CSCL Poster
#959: A Teacher-in-the-Loop Multi-Agent Framework for Transparent Essay Scoring and Feedback
Wed Jun 17, 4:15 PM–5:45 PM · Outdoors
Assessment, Feedback & Formative Practices Generative AI & Large Language Models AI in Education Human-AI Collaboration & Teaming
We present MAESF, a multi-agent essay scoring and feedback, as a teacher-centered AI tool for scoring and feedback. The system distributes evaluation across specialized agents and enables educators to interactively query and refine rubric-aligned assessments. Evaluated on the ASAP dataset, MAESF outperforms single-agent baselines and approaches human inter-rater reliability. By combining transparent scoring with teacher-in-the-loop interaction, MAESF reframes automated assessment as a collaborative process that augments teacher capacity and supports more consistent, timely feedback.
Speakers
- Ali Keramati — Microsoft
Authors
Ali Keramati, Mark Warschauer