Feature

Research Insights: AI and the Future of Human Legal Reasoning

Four Minnesota Law Faculty Ask if AI Can Improve Performance in Legal Tasks

A smattering of dark and light blue pixels, invoking a retro-futurism aesthetic.

Since ChatGPT’s public debut in 2022, the legal profession has wrestled with a fundamental question: Will artificial intelligence enhance lawyers’ work or erode the reasoning skills at the core of the profession? While opinions have been plentiful, evidence has been scarce.

A recent empirical study by Minnesota Law professors Nicholas Bednar ’16, McKnight Land-Grant Professor; David Cleveland, Clinical Professor of Law and Director of Legal Writing & Research; Allan Erbsen, Popham, Haik, Schnobrich/ Lindquist & Vennum Professor of Law; and Daniel Schwarcz, Fredrikson & Byron Professor of Law and Distinguished University Professor, brings new clarity to the debate. Their findings begin to answer the question with data rather than speculation and suggest a more complex and encouraging story than many might expect.

In a randomized controlled study, the research team worked with approximately 100 upper-level law students who were tasked with completing a sequence of foundational legal tasks, including synthesizing source materials, applying them to a client problem, and revising a written memo. Participants were divided into two groups. One group completed most of the work without AI assistance, while the other used AI tools at key stages.

The results confirmed what many have already observed: AI can significantly improve efficiency and performance in certain legal tasks. Students who used AI to synthesize legal materials outperformed their peers by a substantial margin, improving their work by 50% to 70% and completing the task more quickly.

But the study’s most notable finding emerged in the next phase.

Even when all participants were required to complete a legal analysis without AI, those who had previously used it still performed better. Early access to AI appeared to have strengthened, not weakened, their independent reasoning. The researchers suggest that AI may improve the quality of initial work by providing users with a stronger analytical foundation that carries over into subsequent work.

At the same time, the study underscores that AI is not without risk. In the final phase, when all participants were asked to use AI to revise their memos, the results diverged. Students who started with weaker drafts often improved with AI assistance. But those who began with stronger analyses sometimes saw their work decline in quality after incorporating AI-generated revisions. In these cases, AI appeared to substitute for, rather than support, careful thinking.

This tension points to a key finding: AI’s impact depends more on how the tool is used than on the tool itself.

The research also acknowledges its limitations. The study focused on law students rather than practicing attorneys, and the tasks were more structured than many real-world legal assignments. There are still many questions about how sustained reliance on AI may affect long-term learning and professional development.

Even so, the findings offer practical guidance for a profession in transition. AI can be a powerful aid when used to support specific, well-defined tasks, particularly when lawyers can independently assess and build on its output. But it is less effective and potentially counterproductive when it replaces careful reasoning altogether.

As AI evolves, the legal profession must decide not just whether to adopt it but how to do so without losing the intellectual discipline that defines legal practice.

Read the full text of this study on SSRN

Minnesota Law Magazine

Spring 2026
Minnesota Law Magazine wordmark