Adaptation asymmetry in manual tracking

A theory of manual adaptive control requires an understanding of the mechanisms by which human operators alter their control strategies in response to changes in control conditions. We show that significant asymmetries in adaptation occur in a manual tracking task when well-trained subjects are exposed to situations in which track preview, lag of the controlled system, or track bandwidth gradually change. We analyzed the experimental data using our recently developed recursive system identification algorithm which not only estimates non-stationary linear model coefficients but also time varying delay time. We show that adaptation strongly depends on whether the task becomes easier or more difficult as a result of changing conditions. Subjects were not always inclined to improve their strategy when tracking difficulty gradually decreases as a result of changing conditions. When the task gradually became more difficult, subjects seemed predisposed to maintain their current strategy rather than switching to a more effective strategy particularly when switching required an increased work load or attentional demand. The observed adaptation asymmetries indicate that tracking performance cannot be predicted based on momentary task conditions alone. To establish a model of adaptation mechanisms and its driving forces, the recent history of changes in tracking conditions needs to be considered. We attribute the observed asymmetries to a performance judgment process which triggers adaptation when perceived performance falls outside a subjective range that depends strongly on the recent history of tracking conditions