Assembling Accountability: Algorithmic Impact Assessment for the Public Interest

This report maps the challenges of constructing algorithmic impact assessments (AIAs) by analyzing impact assessments in other domains — from the environment to human rights to privacy. Impact assessment is a promising model of algorithmic governance because it bundles an account of potential and actual harms of a system with a means for identifying who is responsible for their remedy. Success in governing with AIAs requires thoughtful engagement with the ongoing exercise of social and political power, rather than defaulting to self-assessment and narrow technical metrics. Without such engagement, AIAs run the risk of not adequately facilitating the measurement of, and contestation over, harms experienced by people, communities, and society.