Measuring Teaching Efficacy Through Comprehension Delta: Large Language Models as Standardized Learner Proxies

Measuring teaching is hard because learners are not standardized. We propose using calibrated LLM proxies, themselves benchmarked against human cohorts, as a stable instrument for measuring the comprehension delta induced by a lesson.

Drawn from a decade of running an educator-centric platform, the paper proposes a methodology for evaluating teaching efficacy through before-and-after comprehension shifts in standardized learner proxies.

Submitted to Teaching in Higher Education and currently under peer review.