Related Experiment Videos
CoMAD: a transferable method for building cognition-aware mathematics assessment datasets for educational AI research
Tibakanya Joseph1, Joyce Nakatumba-Nabende1, Agaba Ezra Joab2
1Department of Computer Science, Makerere University, P. O. Box 7026, Kampala, Uganda.
Methodsx
|August 9, 2026
Summary
We introduce a new framework for creating cognitively aware, curriculum-aligned datasets for AI. This method ensures transparency and reusability, supporting automated item generation and AI model alignment.
Area of Science:
- Artificial Intelligence
- Educational Data Mining
- Cognitive Science
Background:
- Limited availability of cognitively annotated, contextually grounded, and curriculum-aligned assessment datasets.
- Challenges in integrating educational taxonomies, linguistic complexity, and FAIR data principles in dataset construction.
Purpose of the Study:
- To present a transferable, FAIR-oriented framework (CoMAD) for creating cognition-aware, curriculum-aligned datasets.
- To address the need for robust datasets supporting automated item generation and AI model alignment.
Main Methods:
- Developed the Cognition-Aware Mathematics Assessment Dataset construction (CoMAD) framework.
- Utilized computational data engineering and integrated educational taxonomies (Bloom's, Webb's DOK, Hess's CRM).
- Validated the framework with the HiCogMath mathematics dataset.
Main Results:
- CoMAD provides a modular, scalable, and transferable method for dataset creation.
- The framework ensures transparent quality assurance, enhancing dataset reliability and reusability.
- Demonstrated suitability for cognitive modeling, data mining, assessment analytics, and LLM alignment.
Conclusions:
- CoMAD offers a robust solution for constructing high-quality, cognition-aware assessment datasets.
- The framework's modularity and transferability make it adaptable to various subjects and curricula.
- Facilitates advancements in AI-driven educational assessment and personalized learning.