The Effects of Task Partitioning and Computer Skill Level on Operator Engagement and Performance in an Adaptive Biocybernetic Environment

The present study examined task partitioning in an adaptive environment, utilizing the cybernetic closed-loop system of Pope et al. (1995). In that study, participants either performed a tracking task (manual condition) or monitored the tracking task (automatic condition). Changes in task mode were driven by an engagement index derived from the participant's EEG and operated under either positive or negative feedback conditions. One goal of the present study was to examine participant performance under a wider range of tracking conditions. Specifically, the task was partitioned into 3 modes: a manual mode, a partitioned mode, and an automatic mode. A second objective of the present study was to examine the effects of different skill levels of a computer teammate within the closed-loop system. An index of operator engagement was used as a basis for switching the compensatory tracking task among the three task modes. These three modes were combined factorially with two feedback. conditions (positive and negative) and two "skill" levels (expert and novice). Performance was assessed by the level of operator engagement index as well as RMSE on the tracking task. In addition, participants provided estimates of subjective workload on the NASA-TLX. Results of the present study showed that more task allocations occurred under negative feedback than positive feedback. Also, a feedback by task mode interaction for mean engagement showed that the biocybernetics system responded exactly as expected in a multi-level situation, supporting Pope et al. (1994, 1995). Participants paired with an expert-level computer outperformed those paired with a novice-level, showing that the effect of skill level of a computer teammate applies in an adaptive environment. Workload estimates, however, did not show significant differences across skill levels. Tracking performance in the partitioned mode may have been affected by the brief amount of time participants had to acclimate to that mode. Future changes to the cybernetic system may require a damping factor to limit the frequency of task allocations, or the use of a wider range around the baseline mean for the switch criterion.