Evaluating GenAI Feedback in Classroom Assessment: A Meta-Synthesis

• In Event: AERA Roundtable Session Friday 7:45 am - Gold 4
  In Roundtable Session: AI and Immersive Technologies: Exploring Emerging Tools for Assessment and Feedback

Abstract

Generative AI (GenAI) tools are increasingly used to provide feedback in classrooms, yet adoption has outpaced scrutiny of the feedback they produce. This study synthesizes 33 reviews to examine how GenAI feedback is characterized and whether it aligns with research-based principles of effective feedback. Accuracy—often assumed in feedback literature—cannot be presumed with GenAI, whose outputs are based on probabilistic prediction, not instructional intent. Before judging such feedback as effective, its trustworthiness must be foregrounded. As an implication of our findings, we propose ten educator-facing criteria to support feedback literacy. We also map two levels of decision-making—generation and uptake—and identify one concrete implication for the often-vague human-in-the-loop principle: equipping users to evaluate and respond to GenAI-generated feedback.

Authors

• Elie Chingyen Yu, University at Albany - SUNY

• Carla Evans, Center for Assessment