Audit, Don't Explain -- Recommendations Based on a Socio-Technical Understanding of ML-Based Systems
This addresses the problem of ensuring accountability and public trust in ML systems, particularly for regulatory bodies, but is incremental as it builds on existing socio-technical perspectives.
The paper argues that systematic audits are preferable to explainable AI for ensuring public interest in ML-based systems, proposing concrete recommendations for institutions like the German TÜV and Stiftung Warentest.
In this position paper, I provide a socio-technical perspective on machine learning-based systems. I also explain why systematic audits may be preferable to explainable AI systems. I make concrete recommendations for how institutions governed by public law akin to the German TÜV and Stiftung Warentest can ensure that ML systems operate in the interest of the public.