Equity beyond metrics: Co-designing solutions for sociotechnical challenges

The research begins in proposing equity as an alternative to fairness due to its context orientation, fostering of belonging, and the focus on the redistribution of power and resources. Drawing on feminist and decolonial thought, this thesis redefines equity as a theoretical and methodological alternative to fairness within AI governance. To operationalise this definition, the research adopts a participatory and design justice approach, combining methods from feminist human–computer interaction (HCI), AI ethics, and feminist theory. It situates accountability within the broader ecosystem of AI governance and examines how mechanisms such as audits, impact assessments, and due diligence processes can be reframed through an equity lens. It addresses the central research question: How can assessing auditing methodologies from an equity approach contribute to fostering reliable AI and upholding fundamental rights in the context of the EU AI Act? The primary objective of this research was to develop and operationalise the concept of equity as an analytical framework for AI governance. While fairness has long been the dominant ethical principle associated with equality and non-discrimination in the AI ecosystem, it has failed to capture the structural, contextual, and lived dimensions of negative impacts to marginalised communities that often face structural and intersectional harms. This thesis therefore proposes equity as an alternative. Part I establishes the conceptual foundations by tracing the evolution of AI governance and accountability frameworks, from global soft-law instruments to binding regional regulations such as the EU AI Act. It argues that while mechanisms like the Fundamental Rights Impact Assessment (FRIA) from the EU AI Act represent an important step, they remain limited in scope, often procedural rather than substantive, and disconnected from the lived realities of those most affected by AI systems. The second objective was to translate equity into practice through participatory and co-design methodologies. The research developed in Part II a set of eight equity indicators, derived from workshops conducted with two distinct groups: AI experts from the EU Horizon project FINDHR and youth participants representing affected communities. FINDHR was selected for its interdisciplinary composition—spanning civil society, academia, and industry—and for its focus on non-discrimination in AI-based recruitment systems, offering a critical environment to test the distinction between fairness and equity. The youth workshops, despite challenges in participation and continuity, provided essential insights into how younger generations perceive and articulate equity within technological contexts. The third objective was to apply the co-created indicators to existing AI auditing methodologies to evaluate how current practices address or fail to address equity concerns. The assessment presented in Part III focused on four third-party ecosystem audits selected for their ethical or community-based orientations. This application demonstrated how the implementation of equity can be used as a diagnostic tool to analyse accountability mechanisms in real-world contexts. The research makes three major contributions: \1. Conceptual Contribution: It develops a new theoretical framework situating equity as a guiding value for AI accountability, linking ethical reflection to feminist and human rights perspectives. This reframing moves beyond technocratic compliance to address how accountability redistributes power and amplifies the voices of those most affected.} \2. Methodological Contribution: It introduces a participatory co-design method for constructing equity indicators that bring together expert and lay perspectives. This approach bridges the gap between ethical theory and technical evaluation, showing that participatory processes can themselves be instruments of accountability.} \3. Applied Contribution: By applying these indicators to real-world audits, the thesis demonstrates that equity-oriented assessments can uncover limitations often obscured in conventional audits—such as lack of contestability, narrow data representation, and weak enablement for affected communities. The findings show that equity assessment fosters alignment between technical reliability and social justice goals. } It bridges technical, political, legal, and ethical domains demonstrating that auditing can serve as a socio-technical practice of accountability rather than a procedural exercise of compliance. The indicators developed here provide a transferable framework for embedding equity into AI governance across domains, particularly within the EU AI Act’s evolving accountability ecosystem. Ultimately, we argue that equity in AI cannot be captured through metrics alone. It demands rethinking who participates in evaluation, how accountability is conceived, and whose perspectives shape the design and oversight of AI systems. Embedding equity into auditing thus becomes not only a matter of technical improvement, but a transformative practice in the age of algorithmic governance.