Como construir uma inteligência artificial orientada para justiça social? Um estudo de caso do Radar Antigênero sob as lentes do Feminismo de Dados e das Inteligências Artificiais Feministas

Abstract

Research such as that by Taso, Reis, and Martinez (2023) and Neves-Barros (2025) shows that generative artificial intelligence systems exhibit discriminatory biases, particularly related to gender. In the context of investigations into algorithmic sexism, this work aims to analyze the development process of the Radar Antigênero platform, created by the Gênero e Número Journalism Association in partnership with the company Novelo Data. The platform is a data unlocking initiative that uses artificial intelligence to analyze and classify videos containing anti-gender discourse published on YouTube, resulting in the production of counter-data (D’Ignázio, 2024). The overall objective of this research is to understand the tactics, according to Certeau's (1998) conception, mobilized by Radar Antigênero to challenge and subvert sexist biases present in generative artificial intelligence systems. To achieve this objective, the lenses of data feminism (D’Ignázio; Klein, 2020) and feminist artificial intelligence (Toupin, 2024; Jaime, 2024) are used, and a case study methodology (Yin, 2001) is adopted, consisting of a usability analysis of the interface according to Nielsen's heuristics (2024) and semi-structured interviews (Gaskell, 2002) with the team that developed the initiative. The investigation indicates that the platform was conceived and designed with the aim of dismantling systemic violence that results in structural inequalities and follows the trend of machines driven by social justice. The results also demonstrate the crucial role of journalism with an intersectional perspective (Batista; Gonçalves, 2025; Rodrigues; Aguiar, 2023) in producing investigations that examine power and contribute to a more egalitarian society.