Kick-Off Working Group for a Feminist Perspective on EU AI Act Standardization

On 16th of January, our Feminist AI and Digital Policy Roundtable held on the topic of EU AI Act Standardization processes identified the main levers to bring intersectionality into AI policy processes.  

A diverse group debated about effective methods to address the lack of diversity in AI policy.

In the rapidly evolving landscape of artificial intelligence (AI), the European Union's AI Act (EU AI Act) has become a focal point of discussions, aiming to establish a comprehensive legal framework for AI development, deployment, and trade. As decisions are made within this framework, it is crucial to assess the impact on society, especially marginalized and underrepresented groups. This article explores the intersection of feminist perspectives and AI standardization, highlighting the need for inclusivity, transparency, and ethical considerations. 

Background and Urgency 

The urgency of addressing AI standardization is emphasized by the lack of mandatory provisions for marginalized aspects in Article 69 of the EU AI Act. The absence of mandatory considerations for marginalized groups raises concerns about inclusivity and the broader implications for human rights. The Feminist AI and Digital Policy Roundtable, recognizing the urgency, initiated a working group to influence the AI standardization debate, ensuring that marginalized voices take precedence. 

Diversity and Representation 

One significant challenge identified is the lack of diversity in standardization processes. The absence of representation from diverse backgrounds, including women, people of color, LGBTQIA+, and other marginalized groups, hampers the development of inclusive AI standards. The call to action emphasizes the importance of addressing this issue promptly to prevent further perpetuation of discriminatory practices in AI systems. 

Insights into Standardization Processes 

The standardization process under the EU AI Act involves various stages, including the formulation of guidelines and regulations by European standardization bodies. The development of standards necessitates the input of stakeholders, and here lies a critical problem – the underrepresentation of civil society, especially marginalized voices, in these decision-making bodies. The lack of diversity in these forums threatens to reinforce existing power structures and hinder the creation of equitable AI standards. 

Feminist Perspectives on AI Fairness 

As part of our Feminist AI and Digital Policy Roundtable, we delved into the importance of redefining the concept of fairness in AI systems from a feminist standpoint. Instead of focusing solely on "fairness," we propose to embrace the term "justice." This shift aims to acknowledge existing power imbalances and emphasizes the need for human intervention in AI lifecycles to reduce potential discrimination. The argument underscores the significance of setting high standards in AI to avoid ethical-washing practices that could compromise democracy. 

Challenges in AI Metrics and Fairness 

The discussion on fairness metrics in AI systems reveals a technical and ethical challenge. The predominant focus on performance metrics rather than justice theories highlights the gap between the technical discourse on fairness and its ethical implications. The need to bridge this gap is crucial for ensuring that AI standards align with principles of justice and inclusivity. 

Inclusive AI Standard Setting 

The EU AI Act provides a unique opportunity to shape the future of AI, ensuring that it serves the interests of all individuals, especially those from marginalized communities. The sub-working group on European AI standardisation takes on the charge to enable diverse representation, transparency, and ethical considerations in the creation of AI standards that align with principles of justice and inclusivity. 

CALL TO ACTION

This blog article aims to present the main findings and motivate the readers to engage in our sub-working group on EU AI Act standardization. For further information or interest in participation please contact info@fem-ai.com.