Psyche and pixels: AI-generated images and their impact on stigma of mental illnesses in Germany

Artificial intelligence (AI) has rapidly evolved in recent years. Image generating chatbots enjoy great popularity. But they bear the risk of maintaining or enhancing preexisting stereotypes and stigma against severe mental illnesses. The objective of the study is to investigate how AI-generated imagery on psychiatric diseases and institutions are perceived compared to (non-psychiatric) medical ones. Using the chatbots Designer, DALL-E 3 and Midjourney V6 we created images associated with psychiatric and medical contexts (disease, institution, incident). Participants (N = 239, 75.31% women) rated images generated by one of the chatbots upon a self-assessment manikin (SAM) rating, an adjective and emotion rating. Furthermore, participants were asked to title the images. Images containing psychiatric scenes were perceived as more negative and more arousing than other medical scenes (all p < .001). Further, they were often rated as having less control over the situation. Fear and anger were more often elicited by psychiatric than medical scenes. Psychiatric images were rated as more threatening and scarier (all p ≤ .001). In sum, the perception of AI-generated images of psychiatric terms was aligned with pre-existing stigmatizing attitudes. This is the first study to systematically investigate affective perception and potential stigmatizing effects of AI-generated images in a psychiatric context. It raises important discussion about user information, usage guidelines and stricter regulations for generative AI.