User-Based Algorithmic Auditing

Abstract

In the artificial intelligence and cloud computing age, digital platforms like Meta, Google, and Amazon wield immense social, economic, and political power, shaping users’ daily lives. As these platforms gather vast amounts of user data and utilize sophisticated algorithms to personalize services, they also expose users to risks of bias and manipulation. Policymakers seek ways to hold platforms accountable, and algorithmic auditing is emerging as a key approach. However, existing regulations often rely on self-audits by the platforms themselves, leading to conflicts of interest. The shift towards third-party auditing is promising but still falls short of resolving these conflicts. To address this challenge, this article introduces, typologizes and explores a unique and underutilized approach to regulatory algorithmic oversight: “user-based algorithmic auditing.” According to this auditing approach, the platforms’ users lead the audit or assist external auditors in the process. User-based auditing is impartial, as it is entirely independent of the audited platforms. User-based audits are also valuable for corroborating the information the platforms provide in their self-auditing reports. The article explores regulatory frameworks, scrutinizes auditing approaches, and delves into the potential of user-based auditing to shape algorithmic oversight policies effectively.

Keywords: algorithm, audit, AI, Artificial Intelligence, fairness