AI Design: A Responsible Artificial Intelligence Framework for Prefilling Impact Assessment Reports

Impact assessment reports for high-risk artificial intelligence (AI) systems will be legally required but challenging to complete, especially for smaller companies. That is because the current process is complex, costly, and relies on guidebooks with limited assistance. We propose AI Design, a semiautomatic framework for prefilling these reports. It consists of two components: 1) StakeLinker, an interactive tool combining various stakeholders’ perspectives, and 2) FillGen, a large model-based tool that processes stakeholders’ perspectives and produces a report that is reviewed by regulatory experts within a company. We conducted two user studies: the first with 13 AI practitioners who confirmed StakeLinker’s effectiveness in gathering comprehensive input for impact assessment; the second with eight additional practitioners who successfully evaluated a report for a crime analysis system prefilled by FillGen. To show its generalizability, we also made the reports for two other AI systems publicly available.