Why Some People Feel Discomforted with AI-Generated Food Images: Understanding the Concerns

Imperfectly rendered artificial intelligence (AI)-generated food images can induce an unsettling feeling known as the uncanny valley, where realism intersects with discomfort. This phenomenon may stem from a condition called food neophobia, which drives aversion to novel foods and could be influencing our reaction to digital dishes.

A forthcoming study in the journal Appetite delves into why AI-generated images of food evoke unease among viewers. Researchers discovered that participants found these AI images less appealing and more eerie compared to both realistic photographs and cartoonish depictions. The perceived pleasantness initially declined before rising, following a cubic rather than a simple U-shaped curve.

The research team noted that this effect could be driven by food neophobia—aversity towards new foods due to safety concerns—and not by food disgust, which typically leads people to avoid potentially harmful substances like moldy or spoiled items. The findings have significant implications for clinical studies on eating disorders, marketing strategies utilizing AI-generated imagery, and advertising campaigns in the food industry.

Background: The potential influence of AI-generated food imagery on eating behaviors has been explored by researchers. While these images can be useful in clinical research, particularly with individuals suffering from obesity or restrictive eating habits who might perceive them differently, their reception is not universally positive. Many consumers find AI foods off-putting and prefer genuine photographs instead.

This phenomenon aligns with the uncanny valley theory first described by Masahiro Mori in 1970. He noted that as entities become increasingly realistic, our initial affinity towards them decreases until they appear unnatural or eerie. Although this concept has been studied extensively for humanoid robots and other objects, it remains largely unexplored when applied to food.

The current study aimed to determine if AI-generated images of food would trigger a similar effect, particularly when containing slight distortions from expected realism. Researchers also examined whether aversion towards these digital dishes could be linked to disgust or simply the unfamiliarity with novel foods.

Methodology: The participants comprised 95 German speakers (76 White) with an average age of 31.28 years; there were 27 females, 64 males, and four others identified as “other.” The study included images varying in realism—six highly realistic, six low-realistic, twenty intermediate-levels, and six depicting rotten food.

Results: Imperfect AI-generated images were found to be significantly less pleasant and more eerie than both realistic and unrealistic depictions. Unrealistically stylized or cartoonish AI images did not trigger the uncanny valley effect. Additionally, BMI appeared to influence reactions to these digital dishes; individuals with higher BMIs tended to view them more positively.

Conclusions: The study concluded that imperfect but realistic AI-generated food images fall into the uncanny valley, whereas highly stylized depictions do not. This discovery has practical implications for marketers and advertisers using AI imagery in the food industry. Furthermore, future research could investigate how individuals with obesity or eating disorders perceive these digital dishes and whether familiarity with such technology over time might mitigate its unsettling effects.