AI’s Ability to Reason with Human Norms and Values in Legal Judgments

Ascertain whether AI systems, including large language models, can perform the value-laden reasoning required to render legal judgments—such as determining civil liability—that demands making hard tradeoffs and evaluating conduct in light of human norms and values.

Background

The authors argue that judicial decision-making involves deliberative processes and value tradeoffs that go beyond routine legal tasks. They express skepticism that AI can operate on these registers, which raises concerns about delegating legal judgments to AI.

This uncertainty underpins their recommendation to view judicial use of AI skeptically, particularly for decisions that hinge on normative evaluation.