When we look at a photograph of a person holding another photograph, we implicitly judge the individual in the nested picture as having a lesser capacity to think and feel. A new study published in the journal Cognition reveals that this visual bias remains consistent regardless of whether the faces are upside down, covered by masks, or entirely generated by artificial intelligence. The research demonstrates that the structural presentation of a nested image heavily overrides the actual physical details of the human face itself.
Psychologists evaluate how we view the inner lives of others through a concept called mind perception. This theory proposes that people intuitively judge the mental capacities of various entities along two primary dimensions. The way we attribute these mental states dictates our moral judgments, our empathy, and our behavioral expectations in social environments.
The first dimension represents agency, which is the ability to think, plan, and act on one’s own volition. The second dimension represents experience, which describes the capacity to sense the surrounding environment and feel emotions. We naturally attribute high levels of agency and experience to living humans, placing them firmly at the top of the social hierarchy. We attribute significantly lower qualities of mind to animals, robots, and two-dimensional representations like photographs.
Previous research established a hierarchical decline in mind perception based on visual abstraction, a phenomenon dubbed the Medusa effect. Observers consistently attribute reduced mental capacity and realness to a person depicted in a picture of a picture compared to a person shown directly in a photograph. A single photograph represents a primary level of abstraction. A photo within a photo acts as a nested representation, separating the viewer by a secondary level of abstraction.
Kyushu University researcher Jing Han and a team of colleagues conducted the new study to investigate the underlying cognitive mechanics of this psychological phenomenon. The researchers wanted to know if manipulating how we process facial information would disrupt or erase the Medusa effect. They set out to test the bias using culturally adapted photographs, synthetic media, and physical obstructions that interrupt standard visual processing.
Recognizing human faces relies on two parallel visual evaluation pathways. Holistic processing involves recognizing the overall configuration of a face, intuitively taking in the arrangement of the parts as a unified whole. Feature processing relies on identifying specific individual components, such as the shape of the eyes or the curve of a mouth. The researchers designed a series of eight psychological experiments to systematically interfere with these visual pathways.
In the first experiment, the team recruited Japanese participants online and presented them with a new set of culturally adapted images featuring Asian models. The participants viewed images of a primary person holding a portrait of a secondary, nested person. The participants assigned numerical scores between zero and ten evaluating the subjects on agency, experience, and realness. The team found that the Medusa effect held true for the Asian models just as it had for Western populations in previous literature.
The subsequent test targeted holistic visual processing. The researchers took the images from the first experiment and flipped them vertically. Face recognition is disproportionately impaired when faces are presented upside down compared to other objects like houses or vehicles. While the inversion successfully lowered the broad mind attribution scores for all the individuals pictured, the participants still rated the people in the nested photographs lower than the primary subjects holding them.
Next, the team targeted feature processing by occluding specific parts of the face. In three successive experiments, they photographed their models wearing surgical face masks, dark sunglasses, or both accessories at the same time. Covering the lower face or the eyes prevents observers from utilizing necessary visual cues that typically signal emotion and inner mental states.
The physical accessories drastically reduced the general perception of mentality across the entire trial. Observers found it much harder to recognize agency and experience in subjects hiding behind masks and sunglasses. Yet the relative difference in mind perception remained intact. The nested subjects were always judged as possessing tangibly less of a mind than the directly photographed subjects.
The researchers also examined the impact of authenticity and artificial intelligence. The steady proliferation of synthetic media makes it incredibly easy to generate faces that are indistinguishable from authentic human photographs. The researchers used image generation software to create completely artificial scenes featuring synthetic people holding pictures of other synthetic people. The participants evaluated these images without being told they were generated by artificial intelligence.
The observers intuitively attributed less of a mind to the synthetic subjects than they did to real humans in previous trials. Even within these artificial generations, the psychological gap persisted. The artificial mind in the primary photo was rated higher than the artificial mind in the nested photo.
The final manipulation involved spatial scrambling. The team rearranged the internal facial features of their models, unnaturally scattering the eyes, noses, brows, and mouths. Scrambling removes the ability to interpret the stimulus as a coherent social agent entirely. Rating scores plummeted, establishing the lowest mind perception marks of the entire study. Despite evaluating violently distorted faces, observers still demonstrated the Medusa effect by rating the nested scrambled face lower than the primary scrambled face.
The results indicate that the Medusa effect ranks as an incredibly robust phenomenon that defies basic perceptual disruptions. It appears to operate largely independently of the physical or structural information present on a recognizable face. The researchers suggest that the effect might stem from a psychological concept known as Construal Level Theory. This theory posits that creating spatial, temporal, or hypothetical distance prompts more abstract mental associations in the human brain.
A nested photo signals psychological distance, making the individual seem existentially remote to the person evaluating the image. The Medusa effect could also reflect a deeper categorical sorting process. Observers might unconsciously treat an image embedded within another image more like a decorative object rather than a human agent.
The researchers noted a few specific limitations that require broader evaluation in future testing. The current photographs largely restricted the view to neutral faces and upper torsos, eliminating the influence of full body posture. Bodies contribute a vast amount of social information regarding emotion and identity that could skew visual evaluations.
Future research should test whether altering body language or introducing animated and robotic figures changes how nested abstraction impacts our social judgments. Evaluating individual differences in visual processing speed and accuracy might also help explain why some people are more susceptible to this visual bias than others.
The study, “Robust Medusa effect across facial manipulations,” was authored by Jing Han, Kyoshiro Sasaki, Fumiya Yonemitsu, Kaito Takashima, and Yuki Yamada.