Purpose-built AI tutor beats active-learning class in Harvard physics trial
In a 194-student physics trial, the AI-tutor group had a median post-test score of 4.5 versus 3.5 in class while spending less time; the result is context-specific.
A Harvard study involving 194 introductory physics students compared a purpose-built AI tutor with an active-learning classroom session. The tutor used questions, explanations and reasoning checks rather than simply supplying answers.
The reported median post-test scores were 4.5 for the tutor group and 3.5 for the classroom group. Tutor users spent about 49 minutes compared with roughly 60 minutes in class, and the study reported larger learning gains for the tutor group.
The experiment covered limited physics topics in one university setting. It supports further study of tailored tutoring tools, not a conclusion that AI can replace teachers or universities.
