AI-generated food images that fall short of perfection can evoke feelings of unease, a phenomenon akin to the “uncanny valley,” where increasing realism meets discomfort. This study investigates how food neophobia might be behind our aversion to these digital dishes.
A forthcoming publication in the journal Appetite explores why AI-generated food imagery triggers uncanniness. The researchers found that viewers find AI images less pleasant and more eerie than both realistic and unrealistic (cartoonish or abstract) images. Moreover, perceptions of pleasantness and affinity initially decreased before increasing in a cubic rather than a simple U-shaped curve.
Food neophobia, not food disgust, seems to drive these feelings. Food neophobia is the fear of new foods due to safety concerns. The researchers noted that while AI-generated food imagery might be useful in clinical studies on eating disorders—such as those involving individuals with obesity or restrictive eating habits—it has mixed reception among consumers.
AI food imagery can appear off-putting, strange, or wrong. Viewers often prefer images of real food over artificially generated ones. This effect was first described by Masahiro Mori in his seminal 1970 essay on the uncanny valley phenomenon. As an entity becomes more realistic-looking, initial affinity increases before eventually turning to unease.
While the uncanny valley has been studied for robots and other entities but not extensively for food, this study sought to determine if AI-generated food images would elicit a similar effect. The research team explored whether slight distortions or deviations from expected realism in AI-generated foods could lead viewers into an uncanny response.
Unlike moldy, contaminated, or spoiled foods that trigger disgust and avoidance of harmful substances, the study found that food neophobia played a key role in perceptions of AI-generated images. This suggests that humans may have evolved to discriminate subtle differences in food appearance for survival purposes but also might misinterpret these differences as threats.
The research team tested whether other mechanisms, such as violations of internalized dietary preferences or savoriness assessments, influenced responses to AI-generated food. However, they found no significant correlation between these factors and the uncanny valley effect related to AI images.
Participants in the study comprised 95 German speakers (including 76 who identified as White) with an average age of 31.28 years; 27 were female, while 64 were male, and four identified as other. The findings indicated a slight positive correlation between realism and uncanniness but also revealed that unrealistically stylized AI food was perceived less uncanny.
Statistical analysis showed that participants found imperfect AI-generated images to be significantly less pleasant and more eerie than both realistic and unrealistic versions. However, the researchers noted that BMI appeared to influence reactions to AI-generated food imagery; individuals with higher BMIs may view these images more positively over time as familiarity increases.
The study’s conclusions suggest practical implications for using AI-generated food in marketing, advertising, and clinical research. Future studies could explore how different cultural backgrounds and increased exposure to such imagery might reduce the uncanny valley effect among diverse populations.
