Generative AI in Legal Education: A Two-Year Experiment with ChatGPT
Abstract
This report hopes to contribute to the discussion around the use of generative AI in legal education.
The literature is currently divided: some celebrate ChatGPT’s potential to assist drafting and reasoning; others warn of shortcut learning and degraded skill acquisition. I seek to add evidence to this scholarly debate in two ways. First, I review the growing body of studies reporting experiments conducted in law schools, with the aim of offering readers a repository of pedagogical initiatives that they can replicate and refine in their own classroom.
Second, I present findings from a two-year experiment of my own, in which students faced different teaching techniques around generative AI: prohibition, unstructured exposure, and structured training. I find that structured training produced a clear advantage in the first year, but its edge diminished as baseline familiarity with AI spread in the second. The experiment also suggests that failing to teach students how to use AI effectively and responsibly is a missed opportunity, yet its absence does not lead to the systemic breakdown that some have forecast.
I draw practical lessons from these observations. In the short term, law school boards should resist the temptation to impose one-size-fits-all choices on the entire faculty body. Instead, professors should retain freedom to experiment. In the medium term, I suggest that (current) long-form writing (such as master theses) should be retired. Finally, law schools will benefit from investing in AI literacy for both students and professors if they want to remain attractive in the long run.
Read the full piece at SSRN.