Cognitive load-based multi-level annotation model for knowledge acquisition in heritage games

This study proposes a CLT-grounded, three-tier interactive annotation model to mitigate cognitive overload and enhance knowledge acquisition in cultural-heritage serious games. Using the Shimao Ruins virtual platform, we evaluated behavior logs (interaction frequency, completion time), knowledge tests (immediate/delayed), and user feedback. The experimental group outperformed the control group in short-term recall (84.7% vs 64.6%) and long-term retention (72.3% vs 54.1%). Regression showed interaction frequency positively predicted learning (β = 0.87, p < 0.001), whereas task duration negatively correlated with performance (β = −0.29, p = 0.028). The model reduces extraneous load while fostering germane processing through progressive tasks and information granularity. Results support broader applicability and point to future directions in mobile AR/VR integration and adaptive difficulty for real-time load regulation.