Avatars and stereotypes. Algorithmic biases of the female smile generated by artificial intelligence

Abstract

Artificial intelligence dominates our news. As intrigue, given its novelty; as technological development and as creation. Given this scenario, from the perspective of gender and the field of study of visual culture, we analyze how female laughter is represented in generative AI images. A research objective whose starting hypotheses seek to verify whether stereotypes of classical painting, gender biases and issues such as madness or eccentricity are perpetuated in contemporary representations. To do this, as a methodology, a content analysis is carried out with categories on image, femininity and laughter on thirty images selected objectively. With them, it is also possible to debate the process of artificial creation of the Shutterstock image bank that adheres to the creator's ethics and allows warning of harmful images. In this sample, a sexualized trace of the female smile is visualized, symbolized through model women, with prominent necklines, dressed in leather or heels. Some representations that appear superior to the male gender, at which they laugh. On the other hand, there is little raciality and age representation. There are few avatars that show a healthy and fun laugh, based on a continuum of eccentricity and stereotypes of control and desire, which perpetuates old gender gaps and leads to unchanged patterns of beautiful and free laughter.