PeerChat: Scaling Peer Tutoring with AI Tutees

Ojas Karnavat, Soren Rosier

Track: Innovations in AI for Education (Day 1)

Paper Length: short-paper (2 pages + references)

Keywords: Peer Tutoring, Generative Artificial Intelligence, Education Technology

TL;DR: This paper introduces PeerChat, an AI tutee powered by ChatGPT, to address limitations in scaling peer tutoring by enabling students to practice teaching independently outside the classroom.

Abstract: This paper explores the innovative use of generative arti-ficial intelligence (AI) as a tutee in peer tutoring ses-sions, addressing limitations in scaling peer tutoring. Leveraging OpenAI's ChatGPT models, we introduce PeerChat, an AI tutee designed to facilitate student-led tutoring interactions. Peer tutoring, known to benefit both tutors and tutees, is limited to times when a teacher decides to incorporate peer tutoring into their lesson plan and thereby carries high stakes since the tutor is re-sponsible for another student’s learning. By employing generative AI to build a realistic AI tutee in one-on-one peer tutoring sessions, students gain opportunities for independent teaching practice, addressing these limita-tions. Our study, the first of its kind, systematically in-vestigates the creation of an AI tutee and evaluates its performance against predefined criteria. Results indicate promising accuracy, with successive models improving response quality. This novel approach holds the poten-tial to enhance peer tutoring experiences and opens the door to further research in optimizing AI tutees for edu-cation.

Submission Number: 65