AI Lifecycle Records and Paradata: Reconceptualizing Documentation Requirements for Transparency and Accountability

Abstract

Artificial intelligence (AI) governance frameworks are emerging in response to growing demands for transparency, accountability, and regulatory oversight. However, many lack integrated recordkeeping requirements needed to document the AI system lifecycle. This study examines how AI documentation, documentation artifacts, and paradata – information describing the procedures, tools, methods, and decisions associated with AI processes – support transparent, accountable, and trustworthy AI governance. Using a multi-phase qualitative research design, the study combined a survey, literature analysis, regulatory review, and case studies to identify documentation requirements across the AI lifecycle. AI laws and governance frameworks from multiple jurisdictions were analyzed alongside case studies from the Saint Louis Zoo, the Bank of Canada, healthcare governance initiatives, and the NATO Archives. The findings show that existing AI governance frameworks remain fragmented and insufficient to support lifecycle accountability. The study identifies the documentation and paradata required throughout data preparation, model development, deployment, monitoring, and retirement, and highlights the critical role of Records and Information Management professionals in establishing governance policies, metadata standards, retention requirements, and contextual documentation to support transparent, auditable, and legally defensible AI systems.