This project addresses the challenges of static, vague rubrics and the integration of Generative AI (GenAI) in writing assessment. It proposes a human-in-the-loop, multi-agent GenAI architecture that combines unsupervised clustering of authentic student writing with coordinated AI agents—analytic, retrieval, pedagogical, and ethical. This system facilitates the co-creation of rubrics between students and instructors, exemplar calibration, and guided reflection.

By treating rubrics as dialogic and evidence-based learning tools, this pedagogically grounded system enables students to compare their drafts with curated exemplars, engage in guided revision, and document their use of AI. Simultaneously, it provides instructors with analytics on how assessment criteria are interpreted and applied. The project is initially being implemented across six testbed courses in language communication, engineering, and linguistics, reaching over 700 students. Key courses include LANG1422 (Chinese for Workplace Applications), HUMA3030 (Language, Communication and Culture), and the CSE Individual Ethics Essay. Designed for scalability, RubriX’s core components are discipline-agnostic and intended for expansion into a wide range of writing assessments and disciplinary contexts, with the potential for institution-wide adoption.