Cognitive Diagnostic Modeling from Open Educational Game Data: A Cross-Domain Analysis in Science Learning

• In Event: AERA Poster Session Wednesday 3:45pm
  In Poster Session: Modeling, Measurement, and Design in Technology-Enhanced Learning Environments

Abstract

This study addresses the need for interpretable assessments in game-based learning, moving beyond prediction to diagnostic inference. We implement enhanced Cognitive Diagnostic Modeling (CDM) using authentic log data from two science games: Wake: Tales from the Aqualab (marine science) and Antibiotic Resistance (microbiology). Analyzing 186,810 behavioral events from 240 students , we developed systematic, theory-driven Q-matrices and multi-factor scoring algorithms. Enhanced DINA models achieved strong diagnostic accuracy (WTA: 74.2%, AR: 71.8%) and discriminative power (AUC > 0.79). The framework also demonstrated good cross-domain transferability, validating a scalable method for generating fine-grained, diagnostic feedback from open-ended gameplay.

Author

• Jewoong Moon, University of Alabama