Was reminded recently of one of my favorite papers on social learning. It centers on a clever lie! Subjects study some material, then quiz either a peer or a bot about it. They learn much better when they think they're quizzing a peer. Secretly, in both cases, it's a bot!