FAILS: a tool for assessing risk in ML systems
Gonzalo Aguirre Dominguez, Keigo Kawaai, Hiroshi Maruyama · 2021
Quality assurance of AI based systems presents a unique set of challenges to software engineers, making it difficult to assess the risks involved when deploying them. We present a risk assessment tool based on the widely used failure mode effect analysis (FMEA) methodology, as well as quality assurance guidelines released in recent years. The tool aims to support the search for potential risks in machine learning (ML) components used in the design and development of AI products. A preliminary evaluation showed its effectiveness and pointed toward areas for future improvement.