Auditing Artificial Intelligence as a New Layer of Mediation: Introduction of a new black box to address another black box

Abstract

The discourse of auditing artificial intelligence (AI) is emerging as a field of applied AI ethics to address the problems of fairness, bias, and accountability of AI systems. The injunction of auditing AI assumes and proposes an expert figure of an auditor to address the black box problematic of AI. Contrary to this, the text argues that the figure of an expert auditor installs another layer of mediation. This layer of mediation introduces another black box –the auditor, auditing practices and professional scepticism- to address the existing black box problematic. The text argues that it is important to radically interrogate the need or the efficacy of the figure of an expert AI auditor to whom all users of an AI system delegate their representation. The introduction of audit cultures in the AI ecosystem risks foreclosing future debates on AI and its relationship with humans and other non-humans and risks reducing the problem again to a technical one.