Introduction

Imagine a life-sized mannequin—the kind you’d pass by during a casual shopping spree. Now, picture it with hair, teeth, and empty eyes that sink deep into their sockets. It looks almost human, but not quite. Creepy, right? You’re not alone in this feeling. There’s a reason people react this way, and it’s part of a phenomenon known as the Uncanny Valley Hypothesis. This theory describes how we can find human-like artifacts to be both fascinating and disturbing. Why does something that looks so close to humans make us so uncomfortable, and what does this say about the way our brains are wired? 

What is it?

First proposed in 1970 by Japanese roboticist Masahiro Mori, the Uncanny Valley Hypothesis (UVH) is a psychological phenomenon that describes the eerie feeling people may experience towards something that appears human-like, but isn’t. According to this theory, you might feel comfortable with a human-like object until it reaches a certain threshold—when the object becomes too lifelike. At this point, it encounters a “valley,” and its resemblance becomes “uncanny.” Then, your comfort transforms into disproportionate unease. This concept is analogous to riding a rollercoaster; just as a coaster climbs higher and higher, so would your preference for a human-like artifact. But once the coaster reaches the top—when the object reaches a certain peak of realism—it descends into what feels like a free fall. 

Take a prosthetic hand, for example. Early prosthetics, made up of glue, linen, and plaster, demonstrated an obvious distinction from human limbs. They’re bulky, clumsy, and strapped onto the user with leather [1]. Because they appeared unfamiliar, it was simple to distinguish where real flesh began and ended. With today’s technology, that difference becomes minute. Now, there is consideration for appearance as well as functionality for prosthetics. Beyond upgrading from plaster to more durable materials such as titanium, additions like skin-like coverings, motion-sensing technologies, and even intricate details like wrinkles transform a prosthetic hand. Today, it may strike a startling resemblance to that of a real hand in both appearance and motion. However, with these upgrades comes a risk of triggering the uncanny valley reaction. If we imagine shaking the prosthetic hand, its cold, firm grasp—despite its lifelike appearance—may feel uneasy. The contrast of the hand’s lifelike appearance and its artificial properties may evoke discomfort. It’s crucial for this effect to be considered when designing prosthetics to ensure both comfort and a positive experience for users [1]. 

The same can be said for other examples, like when we compare industrial robots to humanoid ones. Industrial robots are designed purely for functionality in manufacturing settings. In contrast, humanoid robots are intended for human-robot interactions in social settings such as customer service, where a humanlike robot would be preferred over the industrial robot. However, this preference backpedals as the robot appears more and more human—such as with facial expressions, realistic skin, or movement—which may negatively impact its interactions with real humans. So what causes this familiarity to reverse completely into aversion?

The Science Behind It

Mori's Paper

When he initially hypothesized the Uncanny Valley in 1970, Masahiro Mori developed models to explain the nature of the UVH response and how it shifts based on different circumstances [2]. Just as a mathematical function y = f(x) increases with variable x, Mori used predictive models to propose that as an object draws nearer to human likeness in size and appearance, human affinity for it increases until it becomes too realistic. This graph shifts when the element of movement is added to the equation. If an object is already creepy, adding movement will only amplify its peculiarity. This is because adding that extra layer of realism to an object intensifies its eeriness for people who already perceive it as unnatural. Based on this theory, Mori suggests alternative designs that steer clear of the Uncanny Valley effect, such as with prosthetic limbs. To illustrate this idea, he draws an analogy between prosthetics and glasses: “eyeglasses do not resemble real eyeballs, but one could say that their design has created a charming pair of new eyes” [2]. By suggesting that designers consider this mindset, he describes prosthetics not as replacements for a human limb, but simply a device or accessory, much like how glasses are considered. Mori also discusses why humans are wired to feel uncomfortable with objects in the Uncanny Valley. In his models, he proposes that the realization that something is not alive may serve as a reminder of death to people, which could explain their opposition towards these objects. Suggesting that the UVH response serves as an integral survival instinct within humans, he encourages further exploration of this phenomenon to better understand why it exists [2]. 

Brain Regions Involved

Although explaining the neuroscience behind a feeling may seem difficult, emerging research helps us understand the Uncanny Valley effect through brain responses. In one such study, researchers investigated neural activity of human participants when reacting to six categories of artificial agents. These categories were composed of humans without any physical impairments, humans with physical impairments, synthetic humans, and robots (android, mechanical, and humanlike). Full-body images of humans with and without physical impairments were sourced from photo databases, while images of synthetic humans were artificially generated based on people who had undergone extreme plastic surgery. The robot images were taken from a previous study that focused on robot appearances [3]. Participants in this study evaluated these categories on levels of likability, ranging from familiar to uncanny. During this task, scientists relied on a type of brain imaging called fMRI scans to investigate the neural patterns involved in the participants’ responses. Using these scans, they demonstrated linear and nonlinear relationships found with brain activities in certain areas, and the neural responses to human and artificial stimuli.

One area that was discovered to encode human likeness on a linear scale was the Temporoparietal Junction (TPJ), a brain region involved in social learning and belief attribution (the act of assigning a particular belief to someone based on a situation). In this linear relationship, as artificial agents became more human-like, activity in the TPJ gradually increased. This indicates its role in tracking human-like characteristics, which can help more complex brain regions in distinguishing between human and non-human characteristics. Activity patterns were not shown to be linear in other brain areas, such as the Ventromedial prefrontal cortex (VMPFC), a region associated with decision-making. Activity seen in the VMPFC reflected a combination of likeability and human-likeness, revealing decreased preference for agents that were highly humanlike but still artificial. This pattern aligns with the UVH, in which likeness increases with human-likeness, but sharply declines when human-likeness becomes too extreme. Other brain areas such as the dorsomedial prefrontal cortex (DMPFC) and fusiform gyrus (FFG) also reflected nonlinear responses. The DMPFC is involved in cognition and helps us understand or infer others’ intentions. This region displayed a shift in neural activity at the point when the UV effect typically occurred, indicating its involvement in identifying and emphasizing between humans and non-humans. This indicates the DMPFC’s role in setting the threshold for the uncanny valley response. The FFG plays a role in processing faces, and it was shown to increase its activity as artificial agents approached realism (but were still considered non-human). This suggests that the FFG helps discriminate between human and non-human objects, and could help reject agents beyond the threshold in the uncanny valley response. The amygdala, traditionally associated with fear responses and emotion, was also studied for its role in the UV response. Scans of the amygdala suggested that the amygdala signaled a greater discomfort or fear towards artificial agents that were too human-like. This reflects the amygdala’s role in inhibiting our interactions with objects perceived as threats due to their unnatural nature. Combined with behavioral choices, the activity in this brain region suggests more caution for people around nonhuman agents. 

Although this study sheds light on human responses to such objects in a neurobiological perspective, the controlled categories of objects exposed to participants may not represent the full spectrum encountered in the real world. Furthermore, the study used a neural network that primarily focused on specific brain areas, potentially overlooking others that may also play a role in the UVH response. Nevertheless, these results suggest that the specific brain regions studied in this experiment may display neural patterns and activity that can help explain the uncanny valley response in a neuroscientific perspective. The different brain activities occurring in these areas based on categories of objects explain why subtle changes in human likeness can cause disproportionate discomfort. In simpler terms, the study’s results support the uncanny valley hypothesis, suggesting that different brain regions show significant neural responses to characteristics of nonhuman agents—which affect how we process and react to non-human agents [3]. 

Perception 

In what ways does something appear uncanny to the human eye? As aforementioned with Mori’s proposal, the Uncanny Valley Hypothesis predicts that with movement, objects fall deeper into the “valley” of uncanniness. In a face, the lack of emotions may be uncomfortable for a viewer, particularly if an object mimics the appearance of a human in other aspects. However, the addition of facial cues, such as smiling or frowning, can still fail to capture how human emotion is naturally expressed. This makes them more unnatural, or shocking, to the human eye. The ability to perceive emotions is an essential survival skill that allows humans to detect danger, and emotions are natural human traits that connect people on a social level. Emotion-recognition is built into the human brain and can even be measured with methods like EEGs to explain how humans can distinguish between different facial cues [4]. 

Although this may explain the biological origins of the Uncanny Valley Hypothesis, another perspective suggests that it might also be provoked by cognitive familiarity. Humans often rely on recognizing faces by identifying them based on the relative positionings of facial features, like eyes, nose, and lips. Prototypical faces are more preferred because they tend to be associated with beauty and symmetry—often seen as indicators of health and genetic fitness. As a result, these faces are more universally familiar and easier to process for the brain. This suggests that we perceive familiar things differently compared to unfamiliar ones, which may play a role in the UVH response. One study tasked human participants to assess 3D computer renderings of realistic objects as attractive, average, or distorted. Researchers noticed that individuals who received training in these tasks prior to the experiment had a faster response to distorted or uncanny-looking renderings than novices, as they’d been shown similar 3D computer renderings and were already familiar with them [5]. Results showed that trained participants rated distorted renderings as more abnormal and less attractive compared to inexperienced participants Based on these results, researchers suggest that perceiving uncanniness in realistic objects is not purely a biological response, but also based on expertise in recognizing deviations from an average appearance. However, this conclusion does not imply that mere familiarity leads to these results, indicating that perhaps the uncanniness of an object is enhanced if we are experts in recognizing them. Thus, more abnormal faces may elicit discomfort based on our unfamiliarity with them, and this effect grows when we train to look for these abnormalities. Furthermore, the presence of facial expressions may intensify our discomfort, suggesting that our cognitive familiarity with both faces and emotions may contribute to the uncanny valley effect.  

Consider six basic emotions: anger, disgust, fear, happiness, sadness and surprise. These emotions all serve as adaptive functions that allow humans to react in a self-protecting manner when in distress. But these emotions are not equal; happiness, surprise, and sadness play a larger role in social settings, while anger, fear, and disgust are more significant in situations that threaten harm. When objects display emotions that are more survival-related, the Uncanny Valley effect increases compared to when objects express non-survival related emotions such as happiness [6]. Researchers applied this idea when studying realistically animated video game characters that had limited emotions on their faces. Because these moving characters lacked emotional depth with their limited facial expressions, they failed to replicate human emotions on the realistic characters, triggering the UVH response. And even if an object displays emotions, any deviation from authentic human ones can cause the same effect [6]. This response is more pronounced for specific emotions that cause distress, as the lack of emotional depth can inhibit effective communication of facial expressions, intensifying our unease around these objects. 

Applications of UV

Robots

With the continued research for the UVH, questions also emerge about how this phenomenon can be applied (or avoided) in real life applications. When the UVH enters a conversation, inevitably it will trigger mental images such as faces, realistic objects, and robots. Robots in particular have many applications for the UVH, especially with the growing popularity of human-robot interactions. In everyday life, it’s becoming more common to sit down at a restaurant and be served by a robot with a smiley face on its screen, or find one roaming around in Lowe’s interacting with customers. In such contexts, it is essential for these social exchanges to feel natural and seamless to maximize cooperation between humans and robots. In ordinary social exchanges, the first thing you may notice upon meeting someone new is their appearance; this observation might reflect insight on their ethnic background, interests, or age. With exchanges between social robots and humans, the UVH response should be considered in research to ensure a comfortable, natural, and personable environment. Such research may include the psychological closeness you might feel to a same-sex robot compared to one of the opposite gender, or vocal features that influence how robots are judged. One study investigated both, suggesting that humans favor social robots with realistic speech cues over synthetic ones, and felt closer to robots similar to them in gender [7]. This opens further pathways to consider how social robots can be used to emotionally connect with humans, which may become more prevalent in the future with the rising trend of loneliness and depression among other mental conditions seen in countries like the US [8]. 

From a healthcare perspective, touchscreen robots must foster trust with patients when delivering medication or assisting with patient care. Touchscreen robots can come in many different forms—while some displays have simplified or no faces at all, others boast their faces with moving parts. Studies have investigated robots with face-displays to understand how they are perceived by human participants, and propose that robots with more humanlike displays are perceived to be more amiable and sociable compared to less humanlike ones [9]. Robots perceived as less friendly may provoke feelings of unease within patients. In this case, the Uncanny Valley Hypothesis can be applied to suggest that a more realistic robot is preferred to strike comfort within patients, as long as patients aren’t confused about whether it is truly human or not. This leaves researchers and designers with the challenge of finding that sweet spot in which the robot is both realistic and personable, yet not lifelike enough to cause distress. With growing consideration for designs such as constructing 3D faces on robots, understanding the Uncanny Valley Hypothesis can guide technological advancements that fuel appropriate robot designs to maximize interactions with patients. However, it is important to acknowledge that as a novel application in the healthcare field, there are still concerns about the reliability of robots, particularly when dealing with more personal tasks traditionally handled by a human provider. While the UVH perspective would aim to match the compatibility of robots with humans on a social level, it would not encompass or correlate with their functionality. While we can improve robot designs for smoother communication with humans, there is still much to be done to ensure robots are safe and fully functional for use in the healthcare field [9]. 

Online/Social Media

Beyond healthcare, the UVH response is also being explored in the digital realm of communication. One example of this is the growing popularity of chatbots as a communication channel on online platforms. Although chatbots don’t mimic humans in appearance, they may elicit discomfort due to their ability to stimulate conversations and act like humans, triggering the uncanny valley effect. Designed for human interaction, chatbots can communicate with users to increase engagement on a platform. However, in order to achieve this, chatbots must effectively snag the user’s attention, which may be difficult as they’re often found idling on the bottom right corner of the screen. Studies have applied the Uncanny Valley Hypothesis in this context to suggest that when chatbots demonstrate positive emotions, they are more quickly noticed than if they demonstrate negative ones [10]. Online platforms can apply this understanding to maximize chatbot-human interactions while still being mindful to not overload users with emotional stimuli. The study of this effect on users may not be consistent, however, as different exposures to chatbot interactions might cause some users to become more habituated to the emotional aspect than others [10]. 

The UVH can also influence decisions like the acceptance of an online “friendship” or a “follow” on social media. For example, one study proposed that users may be more likely to reject female friendship requests from hyper-realistic avatar profiles (realistic, animated graphics) than cartoonish ones [11]. This implication may extend to other conversations, such as how people may impose judgements not just on an avatar profile, but on the user represented by that avatar [11]. Further investigation may also expand this discussion to understand the UVH in this context beyond the female population, and in different circumstances. When tracking the activity on social media platforms of virtual influencers, another study supports the idea that objects that fall into the Uncanny Valley receive more negative engagement than their less-realistic counterparts [12]. Lu do Magalu, Barbie, or the computer-generated counterparts to Aespa’s girl group members are examples of virtual influencers who have risen to popularity in many aspects of social media. While Lu do Magalu approaches a close likeness to a human, with powerful animated videos and images imitating that of a human influencer, Barbie is less realistic, with content primarily featuring unmoving, doll-like animation. Virtual influencing has become a powerful tool in the influencer world, as concerns for mental well-being and boundaries that come with the real world are not applicable to virtual influencers. Although they can be successful, the UVH must be taken into consideration to avoid negative reactions from audiences [12]. 

Popular Culture/Films

In current trends, the Uncanny Valley Hypothesis has been commonly seen on platforms like TikTok. As early as March of 2024, TikTok users have been posting makeup videos in which they imitate a doll-like or robotic appearance in an effort to exploit effects of the UVH reaction. In a tongue twister, these users strive for the irony where humans imitate non-humans that look human. But by also exploiting the Uncanny Valley, they draw in audiences enthralled by the unsettling nature of this content. By using these videos, users have been able to go viral, sparking widespread interest in the phenomenon. 

Popular films often mentioned in regards to this phenomenon include Polar Express (2004), Coraline (2009), and Cats (2019). Although some films, such as Polar Express, don’t intend for the realistic design of their characters to alienate viewers, others like Coraline use intentional visuals to deepen the unsettling atmosphere the audience is meant to feel. In the movie adaptation of Cats, characters are displayed in a mix of both animal and human form. As a high-budget film production that failed to draw in audiences, Cats has been criticized for its “creepy CGI” and “digital fur technology” [13]. The combination of their hyper-realistic fur, ears, and tails with human faces projects ambiguity to the audience, making it difficult for them to empathize and relate to the characters. It’s necessary for the UVH response to be taken into account when developing films that don’t call for that uneasy atmosphere. Otherwise they risk criticism from audiences. The same can be said for other aspects of popular culture, like video games and artificial intelligence (think Meta AI chatbots). This demonstrates both the relevance and commonness of the Uncanny Valley effect in everyday life. As the phenomenon grows in consideration within the technological world, viewers and audiences are gaining interest for its application in the media and entertainment industry. Whether it be technology, education, or the entertainment industry, there is no world that the Uncanny Valley Hypothesis has left untouched. 

Conclusion

In all of these applications and more, the UVH is becoming more prevalent in explaining the fine line between our perception of authenticity and imitation. Prosthetics, robots, and online spaces are only starting points that pave the way for innovations we can only imagine now, but may introduce essential tools for patient care, education, and facial animation in the media world. Whether for the enrichment of entertainment or to facilitate smoother interactions with humans and robots, the Uncanny Valley Hypothesis stands as a careful consideration for designers and researchers alike. By understanding the effect of the Uncanny Valley on a psychological, neurological, and social level, this phenomenon can help guide research in not only emerging fields, but refine established ones that play critical roles in our everyday lives. Just as Mori suggested that awareness of the UVH reaction could improve product design, comprehending its relevance in the context of today’s technological landscape can become a powerful tool in guiding intuitive developments that are both impactful and approachable. 

References

  1. U.S. National Library of Medicine. (n.d.). Prosthetics through the ages | NIH MedlinePlus Magazine. MedlinePlus. https://magazine.medlineplus.gov/article/prosthetics-through-the-ages 
  2. M. Mori, K. F. MacDorman and N. Kageki, "The Uncanny Valley [From the Field]," in IEEE Robotics & Automation Magazine, vol. 19, no. 2, pp. 98-100, June 2012, doi: 10.1109/MRA.2012.2192811.
  3. Pütten, A. M. R. der, Krämer, N. C., Maderwald, S., Brand, M., & Grabenhorst, F. (2019, August 14). Neural mechanisms for accepting and rejecting artificial social partners in the Uncanny Valley. Journal of Neuroscience. https://www.jneurosci.org/content/39/33/6555/tab-article-info 
  4. Mehmood, R. M., Du, R., & Lee, H. J. (n.d.). Optimal feature selection and deep learning ensembles method for emotion recognition from human brain EEG sensors | IEEE Journals & Magazine | IEEE Xplore. Optimal Feature Selection and Deep Learning Ensembles Method for Emotion Recognition From Human Brain EEG Sensors. https://ieeexplore.ieee.org/abstract/document/7997991/ 
  5. Diel, A., & Lewis, M. (2022, September 1). The deviation-from-familiarity effect: Expertise increases uncanniness of deviating exemplars. PLOS ONE. https://journals.plos.org/plosone/article?id=10.1371%2Fjournal.pone.0273861 
  6. Tinwell, A., Grimshaw, M., Nabi, D. A., & Williams, A. (2011). Facial expression of emotion and perception of the Uncanny Valley in virtual characters. Computers in Human Behavior, 27(2), 741–749. https://doi.org/10.1016/j.chb.2010.10.018
  7. Eyssel, F., Kuchenbrandt, D., Hegel, F., & de Ruiter, L. (2012). Activating elicited agent knowledge: How robot and user features shape the perception of social robots | IEEE conference publication | IEEE Xplore. IEEE Xplore. https://ieeexplore.ieee.org/document/6343858/ 
  8. New survey shows increasing loneliness, including on the job. Psychiatry.org - New Survey Shows Increasing Loneliness, Including on the Job. (2020, February 10). https://www.psychiatry.org/news-room/apa-blogs/new-survey-shows-increasing-loneliness-on-the-job 
  9. Broadbent, E., Kumar, V., Li, X., 3rd, J. S., Stafford, R. Q., MacDonald, B. A., & Wegner, D. M. (2013, August 28). Robots with display screens: A robot with a more humanlike face display is perceived to have more mind and a better personality. PLOS ONE. https://journals.plos.org/plosone/article?id=10.1371%2Fjournal.pone.0072589 
  10. Bortko, K., Fornalczyk, K., Jankowski, J., Sulikowski, P., & Dziedziak, K. (2023, July 27). Impact of changes in Chatbot’s facial expressions on user attention and reaction time. PLOS ONE. https://journals.plos.org/plosone/article?id=10.1371%2Fjournal.pone.0288122 
  11. Arsenyan, J., & Mirowska, A. (2021, July 21). Almost human? A Comparative Case Study On the Social Media Presence of Virtual Influencers. International Journal of Human-Computer Studies. https://www.sciencedirect.com/science/article/pii/S1071581921001129#sec0018 
  12. Cantrell, Kate, and Lesley Hawkes. “Empathy, Anthropomorphism, and the Uncanny Valley Effect: Why Audiences Strayed Away from the Film Adaptation of Cats.” Wiley Online Library, June 17, 2021. https://onlinelibrary.wiley.com/doi/abs/10.1111/ecog.03440/. 
  13. Shin, M., Song, S. W., & Chock, T. M. (2019). Uncanny Valley Effects on Friendship Decisions in Virtual Social Networking Service. Cyberpsychology, behavior and social networking, 22(11), 700–705. https://doi.org/10.1089/cyber.2019.0122