Can we predict whether an AI's debugging advice is actually good advice, before a student ever sees it? A new paper by Brendan Toscano tackles exactly that question, using thousands of real student coding submissions to build models that flag helpful feedback from unhelpful fixes.
