Algorithmic Impact Assessment

An algorithmic impact assessment (AIA) is a structured evaluation conducted before or during the deployment of an algorithmic governance system to identify its potential harms, biases, and social consequences. Unlike conventional regulatory review, which evaluates compliance with existing rules, an AIA evaluates the system's generative capacity — its ability to produce new social phenomena that the rules have not yet anticipated. The AIA requires the operator to examine not just what the algorithm is designed to do, but what it is likely to do given the structural coupling between the algorithm and its environment. The most significant obstacle to effective AIAs is not technical but epistemic: the assessor must predict the behavior of a complex adaptive system whose outputs are emergent rather than designed. An AIA that treats the algorithm as a tool rather than a system is not an assessment; it is a ritual of reassurance.