Problem & context
Problem
Provide university students with scalable practice in linear algebra and differential calculus while preserving mathematical correctness, explaining errors and adapting difficulty from recent evidence and mastery.
Intended users
University mathematics students, instructors, researchers and administrators conducting controlled evaluations of automated feedback.
Role and contributions
Research design, mathematical content, adaptive policy, grading architecture and full-stack implementation
Dataset / inputs
The platform currently structures two courses with 18 seeded topics, generated mathematical problems, open-ended student answers and reasoning, mastery histories, misconceptions, remediation activity and research-study events.
Methods & architecture
A layered grading pipeline combines deterministic computer algebra with AI diagnosis and pedagogical criteria. An evidence-weighted policy updates mastery, applies time decay, selects easy, medium or hard tasks from recent performance, and prioritises misconceptions. Research infrastructure supports condition assignment, frozen feedback pools and pre-, post- and delayed quizzes.
Implementation
- Build responsive course, topic, practice, quiz, examination and dashboard flows
- Render mathematical notation with KaTeX and grade supported answers deterministically
- Generate varied problems across linear algebra and differential calculus
- Provide AI-assisted diagnosis, feedback and guided mistake correction
- Track mastery, misconceptions, remediation and research events in Convex
- Add role-based administration, consent controls and technical-validation tooling
Current evidence
The current implementation verifies the project foundation below; broader outcome claims will be added after evaluation.
Limitations & next steps
The current test run verifies the newer local research working tree, which contains substantial work not yet published to GitHub. External AI behaviour, authenticated database integrations and student-learning outcomes require separate controlled validation.
Complete isolated integration tests, freeze a reproducible study release, run the planned controlled evaluation and report learning outcomes separately from technical validation.
Attribution & licence
Copyright 2025 Motognon Wastalas d'Assise Dogbalou. All rights reserved, according to the local project README.