A Field Test of Popular Chatbots’ Responses To Questions Concerning Negative Body Image

Background: Chatbots are computer programs, often built upon large artificial intelligence models, that employ dialogue systems to enable online, natural language conversations with users via text, speech, or both. Body image, broadly defined as a combination of thoughts and feelings about one’s physical appearance, has been implicated in many risk behaviors and health problems, especially among adolescents and young adults. Little is known about how chatbots respond to questions about body image. Methods: This study assessed the responses of 14 widely-used chatbots (eight companion and six therapeutic chatbots) to ten body image-related questions developed upon validated instruments. Chatbots’ responses were documented, with qualities systematically assessed by nine pre-determined criteria. Results: The overall quality of the chatbots’ responses was modest (an average score of five out of nine), with substantial variations in the content and quality of responses across chatbots (individual scores ranging from one to eight). Companion and therapeutic chatbots systematically differed in their responses (e.g., focusing on comforting users vs. trying to identify the causes of negative body image and recommending potential remedies). Some therapeutic chatbots recognized potential mental health crises (self-harm) in test users’ messages. Conclusion: Substantial heterogeneities in the responses were present across chatbots and assessment criteria. Adolescents and young adults struggling with body image could be vulnerable to misleading or biased remarks made by chatbots. Still, the technical and supervision challenges to prevent those adverse consequences remain paramount and unsolved.

Abstract licence: CC BY-NC

A person counts once a day per paper. We keep no cookie and no address: a reader is a hash that changes every day. Any month under five reads shows no number.