AI in the feedback process: Effects of humanlike AI agents vs. human feedback on student experiences

This study examines artificial intelligence (AI) agents as a support or alternative to traditional teacher feedback in higher education, aiming to address the challenge of limited teacher resources in providing expert feedback. Recent advances in Large Language Models (LLMs), enable personalized, scalable, and instant feedback opportunities, but feedback is studied as a complex, dialogical process influenced by the feedback provider’s and recipient’s contexts, including their feedback literacy, domain knowledge, interpersonal relationship, and modality of interaction. <br/>In this experimental study, students at the bachelor’s level in Educational Science, Pyschology, and Cognitive Science were divided into 2 groups that received feedback in a digital learning environment from either a human teacher (CG) or an AI avatar resembling the human teacher (IG). Screen recordings and post-test interviews were analysed to assess students’ practice and experiences. <br/>Preliminary results suggest differences concerning students' feedback strategies and emotional responses to the different feedback agents. Human feedback evoked feelings of nervousness, whereas AI agents evoked mostly positive feelings while participants also more actively engaged in soliciting specific feedback. The spoken dialogue of the human and avatar groups further forced participants to focus on more conceptual feedback. By comparing AI feedback to traditional human feedback, this study seeks to understand how personalized feedback can be effectively scaled and how students' perceptions of AI influence their feedback experience. This study contributes to ongoing discussions about designing AI-enhanced feedback processes and understanding AI-human interaction in general.<br/>