Audit, Don't Explain -- Recommendations Based on a Socio-Technical Understanding of ML-Based Systems

TL;DR

This paper advocates for systematic audits over explainability in ML systems, emphasizing a socio-technical approach and providing recommendations for public institutions to ensure these systems serve the public interest.

Contribution

It introduces a socio-technical perspective on ML systems and proposes concrete audit-based methods as an alternative to explainability for public accountability.

Findings

Systematic audits can be more effective than explainability for public oversight.

Socio-technical understanding is crucial for evaluating ML system impacts.

Recommendations aim to guide public institutions in auditing ML systems.

Abstract

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 T\"UV and Stiftung Warentest can ensure that ML systems operate in the interest of the public.