Case study · 01
MACE
Multi-Agent Course Evaluator
MACE supports course evaluation through structured evidence, responsibility-focused multi-agent modules, deterministic scoring, optional LLM-assisted analysis, and human evaluator review.
- Multi-Agent Systems
- AI Evaluation
- LLMs
Problem
Course evaluation draws on varied evidence, including learning materials, activities, assessment design, and educational criteria. Reviewing that evidence consistently takes time, while a final score alone cannot explain how a conclusion was reached.
Decision-support approach
MACE separates evaluation into responsibility-focused modules. The system structures course evidence, applies deterministic scoring, and can use LLM-assisted analysis where appropriate, while keeping human evaluator review central to the workflow.
- Evidence-based course analysis
- Responsibility-focused multi-agent modules
- Deterministic scoring
- Optional LLM-assisted analysis
- Human evaluator review
Evaluation priorities
The project prioritizes grounded outputs, transparent evidence use, and reviewable conclusions. It is designed to support academic judgment, not make autonomous academic decisions.
Status and limitations
MACE is an active engineering and research project. Public quantitative results are not reported because no verified metrics are currently approved for publication.
Related research
Multi-Agent Systems · AI Evaluation · Decision Support