Data Provenance for Accountability Mechanisms and Properties

Data provenance has been proposed as a component of accountable systems design because it attributes system events to system agents and captures how processes use and generate data. However, achieving “accountability” as part of a mechanism design or as a system property remains imprecise since no consensus exists on the term’s definition. We synthesize prior accountability formalisms to understand their amenability to data provenance, we show how accountability is related to assessment of responsibility and blame, and we consider future challenges in building provenance-driven accountability systems.