Assisting Teachers with Artificial Intelligence: Investigating the Role of Teachers Using a Randomized Field Experiment

Abstract

This study investigates whether artificial intelligence (AI) can transform the teacher’s role by delivering personalized learning to each individual student. Conducting a randomized controlled trial in collaboration with an education company, we evaluate how providing teachers with AI assistance impacts students’ academic outcomes. We find that providing AI-generated reports to teachers significantly improves students’ study effort and performance, although those effects vary by teacher and class characteristics. Consistent with the organizational literature on technology and worker productivity, our findings indicate that technology overload could undermine teachers’ effective use of AI coaching despite its highly accurate diagnostic ability and ready availability. Although some teachers did not utilize the AI coaching program, we found positive effects among the students of both compliers and non-compliers. We, therefore, decompose our results to identify direct and spillover effects on students’ academic performance and show that failing to account for spillover effects across teachers (externality among peers) and within teachers (learning-by-doing across students) may understate the effects of AI coaching. Finally, we provide practical guidelines for implementing technology in educational settings.