A Complexity Approach to Unravelling Traces of Scientific Reasoning Behaviours During Game‐Based Learning

Background Scientific reasoning is central to science learning, yet most existing research measures scientific reasoning with static assessments or aggregated counts that overlook how scientific reasoning unfolds dynamically over time. Game‐based learning environments provide the opportunity for rich behavioural log‐file traces of learners' actions to be captured, but current analytic approaches often fail to reveal how the temporal organization of in‐game actions emerges and reorganizes to form scientific reasoning behaviours during learning. Understanding these behavioural dynamics is critical for characterizing productive learning trajectories and for informing adaptive support for effective scientific reasoning within interactive learning systems that require this skillset for science learning. Objectives This study examined whether and how the temporal emergence of behavioural complexity in learners' scientific reasoning relates to scientific reasoning success and domain learning outcomes during game‐based learning. Methods High‐school students ( N = 170) engaged with Crystal Island, a microbiology game‐based learning environment requiring scientific reasoning to solve a mystery and complete the primary game objective. Guided by complex systems theory, we applied cumulative auto‐recurrence quantification analysis to learners' log‐file data to calculate entropy and model how the structural complexity of in‐game actions evolved over time to demonstrate effective scientific reasoning. Relationships between entropy trajectories, game success and learning gains were examined using nonlinear growth modelling. Results and Conclusions Findings revealed that the timing and rate of complexity growth were associated with learning outcomes. Learners who successfully completed the game demonstrated more structured and gradual increases in behavioural complexity, whereas greater learning gains were associated with steeper increases in entropy. These results suggest that behavioural entropy is neither inherently beneficial nor detrimental; rather, the temporal organization of in‐game actions can provide meaningful insight into how scientific reasoning emerges and offers potential avenues for integrating adaptive support in game‐based learning environments.