Interesting talk both as an introduction to an AI architecture called ACT-R, and as a novel use for it: to model how people think when they program. Also has several great references along the way. (Download the slides first so you can follow along; requires logging in to infoq.)