The journal Big Data & Society has published the article “Double standards of generative AI chatbots: Unveiling (digital) ageism versus sexism through sociological interviews”, co-authored by Francesca Belotti, Mireia Fernández-Ardèvol, Veysel Bozan, Francesca Comunello and Simone Mulargia.
The article analyses how generative artificial intelligence systems may incorporate and reproduce social biases, with particular attention to digital ageism. From a sociotechnical perspective, the study examines how generative AI chatbots represent users according to age and gender, comparing ageist biases with sexist biases.
As a methodological innovation, the authors interviewed five popular generative AI chatbots as if they were interlocutors. Through these sociological interviews, they sought to identify ageist and sexist stereotypes in the systems’ responses, especially in relation to the digital practices of different user profiles.
The findings show that chatbots reproduce two types of “double standard”. The first concerns political correctness: the systems tend to avoid sexist responses about the digital practices of men and women, but they do not show the same caution when dealing with age-related stereotypes. In other words, chatbots appear to be more sensitive to sexism than to ageism.
The second double standard refers to how chatbots imagine the usefulness of artificial intelligence for different user profiles. According to the article, the systems’ responses attribute different uses and needs depending on age and gender, reproducing stereotypical associations about young people, older people, men and women. This shows that social biases do not only appear in data or algorithms, but also in the way generative AI organises and communicates information about social life.
The study argues that these findings reveal a form of generative epistemic injustice, understood as the reproduction of potentially harmful representations of social groups with less symbolic power, such as older people or women. In the case of ageism, the problem is particularly concerning because it is often less socially contested than other forms of discrimination.
The article thus contributes to current debates on artificial intelligence, inequality and digital culture, showing that generative AI systems are not neutral tools. Their responses incorporate cultural repertoires, social imaginaries and hierarchies of value present in the environments in which they are designed, trained and used.
The article is available in open access at: https://doi.org/10.1177/20539517261419407