SMU Law Review Forum
Abstract
Just as humans trust human doctors to treat human ills, we trust human judges to resolve human disputes. A doctor may employ artificial intelligence (AI) tools to aid in diagnosis, but the doctor ultimately decides what to prescribe, and the doctor bears accountability for that choice. This Article argues that judging demands the same boundary: AI may assist as a tool, but it must not replace human decision-making.
Perhaps unsurprisingly, in light of AI’s recent gains in its ability to process large quantities of seemingly complex material, some scholars and practitioners have contended otherwise, urging that at least some judging be handed off to AI entirely. This Article argues against that call. First, there are many reasons to resist any delegation of judicial power to AI, chiefly that the process of determining the “right answer” to a novel legal question is inherently a human one. That is, even if an AI model arrives at a credited “right answer” much of the time, that result is not what makes the answer correct in the first place. Moreover, constitutional and empirical concerns weigh against AI adjudication. Second, drawing on the comprehensive statutory regime governing federal magistrate judges, there are at least three important constraints that we can and should impose upon the courts’ use of AI: Such use should be consistent with well-informed legislation or regulation, it must be meaningfully transparent, and there must be de novo review by an Article III judge over any AI determination.
Recommended Citation
Kyle Singhal,
Dishonorable Machines: Why and How to Resist Trial by AI Judges,
79
SMU L. Rev. F.
155
(2026)
Included in
Computer Law Commons, Criminal Procedure Commons, Internet Law Commons, Law and Society Commons
