An experimental study of dialogue-based communication for dynamic human-computer task allocation

Abstract

The allocation of tasks between human and computer and the merits of a dynamic approach to this allocation are discussed. Dynamic task allocation requires efficient human-computer communication. This communication may be accomplished in an implicit, model-based, or explicit, dialogue-based, manner. A framework for the study of dialogue-based human-computer communication is introduced and a study exemplifying the use of the framework is presented. This study investigated the effects of two input media and four task allocation strategies on the performance of a human-computer system. The task environment represented a simplified version of an air traffic control scenario wherein computer aid could be evoked by the human controller to accomplish task sharing between the human and the computer. Dedicated function keys proved to be a more effective input medium than the standard Sholes QWERTY keyboard in terms of both objective performance and subjective preference measures. Of the task allocation strategies considered, spatial assignment, contingency-based assignment, and assignment by designation achieved the highest levels of overall system performance, while temporal assignment achieved a significantly lower level of performance. Subjective ratings indicated an overall preference for assignment by designation, followed by spatial assignment and contingency-based assignment. Spatial assignment was the most powerful, but the least specific strategy. Assignment by designation was the least powerful strategy, but the most specific and most flexible strategy.