AI teaching systems raise new questions for school districts

AI shifts from teacher support to K-8 instruction

Early AI tools for education largely focused on supporting teachers or providing students with supplemental practice. A newer class of systems is designed to take on more of the instructional work, guiding students through lessons and responding to their individual needs.

Tracy Weeks, senior education policy and strategy director at Instructure, an education technology company, said AI tutors can provide students with help anytime, rather than only when a human tutor is available.

"Many are designed to act more like coaches than answer machines, prompting students to work through problems rather than simply providing answers," she said.

Earlier adaptive-learning systems largely relied on traditional predictive AI, while generative AI can produce new responses and interact more dynamically with students, said Rebecca Winthrop, senior fellow and director of the Center for Universal Education at the Brookings Institution, a non-profit public policy and research organization. But the more dynamic generative AI technology, with all its benefits and risks, also raises reliability concerns, especially for younger children.

"Younger children are less equipped to recognize or compensate for incorrect or misleading information when they're interacting directly with an AI system," Winthrop said.

Weeks said human oversight and AI literacy are especially important when AI systems are used with younger students. "At those earliest grades, we're probably not as likely to see a full-on AI tutor quite as much," she said.

Districts need evidence before scaling AI teaching systems

Personalized instruction and instant feedback are not enough for districts to know whether an AI system is improving student learning.

Michael Horn, co-founder of the Christensen Institute and lecturer at the Harvard Graduate School of Education, said districts should be clear about how an AI system fits into their learning model and what they want it to accomplish.

"They should also examine its guardrails, data use and privacy protections," he said.

He suggested outcomes-based contracts with AI education vendors, tying payments to whether students achieve agreed-upon learning milestones. "If I'm the customer, I want to make sure I'm paying for the results I'm looking for, not just for the presence of the tool," he said.

Winthrop said districts should apply an evidence standard to AI systems similar to what they use for other educational interventions, including external evaluations that test the systems in settings similar to their own. She also added that researchers have conducted relatively few external evaluations of generative AI interventions in school districts.

Districts should look for external, third-party research showing that an AI teaching system has a positive impact before deploying it widely, according to Horn. "That requires districts to look beyond usage rates or time saved and consider whether the technology is improving outcomes such as student performance and comprehension," he said. 

AI tutors require more than a classroom rollout

AI teaching systems don't necessarily mean schools are preparing to replace teachers. Instead, the technology could shift teachers toward more individualized and social aspects of instruction.

Horn said teachers could spend more time understanding what motivates students, providing individual explanations when students get stuck and facilitating group and experiential learning. That could mean spending less time on traditional whole-class instruction.

"Ideally, teacher lesson planning and delivering lectures to an entire class would go away," he said.

That change in teachers' roles also creates new implementation challenges for school districts. Weeks said teachers need training before students use AI because educators are still developing best practices. Instructure's recent research found that 68% of K-12 educators reported using AI at least occasionally, while 45% said they had received no formal AI training.

For districts, adopting an AI teaching system isn't simply a technology decision. It can change how instruction is delivered, how teachers spend their time and how schools measure student progress.

But the bigger question is how much of that work districts are prepared to turn over to AI and what role they want teachers to play as they do.

Kinza Yasar covers AI and emerging technology for TechTarget, with a focus on ethics, enterprise adoption, governance and business strategy. Before moving into journalism, she worked in IT and network support roles, giving her a systems-level perspective on how enterprise technologies are built, deployed and managed.