LAUSR

This study examines the use of Bayesian Networks (BNs) for Stealth Assessment (SA) in digital game‐based learning (DGBL) environments. By integrating in‐game behavior tracking with embedded assessment scores, we investigate how interactive gameplay fosters learning gains. Data were collected from 632 middle school students participating in Mission HydroSci (MHS), a first-person 3D narrative adventure designed to teach water science and scientific argumentation. Using Static Bayesian Networks (SBNs), we modeled probabilistic dependencies among various in‐game behaviors, including evidence‐based argumentation, tool usage, dialogue engagement, and spatial exploration, and corresponding learning outcomes measured via pre‐ and post-assessments. Our analysis reveals distinct behavioral profiles strongly linked to positive learning gains. In particular, behavior patterns, including repeated engagement in argumentation tasks, strategic tool usage, and goal-oriented spatial exploration, emerge as key predictors of enhanced performance. Insights from the BN analysis inform the design of more effective DGBL environments and highlight the potential for real-time, adaptive assessment mechanisms that maintain gameplay immersion. Overall, this research offers a data-driven framework for understanding and optimizing learning trajectories in DGBL, providing practical guidelines for educators and game designers to enhance digital learning interventions.