Auditing of AI: Legal, Ethical and Technical Approaches

Abstract

AI auditing is a rapidly growing field of research and practice. This review article, which doubles as an editorial to Digital Society’s topical collection on ‘Auditing of AI’, provides an overview of previous work in the field. Three key points emerge from the review. First, contemporary attempts to audit AI systems have much to learn from how audits have historically been structured and conducted in areas like financial accounting, safety engineering and the social sciences. Second, both policymakers and technology providers have an interest in promoting auditing as an AI governance mechanism. Academic researchers can thus fill an important role by studying the feasibility and effectiveness of different AI auditing procedures. Third, AI auditing is an inherently multidisciplinary undertaking, to which substantial contributions have been made by computer scientists and engineers as well as social scientists, philosophers, legal scholars and industry practitioners. Reflecting this diversity of perspectives, different approaches to AI auditing have different affordances and constraints. Specifically, a distinction can be made between technology-oriented audits, which focus on the properties and capabilities of AI systems, and process-oriented audits, which focus on technology providers’ governance structures and quality management systems. The next step in the evolution of auditing as an AI governance mechanism, this article concludes, should be the interlinking of these available—and complementary—approaches into structured and holistic procedures to audit not only how AI systems are designed and used but also how they impact users, societies and the natural environment in applied settings over time.

Algorithmic Decision-Making, Agency Costs, and Institution-Based Trust

“…However, the motivation for such audits and the institutional features required to support them are quite different. Audits for algorithmic externalities have been motivated by activists concerned with social injustice and companies complying with current or prospective regulation (Krafft et al, 2021;Mökander, 2023). On the other hand, the pressure to address algorithmic internalities must come primarily from the human principals themselves.…”

Artificial intelligence, rationalization, and the limits of control in the public sector: the case of tax policy optimization

“…It is also true that many AI systems fail to live up to the claims made by the organizations that deploy them. In these cases, more robust software development practices (Kearns & Roth, 2020), broader impact requirements (Prunkl et al, 2021), guardrails for the use of AI systems (Gasser & Mayer-Schoenberger, 2024), and independent AI audits (Mökander, 2023) can help mitigate risks and prevent harms. However, with respect to the cold, impersonal treatment of decision subjects (O'Neil, 2017), the quantification of social relationships (Mau, 2019), the centralization of decision power and the standardization of decision criteria (Kleinberg & Raghavan, 2021), the automation of cognitive tasks (Brynjolfsson & McAfee, 2014), or the shaping of social preferences through nudges (Thaler & Sunstein, 2008), the problem is not that AI systems 'don't work' but that they work 'too well' as tools for instrumental rationalization.…”