<\/p>\n\n\n\n
On Monday 29 January 2025, the GRAIN network organised an online training session on \u00abArtificial Intelligence and Gender\u00bb for the beneficiaries of the Grain project, who are also members of the GRAIN network.<\/p>\n\n\n\n
The aim of the training was, amongst other things, to enhance participants\u2019 understanding of how to integrate gender perspectives into AI initiatives, to explore the role of evidence in designing effective and inclusive AI solutions, and to provide practical tools for applying gender-sensitive approaches to AI projects.<\/p>\n\n\n\n
Led by Sixtus Onyekwere, a researcher in international development and gender expert at the Centre for the Study of African Economies (CSEA), the training session raised awareness of the importance of integrating a gender perspective into technological innovation processes, particularly in the field of AI; and to rethink the design, deployment and evaluation of AI systems through the lens of equity, inclusion and social justice.<\/p>\n\n\n\n
The session provided an opportunity for the gender expert to revisit key concepts such as algorithmic bias, intersectionality and the ethics of AI. She emphasised the idea that AI can either reinforce existing inequalities or act as a catalyst for positive change, depending on how it is designed and implemented.<\/p>\n\n\n\n
Indeed, whilst artificial intelligence is often perceived as objective, it is nevertheless subject to biases that may be present in the data, the algorithms or the teams that design it. Gender bias can have tangible consequences, including the exclusion of certain groups, the reinforcement of stereotypes, and discriminatory decisions in recruitment, credit assessment or predictive justice algorithms. The trainer shared examples of gender bias, including: facial recognition; AI-based human resources tools; and medical diagnosis, etc.<\/p>\n\n\n\n
The session also provided an opportunity to discuss methods and tools for considering how to better integrate gender, such as: regular audits of AI models to detect gender and racial bias; incorporating diversity and inclusion criteria into technology performance indicators; encouraging a diverse range of profiles within design teams (women, minorities, people from the social sciences); working with a diverse range of end-users to test tools in varied contexts; collecting representative data that takes gender differences into account; considering the social impact of innovations right from the design stage, rather than as an afterthought; and implementing methods for anonymising and de-biasing data.<\/p>\n\n\n\n
Following the presentation, the discussions provided an opportunity for network members to express the difficulties they faced in putting these best practices into practice in certain contexts (notably a lack of resources, low awareness and institutional resistance). They therefore believe it is important to train developers and decision-makers in critical thinking and a gender-sensitive approach from university level or within technical courses.<\/p>\n\n\n\n