In Study, AI Helped Law Students Without Hurting Reasoning

Published
Score
13

Why it matters

A randomized controlled trial from researchers at the University of Minnesota and University of Michigan law schools found that two AI tools—Vincent AI (a retrieval-augmented generation platform from vLex) and OpenAI's o1-preview reasoning model—improved upper-level law students' performance on six legal tasks by 38 to 140 percent. The study, published on SSRN in April 2026, contrasts sharply with earlier research on GPT-4 and similar models, which showed that while AI accelerated task completion, it degraded reasoning quality. Vincent AI kept hallucinations at baseline levels, while o1-preview enhanced analytical depth but introduced occasional errors.

The trial tested whether newer AI architectures could overcome the reasoning-quality tradeoff documented in prior studies. Critically, researchers found no evidence that students' independent reasoning abilities suffered when AI tools were unavailable after the task concluded. The full details of the study methodology and results remain available only in preprint form.

For practicing attorneys and firms evaluating AI adoption, the findings suggest that current-generation tools may genuinely augment rather than atrophy core legal skills—a material shift from earlier warnings about AI's pedagogical risks. With 55 percent of U.S. law schools now offering AI courses and integration accelerating since 2023, this empirical evidence arrives at a pivotal moment for legal education and junior attorney training. Firms considering AI-assisted workflows should monitor whether these results hold across practicing lawyers and real client work.

mail

Get notified about new Artificial Intelligence developments

Primary sources. No fluff. Straight to your inbox.

See more entries tagged Artificial Intelligence.

Also on LawSnap