Qualification Cases for AI-Supported Workflows
Abstract
Existing approaches for tool evaluation and qualification are based on assumptions that do not hold for AI-enabled tools. Building on the idea of preparing an assurance case for a safety-critical system, this paper introduces the notion of a “qualification case” that marshals arguments and evidence about the adequacy of AI-supported workflows used in the development of critical systems. An AI-supported workflow for preparing Failure Modes, Effects, and Diagnostic Analysis (FMEDA) is used to illustrate the notion of a qualification case.