The Students AI Tutoring Still Leaves Behind

We tend to assume that putting an AI tutor in a classroom automatically spreads support more evenly. A new study says otherwise. Researchers at North Carolina State University and Carnegie Mellon analyzed more than 1.4 million student-system interactions across 339 middle school math students in 10 schools, and the finding is uncomfortable: teachers using intelligent tutoring systems kept circling back to the same students they had already helped.

The pattern held even after the researchers accounted for which students were actually engaged or disengaged. In the words of lead researcher Qiao Jin, "teachers are more likely to interact with students that they have interacted with before, even after considering who is engaged and disengaged." The help was real. It was just sticky. And sticky help is not the same as fair help.

The Tool Was Not the Bottleneck

It would be easy to read this study as a knock against AI tutoring. It is not. The intelligent tutoring system was doing its job, generating practice and surfacing where students struggled. The constraint was human bandwidth. A single teacher with 28 students and 50 minutes cannot manually reach everyone, so attention flows along the paths of least resistance: the student who asked yesterday, the one already on the radar, the familiar face at the front table.

That is not a character flaw. It is arithmetic. And it is exactly the arithmetic that AI tutoring should be relieving rather than reproducing. The study landed on a sensible recommendation: give teachers dashboards that show their own intervention patterns so they can allocate help in line with their stated fairness principles. We would push the idea one step further.

Reaching the Student Who Never Asks

This is the gap Socrat, the AI platform inside Core Learning Exchange, is designed to close. The point is not to flood every student with attention. It is to make sure the quiet student, the multilingual learner translating the question twice before answering, and the learner reading two grade levels below the text all receive support without first having to flag themselves to a busy adult.

Two design choices matter here. The first is that the tutor is available the moment a student stumbles, not the moment a teacher happens to walk by. The second is that it works through Socratic dialogue rather than answer delivery, keeping students in the productive struggle that actually builds understanding instead of handing over a solution that builds nothing.

Data Should Widen the Circle, Not Confirm It

The NC State team is right that dashboards help. But a dashboard that only reports teacher behavior risks becoming a mirror. The more useful signal points outward, toward the students themselves: who has not engaged the tutor in three days, whose mastery curve has flattened, who is quietly stuck on the prerequisite that everything else depends on. Continuous mastery tracking is built to identify those learners earlier than traditional alerts, which tend to fire only after a grade has already slipped.

Used that way, the same analytics that exposed the attention gap can help close it. The teacher still decides where to spend their limited minutes. The platform simply makes sure the list of who needs them is complete, and that the students who never ask are on it.

AI in education does not earn its keep by helping the students who were already going to be fine. It earns it by reaching the ones the system has always been most likely to miss. That is the bar. This study is a useful reminder of how easy it is to fall short of it, and of why the design of the tool matters as much as its presence in the room.

Read the Full Article

Read "Teachers tend to help the same kids repeatedly when using AI-powered tutoring tools" on Phys.org

About Core Learning Exchange: We provide turnkey Career and Technical Education (CTE) solutions for grades 6-14, offering 450+ courses from 20+ providers aligned to state standards and industry certifications. Our AI platform uses proven Socratic methodology to develop critical thinking skills through personalized, adaptive learning—deployed in hours via LTI integration.

Related Posts