Transparency and Explainability for Public Policy
Kate Vredenburgh · LSE Public Policy Review · 2024
Governmental decision-making ought to be transparent and understandable by the political community. However, predictively accurate but opaque AI systems have raised moral and legal challenges for governments wishing to use AI for public policy. This paper makes main claims. The first claim is that both transparent and explainable AI are important. The second claim is that the current generation of AI tools are not sufficient to make governance with AI appropriately explainable. The third claim is that care needs to be taken in crafting decision-maker incentives, and that members of the political community who have to use explanations are owed resources and other support.