A comparison of human and near-optimal task management behavior

The primary contribution of this work is the development of an abstract framework to which a variety of multitasking scenarios can be mapped. The metaphor of a juggler spinning plates was introduced to represent an operator performing multiple concurrent tasks.

This allowed seeking a quantitative model for management of multiple continuous tasks instead of a model for completing multiple discrete tasks, which was considered in previous studies.

The multitasking performance of 10 participants in five scenarios was measured in a low-fidelity simulator (named Tardast), which was developed based on the concept of the juggler metaphor. This performance was then compared with a normative model, which was a near-optimal solution to a mathematical programming problem found by tabu search heuristics.

Tabu outperformed the participants overall, although the best individual performance nearly equaled that of tabu. It was also observed that participants initially tended to manage numerous tasks poorly but that they gradually learned to handle fewer tasks and excel in them.

This suggests that they initially overreacted to the penalization associated with poor performance in the software. Participants' strategic task management (e.g., what tasks to handle) was more significant in obtaining a good score than their tactical task management (e.g., how often to switch between two tasks).

Potential applications include better design of equipment, procedures, and training of operators of complex systems.