Exploring Bias in Text-To-Image Models: From Body Representation of Teenage Students to Perspectives for the Aging Society

Starting from the results of a study on body shaming conducted between Italy and Romania in ay 2020/21, this investigation explores the presence of bias in images produced by generative Text-To-Image (TTI) models. Specifically, it proposes an analysis and evaluation of the representation of adolescent students’ bodies in the outputs of two different TTI systems, highlighting social, cultural, and technological factors. The findings reveal significant biases in the representations of students, with notable differences between the platforms examined, and underline the need for targeted educational interventions to counter distorted perceptions of the body. This research provides a reflective contribution on the impact of TTIs on society and on the representation of body image that extends to different social and demographic categories, among these, the elderly, who are often victims or passive users of these technologies.