AI and human tutoring yield equivalent GRE learning gains
10 hours ago
- StudentBench is a public platform with over 175,000 student-AI messages for evaluating AI teaching.
- A study of 2,383 participants found AI tutoring statistically equivalent to expert human tutoring for GRE learning gains (p = .015).
- In five of seven GRE domains, the best AI tutor outperformed the human tutor on average.
- A second study had human experts rate LLM-generated lesson plans and practice problems via 2,028 pairwise rubric evaluations.
- The studies distinguish AI tutors across lesson planning, practice-problem creation, conversational pedagogy, cost, and engagement.
- One AI tutor matched human tutoring at 918 times lower cost: $0.0052 vs. $4.81 per percentage point gained (p = .044).
- Faster AI replies correlated with more student messages, more messages with more correct practice, and more correct practice with larger learning gains (all p < .002).
- The StudentBench platform is freely available online.