CO-FREDA: A Co-participative, Frugal and Justice-Oriented AI Methodology for Responsible Data-Algorithm Driven Science

Abstract

This paper introduces CO-FREDA (co-participative Frugal, Responsible, Equitable Data-Algorithm Driven Science) as a methodology for designing AI systems that are technically robust, resource-aware, and socially accountable. CO-FREDA combines contemporary computing methods with feminist and decolonial perspectives to move beyond technosolutionism and support end-to-end socio-technical design. It integrates sovereignty-aware data architectures, privacy-preserving and distributed learning, and resource-aware deployment across cloud–edge–fog infrastructures. The methodology translates collective and horizontal approaches, including Data Feminism, Design Justice, the methodology COIA (Co-diseñando una IA feminista), and community-led initiatives into concrete technical decisions: problem formulation, dataset documentation, evaluation protocols, governance mechanisms, consent and refusal processes, and auditable decision-making. Through case-driven experimentation, CO-FREDA supports reproducible pipelines and model stress-testing under real-world constraints, including heterogeneous devices, missing data, biased labels, limited connectivity, and unequal access to infrastructure. The result is a practical framework for building AI systems that are performant, explainable, deployable, and aligned with justice-oriented research rather than extractive optimisation.