Back to blog

Why does your AI tutor work in trials but flop in classrooms?

Thắng Đoàn
Thắng Đoàn

Your AI tutoring pilot may have looked brilliant, then stalled with real students. The supplied studies suggest the gap is not the model. It is who chooses to use it.

The Harvard result is real, and narrow

The Times of India reports that Harvard students learned more than twice as much with an AI tutor instead of a lecture. These were motivated adults inside a university that ran the experiment. The result proves AI tutoring works when the learner already wants to learn.

Twelve-year-olds walked away

Children will not opt into a tutor that refuses to help them finish homework fast. Futurism describes scientists offering Khan Academy's Khanmigo, a generative AI tutor, to students across Tennessee middle schools. Per Futurism, the tool was directed not to give up correct answers so it could slowly guide students through the learning process. The students largely ignored it. The problem was not the pedagogy. The problem was that children will not opt into a tutor that refuses to help them finish homework fast.

Demand decides usage, not features

Both tools followed the same Socratic design: guide, do not just answer. It failed with kids and succeeded with Harvard undergraduates. Channel News Asia reports a third case, an AI tutor from NCS and Ngee Ann Polytechnic for continuing education learners, many of whom hold full-time jobs and have limited chances to consult lecturers. Here the tool survives because it fills a real gap in support outside class hours. Age, stakes, and intrinsic motivation decide whether guided tutoring feels like help or an obstacle.

What this does not solve

None of these sources show AI tutoring reaching disengaged learners, the group that needs help most. That is partly mechanical. A Socratic tutor demands effort before reward, so the less a student wants to learn, the more the tutor feels like friction. Futurism also notes educators watching students struggle to complete homework without a chatbot. If your students already lack motivation, an AI tutor is unlikely to reach them.

Which of your learner segments actually wants guided help, versus a faster answer?
What would you measure to know whether your AI tutor is being used or quietly ignored?

Pull one week of usage logs and check what share of your target learners opened the AI tutor more than once.

Sources

Times of India: Harvard AI tutor experiment

Futurism: 12-year-olds ignored an AI tutor

Channel News Asia: AI tutor for polytechnic students

Share:

Recommended for you

Enjoyed this article?

Subscribe for new articles. No spam. Unsubscribe anytime.

By subscribing you agree to receive the newsletter. No spam, and you can unsubscribe anytime.