randomized field experiment

Editorial cover of undergraduate learners and a lecturer examining a course-grounded RAG chatbot alongside flat learning and motivation outcome traces

Journal Paper2026

AI chatbots in higher education: Comparing expectations to evidence

Andrew Thoeni, Luke K. Fryer

Computers in Human Behavior Reports

A semester-long randomized field experiment with 454 undergraduates found that access to a course-grounded RAG chatbot did not significantly improve interest, self-efficacy, engagement, or test performance, despite students reporting that they liked the tool.

RAG chatbotrandomized field experimenthigher education

Read 500-word summary →