The digital mirror: how generative artificial intelligence reflects and amplifies gender bias

Abstract

We live in unprecedented times for women in sport. In the USA, the 2024 collegiate basketball finals marked the first time in broadcast history that viewership of the women’s championship game was higher than the men’s.1 Internationally, the 2024 Paris Games were the most gender equal in Olympic history in terms of numerical parity on the playing field.2 Despite such progress, gender inequities in sport persist, which social scientists partially attribute to the harmful effects of stereotypical and often sexualised representations of women in sports media.3 As generative artificial intelligence (GAI) is increasingly leveraged for visual communication of sport, the risk of amplifying gender stereotypes is profound.4 Recently, authors of this editorial (J.M. and M.M.) created an image for social media use to promote findings from a study on inequities in training environments of elite women golfers.5 A GAI tool was prompted to ‘provide an image of women golfers.’ Dissatisfied with the initial stereotypical result, a second prompt was given: ‘show a more diverse representation of body types and ethnicity.’ There is much to deconstruct in the output (figure 1); at …