Is AI Alone Not Enough? The "Slight Human Support" That Worked in Schools: Why a Small Human Touch Can Multiply the Learning Effect of AI

A research team from Carnegie Mellon University, the University of Hong Kong, and Stanford University conducted a study on "hybrid tutoring," which combines the use of AI tutors with human remote guidance, tracking 7th-grade students in the United States over one academic year (2023–2024). Compared to the previous year's group that used AI only, the group with human support showed a growth difference of 0.36 grade levels by the end of the year. While there was no overall difference in the total scores of standardized tests, an interaction effect was observed where the synergy between human and AI support strengthened as "time on task" with AI increased. Human tutors focused on motivation and early detection of stumbling blocks via Zoom, while AI provided consistent, individualized hints, indicating a functional division of labor. On social media, there is a notable positive reaction emphasizing that human support contributes to equity (support for the disadvantaged), alongside a caution against the excessive expectation that "AI alone is sufficient." However, concerns about the "limitations of comparative design" and the "redefinition of the teacher's role" have also emerged. Moving forward, challenges include refining cost-effectiveness, training methods, and ethical considerations.