David Tan (Deakin Law School) has posted The Legality of Black Boxes in Administrative Law, in Yee Fui Ng & Matthew Groves eds., Automation in Governance: Theory, Practice and Problems (Hart Publishing 2025). Here is the abstract:
Many machine learning models are black boxes in the sense that one cannot understand why these models arrive at their outputs. Nonetheless, these models can sometimes be very reliable. Supposing that we had black box models that performed extremely well, would the use of such black boxes be lawful under judicial review? A school of thought called computational reliabilism supports the position that reliable performance is sufficient for a justified decision. In this paper I argue against computational reliabilism in administrative law.
Recommended!
To receive new posts from Legal Theory Blog by email, get a free subscription to Legal Theory Stack.
Lawrence Solum