Multi-Stakeholder Dialogue for Policy Recommendations on Algorithmic Fairness

Abstract

Multi-stakeholderism is a valuable methodology for governance and policy development. We describe the use of the approach in the UnBias study, which seeks to identify opportunities for effective governance of algorithmic online services. We use the multi-stakeholder methodology to bring together experts from relevant sectors including academia, education, government, regulation, law, civil society, media, and industry and commerce. This paper outlines how our work has facilitated open and constructive debate that can drive the development of meaningful policy recommendations. We also describe some challenges of this approach and our next steps towards producing actionable design and policy recommendations, including engagement with international industry standards development.