How can feminism inform AI governance in practice?

As we navigate the rapidly evolving landscape of artificial intelligence technologies, we must ensure that our ethical frameworks and governance structures reflect the principles of diversity, equity, and inclusion, contributing to the achievement of multidimensional justice. In this text, we will explore how a feminist perspective on AI governance can be instrumental in advancing gender equality and diversity throughout the full AI life cycle. Additionally, I will provide examples of initiatives that showcase the experiences of adopting feminist approaches to AI governance. 

Feminist AI governance 

Feminist AI governance is an emerging field of policy, research, advocacy, and development that aims to ensure that artificial intelligence (AI) systems are governed, designed, developed, deployed, and used in ways that are just, equitable, and inclusive. Feminist AI governance aims to address power imbalances in the AI ecosystem, considering structural inequalities, colonial legacies, and the multidimensional harms that disproportionately affect communities of the global majority.  

Based on feminist principles and a critical understanding of the role of technology in society, feminist AI governance understands that algorithms “are not just mathematical entities,” but are contextual, situated, and shaped by social relations and cultural values. As Cathy O’Neil puts it: “Algorithms are opinions embedded in code.” 

Therefore, instead of being understood just as a heterogeneous set of technologies, algorithmic assemblages result from the ensemble of the mathematical and the infrastructural, the sociopolitical and the cultural. The understanding of algorithms as sociotechnical assemblages, not only as abstract entities, calls attention to the ways in which algorithms can perpetuate inequality and therefore have deep societal implications. 

As AI systems respond to economic, historical, sociopolitical, legal, and cultural contexts in which they emerge and are deployed, feminist AI governance scrutinizes these systems as products of power relationships and exposes their historical and situated nature. The primary contributions of feminist AI governance are tied to a systemic/structural approach, the identification of power asymmetries within the AI ecosystem, the recognition of harm as multidimensional and disproportionately affecting marginalized bodies and territories of the global majority. It is driven by a commitment to achieving integral, planetary justice, considering AI systems throughout their full life cycle, beyond just the AI/data pipeline. 

Some key points about feminist AI governance include the reflection about how to distribute power in a highly concentrated industry, unequal ecosystem, with lack or representation in and governance structures of, countries, territories and populations of the global majority, that are at the same time contributing with data, labor, and natural resources. 

Feminist AI In Practice: technologies, policy, research 

In this section, I will share three initiatives that approach AI governance, development, policy, and research following feminist principles. These initiatives involve various stakeholders: practitioners, civil society, and academia; nation-states; and intergovernmental organizations.  

Incubating Feminist AI in Latin America: practitioners, civil society, and academia 

In the past three years, my work has been deeply intertwined with the Incubating Feminist AI project, focused on the development of feminist AI in three regions: Latin America and the Caribbean (LAC), Middle East and North Africa and South East Asia, funded by the International Development Research Center (IDRC). In this case I will refer only to my experience as leader of the LAC hub of the Feminist AI research network in which the Incubating Feminist AI project emerged, as part of the <A+> Alliance. 

Frequently, I encounter questions probing the feasibility and meaning behind this concept and project. In response, I assert that we can indeed craft technology rooted in care—an approach that involves nurturing processes and challenging the underlying assumptions of what technology is and how it should be developed: caring about people, caring about the environment. Some fundamental questions guide our journey are: Some fundamental questions guiding our journey are: Who can develop AI? Where? How? For what purpose? Why? 

This project is evidence of the possibilities and need to develop technologies guided by principles rooted in social justice and human rights, with a commitment to achieving social justice and combating gender-based violence while being grounded in communities. The 13 proposals incubated in our region address profound societal issues, showcasing the immense potential for women and racialized communities to actively participate in imagining and developing the technologies they need, following the principle “Nothing about us, without us.” 

However, this venture faces constraints. Despite its success, the project’s impact is hampered by limited funding, primarily sourced from abroad. Moreover, the proposals we incubate lack integration into a national or regional public policy that could propel them to technological maturity, ensuring their sustainability over time and solidifying their role as tools for communities showing the need to develop mechanisms to support community-based technological development. 

Feminist AI Policy: Argentine Ministry of Foreign Affairs and UNESCO 

In October 2023, the Argentine Ministry of Foreign Affairs and UNESCO organized the Symposium “Feminist Contributions in Artificial Intelligence: Epistemological Justice and Ethics of Equality” in Buenos Aires. This initiative aimed to collectively formulate basic principles and recommendations for the development of a feminist AI public policy. The active participation of women leading AI projects, government representatives, policymakers in the field of gender equality in science and technology, lawyers, thought leaders, research groups, and networks working on this agenda demonstrated the immense talent and capacity in the region. The recommendations were integrated into a public report presented at the Summit of High Authorities on Artificial Intelligence in Latin America and the Caribbean in Chile. 

Towards Real Diversity and Gender Inclusion in AI Ecosystems: Responsible AI Working Group, Global Partnership on Artificial Intelligence (GPAI) 

Based on a human rights framework, a systemic, feminist, intersectional approach, and utilizing participatory methodologies, and design justice principles, “Towards Real Diversity and Gender Inclusion in AI Ecosystems” aims to provide recommendations, tools, and promising practices for integrating Diversity and Gender Equality (DGE) considerations into the AI life cycle and related policy-making. This project conducted global consultations with nearly 200 participants and represents the first effort to translate the gender dimension considered in the UNESCO Recommendation on the Ethics of AI into an actionable roadmap, framework, repository, and training program for governments and other stakeholders in the AI ecosystem. 

These experiences underscore the importance of institutional support, resources, and political will to amplify the transformative force of feminist AI and feminist AI governance. To realize their full potential, we need sustained funding, strategic integration into national policies, and a shift towards a broader governmental vision that recognizes the importance of these initiatives to advance social justice and access to human rights. By contributing to the construction of inclusive participatory governance structures, we aim to inspire and inform professionals across sectors about the importance of embedding gender equality and diversity in the very fabric of AI.