Stanford review finds mixed learning results in education AI research

A Stanford review published November 10 found mixed effects on student learning from artificial intelligence and promise for some teacher tasks, while identifying possible unintended effects on students.

The November 10 review synthesizes 52 selected causal papers, adding 32 to the 20 included in an earlier review. Its authors considered randomized and quasi-experimental studies covering different tools, populations and outcomes.

According to the review, AI use aimed at understanding and practicing material appears more promising for learning than using it to produce students’ work, though neither approach consistently improves learning. For educators, the researchers found that some tools can save time on lesson planning and support feedback. They also describe research in which a teacher-facing tool reduced student motivation and confidence.

The authors caution that the evidence should not be broadly generalized across U.S. K–12 settings. Many papers concern other countries or postsecondary education, and the review synthesizes findings rather than pooling them into a single effect estimate.

Review author Susanna Loeb told Chalkbeat’s Erica Meltzer that potential benefits depend on thoughtful use and tracking whether students benefit. She cautioned against adopting the tools without the capacity to integrate them carefully.