasing technological discernment on the part of the viewer. One of the goals for achieving a realism that is believable for virtual characters is to overcome the Uncanny Valley where perceived strangeness or familiarity is rated against perceived human-likeness. Empirical evidence shows that the Uncanny can be applied to virtual characters, yet implies a more complex picture than the shape of a deep valley with a sharp gradient as depicted in Mori's original plot of the Uncanny Valley. Our results imply that: (1) perceived familiarity is dependent upon a wider range of variables other than appearance and behaviour and (2) for realistic, human-like characters, the Uncanny Valley is an impossible traverse, is not supported fully by empirical evidence and the concept is better replaced with the notion of an Uncanny Wall.
Full-text available
The development of robots that closely resemble human beings can contribute to cognitive research. An android provides an experimental apparatus that has the potential to be controlled more precisely than any human actor. However, preliminary results indicate that only very humanlike devices can elicit the broad range of responses that people typically direct toward each other. Conversely, to build androids capable of emulating human behavior, it is necessary to investigate social activity in detail and to develop models of the cognitive mechanisms that support this activity. Because of the reciprocal relationship between android development and the exploration of social mechanisms, it is necessary to establish the field of android science. Androids could be a key testing ground for social, cognitive, and neuroscientific theories as well as platform for their eventual unification. Nevertheless, subtle flaws in appearance and movement can be more apparent and eerie in very humanlike robots. This uncanny phenomenon may be symptomatic of entities that elicit our model of human other but do not measure up to it. If so, very humanlike robots may provide the best means of pinpointing what kinds of behavior are perceived as human, since deviations from human norms are more obvious in them than in more mechanical-looking robots. In pursuing this line of inquiry, it is essential to identify the mechanisms involved in evaluations of human likeness. One hypothesis is that, by playing on an innate fear of death, an uncanny robot elicits culturally-supported defense responses for coping with death’s inevitability. An experiment, which borrows from methods used in terror management research, was performed to test this hypothesis. [Thomson Reuters Essential Science Indicators: Fast Breaking Paper in Social Sciences, May 2008]
Full-text available
This article presents a longitudinal study with four children with autism, who were exposed to a humanoid robot over a period of several months. The longitudinal approach allowed the children time to explore the space of robot–human, as well as human–human interaction. Based on the video material documenting the interactions, a quantitative and qualitative analysis was conducted. The quantitative analysis showed an increase in duration of pre-defined behaviours towards the later trials. A qualitative analysis of the video data, observing the childrens activities in their interactional context, revealed further aspects of social interaction skills (imitation, turn-taking and role-switch) and communicative competence that the children showed. The results clearly demonstrate the need for, and benefits of, long-term studies in order to reveal the full potential of robots in the therapy and education of children with autism.
Conference Paper
Full-text available
The Uncanny Valley hypothesis has been widely used in the areas of computer graphics and human-robot interaction to motivate research and to explain the negative impressions that participants report after exposure to highly realistic characters or robots. Despite its frequent use, empirical proof for the hypothesis remains scarce. This study empirically tested two predictions of the hypothesis: a) highly realistic robots are liked less than real humans and b) the highly realistic robot's movement decreases its likeability. The results do not support these hypotheses and hence expose a considerable weakness in the Uncanny Valley hypothesis. Anthropomorphism and likeability may be multi-dimensional constructs that cannot be projected into a two-dimensional space. We speculate that the hypothesis' popularity may stem from the explanatory escape route it offers to the developers of characters and robots. In any case, the Uncanny Valley hypothesis should no longer be used to hold back the development of highly realistic androids.
Full-text available
As virtual humans approach photorealistic perfection, they risk making real humans uncomfortable. This intriguing phenomenon, known as the uncanny valley, is well known but not well understood. In an effort to demystify the causes of the uncanny valley, this paper proposes several perceptual, cognitive, and social mechanisms that have already helped address riddles like empathy, mate selection, threat avoidance, cognitive dissonance, and psychological defenses. In the four studies described herein, a computer generated human character’s facial proportions, skin texture, and level of detail were varied to examine their effect on perceived eeriness, human likeness, and attractiveness. In Study I, texture photorealism and polygon count increased human likeness. In Study II, texture photorealism heightened the accuracy of human judgments of ideal facial proportions. In Study III, atypical facial proportions were shown to be more disturbing on photorealistic faces than on other faces. In Study IV, a mismatch in the size and texture of the eyes and face was especially prone to make a character eerie. These results contest the depiction of the uncanny valley as a simple relation between comfort level and human likeness. This paper concludes by introducing a set of design principles for bridging the uncanny valley.
Various rapidly developed technologies are supporting the daily lives of younger users, but have left older adults behind. However, in a super-aged society, technologies could (and should) contribute to helping older adults maintain independent lives. The chapter introduces “gerontechnology,” which means the study of technology and aging, for the improvement of the daily functioning of the elderly. Through such means as behavior observation, eye tracking, and the subjective evaluation of participants in a daily situation, the difficulties older adults encounter in using technologies are clarified, and products aimed at the elderly are designed and produced based on the obtained data. More empirical studies are still required to verify the effectivity of products aimed at the elderly. At the same time, efforts should be made to help more people become familiar with gerontechnology.
The uncanny valley-the unnerving nature of humanlike robots-is an intriguing idea, but both its existence and its underlying cause are debated. We propose that humanlike robots are not only unnerving, but are so because their appearance prompts attributions of mind. In particular, we suggest that machines become unnerving when people ascribe to them experience (the capacity to feel and sense), rather than agency (the capacity to act and do). Experiment 1 examined whether a machine's humanlike appearance prompts both ascriptions of experience and feelings of unease. Experiment 2 tested whether a machine capable of experience remains unnerving, even without a humanlike appearance. Experiment 3 investigated whether the perceived lack of experience can also help explain the creepiness of unfeeling humans and philosophical zombies. These experiments demonstrate that feelings of uncanniness are tied to perceptions of experience, and also suggest that experience-but not agency-is seen as fundamental to humans, and fundamentally lacking in machines.
This paper focuses on the role of emotion and expressive behavior in regulating social interaction between humans and expressive anthropomorphic robots, either in communicative or teaching scenarios. We present the scientific basis underlying our humanoid robot's emotion models and expressive behavior, and then show how these scientific viewpoints have been adapted to the current implementation. Our robot is also able to recognize affective intent through tone of voice, the implementation of which is inspired by the scientific findings of the developmental psycholinguistics community. We first evaluate the robot's expressive displays in isolation. Next, we evaluate the robot's overall emotive behavior (i.e. the coordination of the affective recognition system, the emotion and motivation systems, and the expression system) as it socially engages nave human subjects face-to-face.