Awesome exam! Last semester I had students had pick a task for AI (several chose legal-related), make a hypothesis for how it'd do, pick measures to evaluate it, and then test it out. Helped expose gaps between expectations and reality (AI mostly did better than they expected). Interesting times.