Human-Labeled Feedback Signals More Investment: Explaining Differential Perceptions of AI vs Human Instructor Feedback

The ubiquity of AI-generated feedback in education raises fundamental questions about how the source of feedback shapes learners’ processing and engagement with it. Evidence is emerging that AI feedback might elicit distinct psychological reactions differing from those elicited by hu-man feedback. However, the mechanisms behind empirically found differences are not yet well understood. The present study adds to this by examining perceptions of investment and judgment associated with social aspects of feedback in addition to feedback usefulness and actionability. We report experimental findings using a fake label paradigm, where the same feedback message is labeled as either coming from an AI system or a human instructor, in a sample of 282 US col-lege students on Prolific. We test the effects of provider labels on feedback perceptions and be-havior (revision engagement). Results show that students perceive the same feedback message differently: students report higher investment perceptions, personal relevance, actionability, and usefulness of the feedback when it is labeled as coming from a human instructor. We did not find significant differences in judgment perceptions and behavioral engagement regarding the feed-back and revision task. Further analyses indicated that investment perceptions, but not covariates such as AI competence perceptions, explain the differences between feedback conditions regard-ing usefulness and actionability. For theory, this implies that perceived investment might be one key relational appraisal through which provider labels (human vs. AI) shape perceived useful-ness and actionability, even with identical feedback content. Additionally, we explore the practi-cal implications for effectively implementing AI-generated feedback in the classroom.