AI Decisions Are Sound, Nothing but Sound

Administrative agencies are progressively relying on AI systems, a phenomenon scholars dub “government by algorithm.” Agencies might use AI systems in the way that police officers use breathalyzers, that is, as means of ascertaining (or predicting) factual information that feature in law’s mandates. Alternatively, they might use such systems as deciders. I will call this latter use of AI systems “automation.” Some prominent voices are proposing automation as a way of addressing the purported influence of bias and noise on human decisions. I will argue that this is woefully ironic and that automation will instead drown the administrative state in bias and noise. To that end, I first show how the proponents of automation conflate bias and noise, which are process faults, with inaccuracy and inconsistency, which are output faults. As I shall explain, it is due to this conflation that the proponents falsely take automation’s (theoretical) potential to reduce inaccuracy and inconsistency as a potential for it to reduce bias and noise. In reality however, if automation could reduce inaccuracy and inconsistency, it could only do so by way of increasing bias and noise. Thus, addressing inaccuracy and inconsistency with automation would come at the cost of increasing bias and noise. After dispelling the foregoing misconceptions, I go on to argue that such a tradeoff is not worthwhile, meaning we should oppose ‘government by algorithm,’ even despite putative promises of greater accuracy and consistency.

Amin Ebrahimi Afrouzi is a Law & Philosophy Fellow at UCLA School of Law and teaches Legal Philosophy. His research lies in Jurisprudence, Interpretation, and the Justness of Political Procedures. He previously held the Knight Digital Public Sphere Fellowship at Yale Law School’s Information Society Project, where he worked on AI and the Law. Amin received his BA and JD from UC Berkeley. He also holds a master’s degree in Philosophy from Oxford, a master’s degree in Classics from Cambridge and a PhD in legal philosophy from UC Berkeley. Amin is the inventor of various patented or patent-pending AI and robotics technologies. His legal philosophy publications have appeared or is forthcoming in Legal Theory, Oxford Journal of Legal Studies, and the Canadian Journal of Law and Jurisprudence, among others.

Moderator: Benjamin Chen, Associate Professor & Director of the Law and Technology Centre, The University of Hong Kong Faculty of Law

To register, please go to https://hkuems1.hku.hk/hkuems/ec_regform.aspx?guest=Y&UEID=98580. A paper will be circulated in advance and attendees will be expected to have read the paper before the seminar.