The Power of Performance and Physiological State: Approaches and Considerations in Adaptive Game-Based Simulation

Adaptive training provides the capability to enable individualized learning and keep users in the Zone of Proximal Development. However, when an individual should adapt is still an open area of research. This paper discusses findings from a recent driving-based simulation in which trainee state was assessed by multiple means. Modalities of measurement included real-time physiological information (heart rate and fNIR data) used to assess mental workload, subjective ratings of workload from participants after each trial (NASA TLX), and performance data on key components related to the execution of the scenario tasks. These methods were used in tandem to adapt the difficulty of the simulation environment. Results show that measures with high sample rates, such as average miles per hour and reported relative oxygenated hemoglobin (HbO2) readings, outperformed less temporally sensitive measures, such as task response correct percentages or NASA TLX ratings. These findings emphasize that the granularity of the measurement method should align with the system’s desired sensitivity in adapting to mental workload. More granular measures should be used when needing to adapt more frequently and/or to smaller changes in workload. Empirically supported methods that lack high temporal or spatial resolution are unlikely to suffice for applications that require rapid adaptation.