Building on Jeremy Paul’s post on a recent paper testing the effect of AI use on law student learning, titled (the post, by Jeremy) “Is AI the Devil?”:
One practitioner posted this thoughtful reflection on the paper and on what its findings imply when lessons on AI-in-law-school are transposed into rough and tumble of law practice.
A brief taste, after the jump:
The study’s authors offer three sensible practice guidelines. Only use AI on tasks where you can independently assess, adapt, and explain the output. Keep AI use narrow and bounded rather than broad and sprawling. Never rely on AI when you’re cognitively depleted or crushed by time pressure.
These are good rules. But there’s a fourth principle embedded in the findings that the authors don’t quite state directly: know what kind of task you’re in.
Synthesis is a good AI task. You’re building a map. AI helps you build it faster and more completely, and that investment pays dividends when you have to navigate the territory alone. Revision of your own best work is a dangerous AI task — especially if that work represents genuine strategic thinking that AI’s averaging tendencies will mistake for an error to be smoothed out.
