A new study from NC State University has uncovered an unexpected side effect of AI-powered tutoring tools in classrooms: when teachers use these platforms, they tend to concentrate their help on the same subset of students rather than rotating attention across the whole class. In other words, the AI might be great at identifying which students need help, but teachers are gravitating toward familiar faces instead of following the data.
This matters because one of the biggest promises of AI in education is personalization — the idea that every student gets the right support at the right time. If the human side of that equation isn’t adapting, the technology alone can’t close the gap. The researchers suggest that better dashboard design and teacher training could help, but the finding highlights a real tension between how AI tools surface information and how teachers actually use it.
For anyone learning about AI, this is a valuable case study in a concept called the ‘human-in-the-loop’ problem. AI can provide excellent data and recommendations, but the outcomes still depend on how people act on that information. The tool is only as good as the workflow around it.
The broader context: 41% of teachers report feeling unprepared to use AI in their curriculum. That gap between adoption and readiness is exactly where stories like this come from — and exactly where training programs and AI literacy efforts can make the biggest difference.