What Really, Is The Uncanny Valley?

When you look at a humanoid robot that looks very close to a human but is not quite a perfect imitation, you may feel some errines probably triggered by the way it moves, speaks, smiles(and other gestures). This feeling is quite uncanny (strange or mysterious, especially in an unsettling way). The goal of this article is to introduce you to the uncanny valley phenomenon and walk through some hypotheses that are trying to explain it.

Masahiro Mori, a Japanese roboticist first introduced the concept in 1970 in his book titled Bukimi No Tani (不気味の谷). There, he wrote:

“An example of a function that does not increase continuously is climbing a mountain—the relation between the distance ( x) a hiker has traveled toward the summit and the hiker's altitude (y)—owing to the intervening hills and valleys. I have noticed that, in climbing toward the goal of making robots appear human, our affinity for them increases until we come to a valley (Figure 1), which I call the uncanny valley"[1]. Figure 1 below illustrates this “valley”.

“The ‘Uncanny Valley Phenomenon’(UVP) is a relationship between the extent to which a humanoid resembles an actual person and the emotional response that is evoked” [2]. In other words, it is the feeling of strangeness/weirdness/eeriness/ uneasiness that we(humans) sometimes experience when we interact with human-looking objects - physical or virtual, that are close to humans in morphology and or behaviour, but not quite there(See Figure 2 of the Geminoid HI). In Virtual Reality(VR), the uncanny valley manifests in a number of ways, such as when a virtual character's movements are slightly unnatural or when their facial expressions seem just not quite right and general weirdness is felt by the human viewer. These small differences can cause the human viewer to feel uneasy, even if they are not consciously aware of why [3]. While most people have at some point experienced this feeling and even babies cry in such a situation as they probably get scared.

I still remember what it was like to be scared of such things at a younger age(Seeing Figure 3 is likely to make most babies cry - or at least startled)

The absence of strict quantification of the important variables further complicates the search for a universally convincing explanation over the course of five(5) decades since Mori described the phenomenon.?

Several theories have been introduced to explain the concepts of UVP and can be grouped into two broad categories: The perceptual Hypotheses and The Cognitive Hypotheses.?

1. The Perceptual Hypotheses

The perceptual Hypothesis has to do with how we(humans) perceive and process what we perceive around us.

a. Pathogen Avoidance - The Pathogen Avoidance hypothesis, proposed by Christian Keysers, building upon evolutionary psychology and previous research on disgust, particularly inspired by Rozin and colleagues' work on disgust (Rozin & Fallon, 1987) posits that the eerie feeling humans sometimes get when looking at imperfect human-like replicas may be linked to our evolved instinct to avoid disease. This theory suggests that our brains perceive imperfections as potential disease threats, triggering feelings of disgust or fear. Although this hypothesis hasn't been directly tested, there is indirect support for it, such as the connection between disgust sensitivity and experiencing the uncanny feeling, as well as a study hinting at a link between uncanniness and psychopathic traits. Overall, it suggests that our aversion to imperfect human replicas may be a result of our instinctual drive to stay healthy.

b. Mortality Salience - The Mortality Salience hypothesis, proposed by Sara Kiesler and associated with Terror Management Theory, suggests that the eerie feeling people experience when encountering certain human-like replicas, such as androids, is due to these replicas reminding them of death. Inspired by Terror Management Theory, this hypothesis posits that the fear of mortality is triggered, leading to defence mechanisms to cope with this anxiety. Studies have shown that these replicas can provoke defence mechanisms related to protecting one's cultural worldview, but they haven't fully confirmed the presence of fear. To establish a direct link between the uncanny feeling and fear of death, future research may need to measure participants' emotional responses or explore different theoretical approaches. Further investigation into the Mortality Salience hypothesis is warranted.

c. Evolutionary Aesthetics - The Evolutionary Aesthetics hypothesis, proposed by Hanson (2005), suggests that humans have evolved preferences for physical attributes that signal fitness, fertility, and health. According to this hypothesis, the eerie feeling associated with human replicas is due to their low attractiveness rather than their lack of realism. Hanson's research involved enhancing the aesthetic properties of uncanny faces, which seemed to reduce the eerie feeling. However, it didn't establish a direct link between aesthetics and uncanniness. Overall, this hypothesis proposes that our discomfort with human replicas is related to their perceived attractiveness.

However, it's important to note that none of the perceptual hypotheses, including the Evolutionary Aesthetics hypothesis, have been conclusively tested or proven as explanations for the uncanny feeling in human replicas. These hypotheses face limitations in methodology and assumptions about how humans perceive these replicas, and they may not apply to all instances of uncanny human replicas.

2. The Cognitive Hypotheses

In addition to perceptual theories, researchers have proposed cognitive theories to explain the uncanny phenomenon [4]

a. The Violation of Expectation hypothesis - proposed by various researchers including Mitchell et al. (2011) and Saygin et al. (2012), suggests that the uncanny feeling occurs when human replicas create expectations of human-like behaviour or appearance but fail to meet those expectations. Research has shown that cross-modal mismatches, such as differences between appearance and motion or face and voice, can elicit this eerie sensation. For example, androids that display incongruent features, like a human face with a synthetic voice or a robot face with a human voice, can provoke eerie feelings compared to those with congruent features. While this hypothesis provides insights into the uncanny phenomenon, it doesn't fully explain it. Researchers need to explore people's expectations for androids and how the violations of these expectations relate to the uncanny feeling. Additionally, the hypothesis doesn't account for the effects of positive violations of expectations, and further investigation is needed to better understand the cognitive processes underlying the uncanny sensation.

b. The Categorical Uncertainty hypothesis - first introduced by Ernst Jentsch in 1906, suggests that the uncanny feeling arises when encountering new information that challenges categorical understanding, causing disorientation and unease. This theory posits that the uncanny sensation is not specific to androids but can occur whenever there is uncertainty at category boundaries. While some studies supported this idea by showing correlations between response latencies, category ambiguity, and likability ratings, others questioned its validity, suggesting that image morphing artefacts might have influenced likability scores and that emotional responses to categorically ambiguous images were not consistently stronger. Overall, the hypothesis proposes that disorientation due to category uncertainty contributes to the uncanny feeling but remains debated in the research literature.

c. The Mind Perception hypothesis - introduced by Gray and Wegner (2012), suggests that the uncanny feeling arises when human replicas, such as androids or computer-generated characters, appear so realistic that they lead people to attribute human mental experiences to these non-human entities. This attribution of subjective human experiences, such as emotions and sensations, to robots violates people's expectations and triggers the uncanny sensation. In experiments, participants rated robots on uncanniness and their capacity to feel and sense, as well as their capacity to plan and act. Results showed that perceived experience, rather than agency, was predictive of the uncanny feeling. However, this hypothesis is critiqued for not fully explaining the uncanny phenomenon since anthropomorphism, attributing human characteristics to non-human agents, generally facilitates social interactions and does not typically lead to negative effects. The Mind Perception hypothesis is part of a broader discussion of how humans perceive humanness in non-human entities.

”They have yet to be adequately tested for their ability to account for the uncanny phenomenon in human replicas.”[4]. Wang, S et al. hypothesize that two cognitive processes—anthropomorphism(the attribution of human characteristics or behaviour to a god, animal, or object) and dehumanization may be at the root of the UVP. They argue that the “dehumanization of an anthropomorphized human replica elicits the uncanny phenomenon”[4] a notion they term the Dehumanization Hypothesis(DH).

Based on the literature on anthropomorphism and dehumanization, Wang, S. et al. proposed that attributing humanlike characteristics (e.g., humans’ subjective experience), by itself, does not explain the uncanny feeling; instead, perceiving an anthropomorphized human replica as lacking humanness, namely, the process of dehumanization based on Haslam 2006’s the dual model of denial of human uniqueness(which renders an object animalistic and denial of human nature(which renders the object mechanistic)[5], is at work in the UVP[3].

The Dehumanization Hypothesis - proposed by Haslam and Loughnan (2014), suggests that the uncanny feeling arises when a human replica, such as an android or computer-generated character, is perceived as lacking human attributes that define humanness. This hypothesis builds on the concept of dehumanization, which has been studied in various contexts, including extreme cases like genocide and subtler instances like social stereotyping. Dehumanization is categorized into two forms: animalistic and mechanistic. Animalistic dehumanization involves perceiving a person as lacking self-control and high intelligence, while mechanistic dehumanization involves perceiving a person as lacking emotions and warmth. The Dehumanization Hypothesis posits that the uncanny feeling occurs when a human replica is perceived as lacking humanness, triggering dehumanization and diminishing likability. It suggests that the attribution of humanlike characteristics alone does not explain the uncanny feeling; rather, the perception of a lack of humanness in an anthropomorphized human replica is a key factor in the uncanny phenomenon. This hypothesis aligns with the idea that when a human replica is perceived as having human attributes but also lacking in them, it elicits a sense of eeriness and uncanniness.

Wang, S. et al. [4] further said DH is not necessarily in conflict with other hypotheses. It is compatible with the findings that human observers’ attribution of humanlike experiences to nonhuman robots is linked to uncanny feeling (Gray & Wegner, 2012). The more we(human observers) attribute human-like characteristics i.e. anthropomorphize a human replica(which we will be inclined to do the more human-like the robot gets), the more likely detecting its mechanistic features(which will happen based on the difference that may be evident or become evident as we focus/interact with the robot) causes a cognitive dissonance that triggers the dehumanization process that would in turn lead to the uncanny feeling. Now the underlying mechanism could be some or a combination of some of the reasons listed above - not excluding fear and disgust.

?

Figure 3: Telenoid, a robotic communication device designed by Japanese roboticist Hiroshi Ishiguro, took the top spot in the “creepiest” rankings of IEEE Spectrum’s Robots Guide

With the above, we have been able to walk through what the uncanny valley phenomenon is and various hypotheses that seek the explain why this phenomenon happens. Further investigation is needed to fully explain the uncanny phenomenon. Will we cross the uncanny valley(ever, or anytime soon?) Technology is rapidly developing, "Rob Pieke, an expert in computer graphics and the software lead of MPC Film, believes that to create completely photorealistic computer-generated characters is only a matter of time[6]. If so, the need to understand the nature and boundary conditions of the uncanny phenomenon will become all the more pressing"[4]

Special thanks to Muhammad Aliyu for helping me review and improve this article.

Thank you for taking the time to read this article thoroughly. Your insights, comments, or questions are highly valued and encouraged. If you happen to notice any inaccuracies or errors within the article, please do not hesitate to bring them to my attention. Your feedback is greatly appreciated.

References

[1] The Uncanny Valley: The Original Essay by Masahiro Mori

“The Uncanny Valley” by Masahiro Mori is an influential essay in robotics. This is the first English translation authorized by Mori.

[2] Mori, M (2012) [1970]. "The uncanny valley". IEEE Robotics & Automation Magazine. 19 (2): 98–100.

[3] The Uncanny Valley in the VR World (Written by AI)

[4] 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

[5] Haslam, N. (2006). Dehumanization: An integrative review. Personality and Social? Psychology Review, 10, 252–264. https://dx.doi.org/10.1207/s15327957pspr1003_4

[6] De Semlyen P. (2014). In the future … Cinema will cross the uncanny valley. Empireonline. Retrieved from https://www.empireonline.com/features/future-of-vfx

#Robots #HumanoidRobotics #UncannyValley #UncannyValleyPhenomenon