Jump to content

Motivation and emotion/Book/2025/Uncanny valley and emotion

From Wikiversity
Uncanny valley and emotion:
What is the uncanny valley phenomenon, what are its consequences, and what can be done about it?

Overview

[edit | edit source]
Figure 1. A humanoid robot head
Imagine this ...

It’s late at night, and you’re walking alone around your university campus. You pass the robotics lab and glance through the window. Inside there are a dozen humanoid robots that are half-finished and frozen in various poses. One robot [grammar?] in particular catches your attention. Its gaze feels ... human, too human somehow (see Figure 1). You stop, mesmerised by its face, until a flicker of doubt runs through your mind:

Did it just blink? Did it just move?!

In the dim moonlight, its expression seems to shift slightly, almost like a person were sitting there silently ... watching you.

A chill crawls up your back, your shoulders tense, and your heart pounds as fight-or-flight takes over.

Why do I feel this way? It's just a robot...

You wonder.

That sense of discomfort, the eeriness, is an example of the uncanny valley phenomenon. It is the uncomfortable feeling that humans experience when they come across objects or beings that are almost, but not quite, human. Understanding why this phenomenon occurs is important, especially due to the increase of humanoid robots, CGI characters, and AI[explain abbreviation] -generated content that are becoming common in humans' lives, and influence their emotional responses, trust, and social behaviour (Wang et al., 2015).

The uncanny valley goes beyond just robotics. In the context of horror games, for instance, CGI[explain abbreviation] characters that display slightly unnatural movements and facial features can have the ability to trigger players' amygdala, even in situations where there is no real or tangible threat (Ratajczyk, 2022). The eeriness is not the result of actual danger, but rather a subtle discordance between what the brain expects and what it perceives. Similarly, early CGI films such as The Polar Express or The Adventures of Tintin: The Secret of the Unicorn unnerved their audiences, not because the characters were overtly frightening, but because their facial expressions and movements were almost human, yet off at the same time (Vukadinović et al., 2023). These slight imperfections creates a sense of unease and discomfort, demonstrating that even a minimal movement that is unnatural from natural human motion can create strong emotional reactions. Beyond digital media, ordinary objects such as hyper-realistic dolls or mannequins can provoke a similar feeling of discomfort or eeriness (Grebot et al., 2022). Across media, technology, and daily life, the uncanny valley shows just how sensitive humans are to stimuli that closely resemble humans without fully replicating human features or behaviours, illustrating the brain’s sensitivity to human likeness and the importance of perceptual congruity (Wang et al., 2015; Mara et al., 2022).

Focus questions

  • What is the uncanny valley?
  • What are the psychological and evolutionary aspects of the uncanny valley effect on humans?
  • What are the social and behavioural consequences of the uncanny valley effect?
  • Where can the uncanny valley appear in (media and physical objects)?
  • How can the uncanny valley effect be reduced?

Understanding the uncanny valley

[edit | edit source]

The term “uncanny valley” was first introduced by Masahiro Mori in 1970 to describe the non-linear relationship between human-likeness and emotional acceptance in humanoid robots (Wang et al., 2015). Masahiro Mori observed that the more robots became human-like, the more people generally felt affinity and grew to feel comforted by them. However, this wouldn't be long. Beyond a certain line of realism, subtle imperfections in appearance or behaviour would cause humans to feel more fearful and uncomfortable, rather than the comfort they felt before, which lead to a dramatic drop in positive emotional responses between humans and robots (Berns & Ashok, 2024). In other words, there exists a “valley” in the relationship between human-likeness and comfort, which is defined by the level of unease when robots appear almost human.

Early observations

[edit | edit source]

Even before the development of advanced robotics, the uncanny valley phenomenon were found in highly realistic dolls, mannequins, and artificial CGI faces. Slight asymmetry in a doll’s facial expression or minor delays in mechanical movement were enough to evoke fight or flight, which indicates that humans are finely attuned to even the smallest deviations from an expected human appearance and motion (Wang et al., 2020). These early observations are important as to understand why near-human anomalies provoke strong emotional responses and laid the groundwork for contemporary research on human-robot interaction and CGI character design (Mathur & Reichling, 2016).

Figure 2. Predictive coding

Emotional responses

[edit | edit source]

Engineers and designers developing humanoid robots and prosthetics found that certain technical designs, while impressive its in craftsmanship and realism, often seems to make those who come across them feel uncomfortable. Dolls or artificial faces with hyper-realistic features frequently felt “wrong” and unnerving to viewers despite their lifelike intentions (Ferrey et al., 2015; Seyama & Nagayama, 2007). Essentially, when the human brain detects features that are nearly, but not quite human, it experiences difficulty categorising whether something is human or not. This causes cognitive dissonance, predictive processing errors, and visceral emotional responses to occur (Urgen et al., 2018).

Predictive processing theory, also referred to as predictive coding (see Figure 2), posits that the brain continuously generates expectations about any incoming sensory information, including cues from human appearance and movement. When an object that resembles a human being behaves slightly differently than what is its normal prediction, the brain detects it as a mismatch between expectation and perception. This mismatch triggers the amygdala and the feelings of discomfort or eeriness increase (Seyama & Nagayama, 2007; Ferrey et al., 2015; Urgen et al., 2018). This theory also aligns with other theories of the uncanny valley, including conflicting perceptual cues and violations of typical human norms, which occur when near-human anomalies challenge internal mental models of the standard human behaviour (Mathur & Reichling, 2016; Kätsyri et al., 2015).


Pop Quiz

1

Masahiro Mori coined the term “uncanny valley” first.

True
False

2

The uncanny valley refers to a point where objects that are almost human provoke:

Comfort
Eeriness or discomfort
Indifference

3

Early uncanny valley observations only occurred in robotics.

True
False

Psychological and evolutionary aspects of the uncanny valley

[edit | edit source]
Figure 3. Fans cosplaying as characters from Silent Hill

A primary psychological explanation for the uncanny valley involves cognitive dissonance. The human brain expects consistency between appearance and behaviour, particularly in entities that resemble humans. When a humanoid robot or CGI/ AI character appears human but moves unnaturally or its facial feature appear distorted, this expectation is violated, creating cognitive conflict (Seyama & Nagayama, 2007; Wiese & Weis, 2020). Predictive processing theory (Predictive coding) further clarifies that violations such as that produces a prediction error signal in the brain, which triggers discomfort, aversion, or unease (den Ouden et al., 2012; Urgen et al., 2018). Importantly, this reaction can occur even in the absence of actual real danger. For example, horror games like Silent Hill (see Figure 3) exploit subtle and unnatural movements in human-like enemies, provoking terror precisely because these characters defy expected human behaviour patterns (Ratajczyk, 2022).

Figure 4. Laffing Sal

Sensory mismatch

[edit | edit source]

Humans rely on a multitude of sensory cues to interpret emotions, intentions, and social cues. Near-human entities with slightly unnatural movement, gaze, or facial expression introduce inconsistencies that the brain struggles to accept (Sasaki et al., 2024). According to predictive processing theory, the brain generates expectations for motion, facial cues, and auditory signals; any contradiction of these expectations can create perceptual discomfort (Urgen et al., 2018). This is why even small glitches in CGI character movements or facial expressions in films may be perceived as unsettling, despite otherwise realistic rendering (Vukadinović et al., 2023).

Evolutionary perspectives

[edit | edit source]

From an evolutionary standpoint, near-human anomalies may activate innate threat-detection systems. Objects or entities that appear almost human, but with subtle deviations, may unconsciously signal injury, disease, or death, situations historically associated with survival risks (Grebot et al., 2022). Predictive processing amplifies this sensitivity, increasing the salience of near-human anomalies and explaining why the uncanny valley elicits such strong, cross-cultural reactions (Hoorn & Huang, 2024). These evolutionary mechanisms shows how discomfort is not arbitrary but is grounded in adaptive processes that have historically promoted safety and avoidance of potential hazards in humans.

Table 1.

Psychological and evolutionary mechanisms [Provide more detail]

Mechanism Description Example
Cognitive dissonance Discomfort due to mismatches of appearance and behaviour Robot with human-like face but jerky unnatural movements
Sensory mismatch/ Predictive processing Inconsistencies across visual, auditory, or motion cues generate predictive processing CGI character with strange blinking patterns
Evolutionary threat detection Subconscious avoidance of anomalies linked to disease or danger Hyper-realistic doll triggering fear


Quiz

1

Cognitive dissonance in the uncanny valley occurs when:

Appearance and behaviour are misaligned
An object matches human expectations perfectly
There is no human-like appearance

2

True or False: Prediction errors occur when sensory cues from a near-human object do not match brain expectations.

True
False

3

From an evolutionary standpoint, near-human anomalies may trigger:

Joy and curiosity
Threat detection and avoidance
Indifference

Social and behavioural consequences to the uncanny valley

[edit | edit source]

Human responses to the uncanny valley can range from mild uneasiness to intense fear or even fascination of it (Ratajczyk, 2022). Context, prior experience, familiarity, and the degree of realism all influence the emotional response. Many people actively seek out fear stimuli from novel or thrilling media, such as from horror films or horror games, which explains why creators frequently use uncanny valley in horror media. Films such as The Ring and Annabelle use hyper-realistic dolls or CGI human-like figures to provoke suspense and tension, while horror video games use humanoid monsters with subtle human features to enhance immersion and psychological engagement (Ratajczyk, 2022; Grebot et al., 2022).

Social and behavioural impacts

[edit | edit source]

The uncanny valley can also affect social interaction and trust. Humans may demonstrate reduced willingness to engage with humanoid robots in domestic, therapeutic, or workplace environments if the robot’s facial expressions or movements appear almost human but fall short of true realism (Mara et al., 2022; Berns & Ashok, 2024). For example, a robot designed to assist elderly individuals may fail to elicit cooperation or engagement of the elderly if it provokes discomfort in them. However, repeated exposure and familiarity can mitigate negative reactions over time, allowing individuals to get used to these robots and interact with them more comfortably (Yam et al., 2021). In media and gaming, designers balances the games realism carefully to raise engagement without triggering their aversions (Vukadinović et al., 2023).

Cultural and individual differences

[edit | edit source]

Responses to the uncanny valley are also shaped by cultural norms, individual personality traits, and prior experience. Some cultures exhibit higher tolerance for anthropomorphised robots or hyper-realistic dolls, while others are more sensitive to it. Individual differences, including personality traits, previous exposure to horror media, and personal experiences, further influences their reactions (Wang et al., 2015; Hoorn & Huang, 2024).


Pop Quiz

1

True or False: The uncanny valley can provoke fascination as well as fear.

True
False

2

Which of the following is true regarding social interaction with near-human objects?

It always increases trust
It has no social effect
It can decrease willingness to engage

3

Which factor influences reactions to near-human entities?

Cultural background
Personality traits
Past experiences
All of the above

Where the uncanny valley can appear

[edit | edit source]
Figure 5. Sophia (AI-robot)
Figure 6. Pepper (robot)

Humanoid robots are becoming prevalent in research labs, homes, and workplaces (Mara et al., 2022). Minor imperfections in movement, gaze, or facial expression can trigger discomfort. Robots such as Sophia (see Figure 5) (Berns & Ashok, 2024) or Pepper (see Figure 6) might provoke mixed reactions. While some may find them fascinating and engaging, others find them unnerving. Designers often use abstracted or deliberately non-human features to reduce the possibility of triggering human's fears while retaining enough human-like qualities to enable meaningful interactions (Hoorn & Huang, 2024).

CGI, AI, and animation

[edit | edit source]
Figure 7. AI-generated people

CGI, AI, and animated media are particularly prone to the uncanny valley effect. Even small inconsistencies in facial expression, movement, or gaze can make the characters look and feel “off”, creating discomfort in viewers (Vukadinović et al., 2023). With the rise of artificial intelligence, this effect has become even more noticeable as AI-generated faces (see Figure 7), voices, and gestures often appear almost human, but still looks off, which disrupts the brain’s expectations and triggers fear (Urgen et al., 2018). A clear example of the uncanny valley in CGI is The Polar Express, where subtle deviations from natural human movement made the characters feel eerie rather than lifelike. In CGI games and animations, designers often adjust the realism of non-player characters (NPCs), finding that slightly stylised or less human-like designs help maintain the immersion without triggering the effect (Ferrey et al., 2015).

Figure 8. Mannequin Heads

Mannequins and dolls

[edit | edit source]

Highly realistic dolls, mannequins (see Figure 8), wax figures, and statues can elicit unease due to subtle imperfections in facial expressions, gaze, or posture (Grebot et al., 2022). Designers may exaggerate certain features intentionally to reduce discomfort. Many horror media creators frequently capitalises on this effect, using dolls and mannequins as central elements to induce suspense and fear (Ratajczyk, 2022).

Figure 9. The Ring (film) logo

Horror media

[edit | edit source]

Horror films and games frequently exploit the uncanny valley to amplify fear and tension. Directors and designers understand that humans are instinctively sensitive to non-human entities that possess human-like features, and they leverage on this to create a suspenseful experience (Ratajczyk, 2022). Films like Annabelle and The Ring (see Figure 9) have dolls with hyper-realistic facial features and special effects make-up that trigger the effect. Horror games manipulate NPC gaze, movement, and timing to elicit fear into players (Ratajczyk, 2022; Grebot et al., 2022). Even social media users have explored the uncanny valley effect, with makeup or digital filters that exaggerated their facial proportions, provoking viewer's feelings of eeriness.


Pop Quiz

1

Which robot might elicit mixed reactions due to near-human features?

Cozmo
Sophia
Bender
C-3PO

2

Slight inconsistencies in motion of CGI characters can lead to:

Unease or discomfort
Increased empathy
Indifference

3

True or False: Highly realistic dolls or mannequins can trigger uneasiness?

True
False

4

Horror media often uses the uncanny valley to:

Make viewers laugh
Induce suspense and fear
Promote social bonding

Solutions to reducing the uncanny valley effect

[edit | edit source]

Stylisation, exaggeration, and deliberate non-human cues can help reduce prediction errors and discomfort. Focusing on enhancing synchrony in movement, gaze, and facial expression will allow the brain to generate accurate predictions, which will improve emotional comfort. Avoiding direct imitation of human features is also particularly important for robots who are meant to be around humans (Mara et al., 2022). In CGI, slightly cartoonish or stylised designs can prevent viewers' discomfort all the while still maintaining emotional engagement and narrative immersion (Vukadinović et al., 2023).

Figure 10. Cozmo (not turned on)

Applications in media and gaming

[edit | edit source]

Animators and game designers often adjust character designs to foster comfort rather than unease. Stylised characters in games such as Overwatch or The Legend of Zelda: Breath of the Wild avoid triggering the effect while maintaining human-like qualities (Vukadinović et al., 2023). Horror games, by contrast, may deliberately leverage almost-human designs to induce fear and anxiety, which demonstrates that the uncanny valley effect can be context-dependent (Ratajczyk, 2022).

Applications in robotics

[edit | edit source]

Human-robot interactions improve when designers account for the uncanny valley. Robots with abstract or stylised features, such as Cozmo (see Figure 10), are generally perceived as more comfortable and trustworthy (Yam et al., 2021). In contrast, highly human-like robots, such as Sophia, may still provoke discomfort, emphasising the necessity of careful design calibration (Berns & Ashok, 2024).


Quiz

1

True or False: Using slightly abstract or stylised features in robots helps reduce the uncanny valley effect.

True
False

2

In animation and video games, avoiding overly realistic characters helps:

Increase viewer discomfort
Promote empathy and engagement
Confuse the audience

3

One effective way to reduce the uncanny valley in robots is to:

Make them as human-like as possible
Ignore human reactions entirely
Use slightly abstract or stylised features

Conclusion

[edit | edit source]

The uncanny valley phenomenon shows the complex interactions between human perception, emotion, and near-human realism. Subtle mismatches in appearance or behaviour trigger discomfort through mechanisms including cognitive dissonance, sensory mismatch, predictive processing, and evolutionary threat sensitivity (Seyama & Nagayama, 2007; Urgen et al., 2018; Grebot et al., 2022). These reactions influence social interaction, trust, and engagement with near-human entities, ranging from AI/ robots to CGI characters.

Understanding the psychological and evolutionary foundations of these responses can help designers, animators, and roboticists create robots and characters that people find socially and emotionally comfortable. They can do this by utilising stylisation, deliberate non-human cues, and cautiously using realism, helping to reduce the phenomenon (Mara et al., 2022; Vukadinović et al., 2023). Though, in horror media and gaming, the uncanny valley can also be leveraged to create tension, suspense, and immersive experiences, showing that the phenomenon is context-dependent and not inherently a negative thing in terms of creativity (Ratajczyk, 2022; Ferrey et al., 2015)[grammar?].


Quiz

1

True or False: The uncanny valley is the uncomfortable feeling people experience when encountering objects that are almost, but not quite, human.

True
False

2

Who first introduced the term “uncanny valley”?

Sigmund Freud
Isaac Asimov
Masahiro Mori

3

Which of the following best explains why near-human objects induce discomfort?

The size of the object
Lack of colour or texture
Cognitive dissonance, sensory mismatch/ predictive processing, and evolutionary threat detection

4

True or False: The uncanny valley only appears in robots.

True
False

5

What is one effective way to reduce the uncanny valley effect in robots, AI, or CGI characters?

Make them perfectly human-like
Use stylised or deliberately non-human features
Ignore human reactions entirely

See also

[edit | edit source]

References

[edit | edit source]
Berns, K., & Ashok, A. (2024). “You scare me”: The effects of humanoid robot appearance, emotion, and interaction skills on uncanny valley phenomenon. Actuators, 13(10), 419–419. https://doi.org/10.3390/act13100419

Ferrey, A. E., Burleigh, T. J., & Fenske, M. J. (2015). Stimulus-category competition, inhibition, and affective devaluation: a novel account of the uncanny valley. Frontiers in Psychology, 6. https://doi.org/10.3389/fpsyg.2015.00249

Grebot, I. B. d. F., Cintra, P. H. P., de Lima, E. F. F., de Castro, M. V., & de Moraes, R., Jr (2022). Uncanny valley hypothesis and hierarchy of facial features in the human likeness continua: An eye-tracking approach. Psychology & Neuroscience, 15(1), 28–42. https://doi.org/10.1037/pne0000281

Hoorn, J. F., & Huang, I. S. (2024). The media inequality, uncanny mountain, and the singularity is far from near: Iwaa and Sophia robot versus a real human being. International Journal of Human-Computer Studies, 181, 103142–103142. https://doi.org/10.1016/j.ijhcs.2023.103142

Kätsyri, J., Förger, K., Mäkäräinen, M., & Takala, T. (2015). A review of empirical evidence on different uncanny valley hypotheses: Support for perceptual mismatch as one road to the valley of eeriness. Frontiers in Psychology, 6(1). https://doi.org/10.3389/fpsyg.2015.00390

Mara, M., Appel, M., & Gnambs, T. (2022). Human-like robots and the uncanny valley. Zeitschrift Für Psychologie, 230(1), 33–46. https://doi.org/10.1027/2151-2604/a000486

Mathur, M. B., & Reichling, D. B. (2016). Navigating a social world with robot partners: A quantitative cartography of the Uncanny Valley. Cognition, 146, 22–32. https://doi.org/10.1016/j.cognition.2015.09.008

Ratajczyk, D. (2022). Shape of the uncanny valley and emotional attitudes toward robots assessed by an analysis of youtube comments. International Journal of Social Robotics, 14, 1787–1803. https://doi.org/10.1007/s12369-022-00905-x

Sasaki, K., Yonemitsu, F., & Ariga, A. (2024). The uncanny valley phenomenon can be explained by categorization failure rather than categorization difficulty. Visual Cognition, 32(5), 388–399. https://doi.org/10.1080/13506285.2024.2448697

Seyama, J., & Nagayama, R. S. (2007). The Uncanny Valley: Effect of Realism on the Impression of Artificial Human Faces. PRESENCE: Teleoperators & Virtual Environments, 16(4), 337–351. https://doi-org.ezproxy.canberra.edu.au/10.1162/pres.16.4.337

Urgen, B. A., Kutas, M., & Saygin, A. P. (2018). Uncanny valley as a window into predictive processing in the social brain. Neuropsychologia, 114, 181–185. https://doi.org/10.1016/j.neuropsychologia.2018.04.027

Vukadinović, M., Ratković Njegovan, B., & Njegovan, M. (2023). On the ugliness and distortedness: The observers’ perception of the "uncanny valley" phenomenon in photorealistic computer animated faces. Studia Psychologica, 65(4), 364–377. https://doi.org/10.31577/sp.2023.04.886

Wang, S., Cheong, Y. F., Dilks, D. D., & Rochat, P. (2020). The uncanny valley phenomenon and the temporal dynamics of face animacy perception. Perception, 49(10), 030100662095261. https://doi.org/10.1177/0301006620952611

Wang, S., Lilienfeld, S. O., & Rochat, P. (2015). The uncanny valley: Existence and explanations. Review of General Psychology, 19(4), 393–407. https://doi.org/10.1037/gpr0000056

Wiese, E., & Weis, P. P. (2020). It matters to me if you are human - Examining categorical perception in human and nonhuman agents. International Journal of Human-Computer Studies, 133, 1–12. https://doi.org/10.1016/j.ijhcs.2019.08.002

Yam, K. C., Bigman, Y., & Gray, K. (2021). Reducing the uncanny valley by dehumanizing humanoid robots. Computers in Human Behavior, 125, 106945. https://doi.org/10.1016/j.chb.2021.106945

[edit | edit source]