Dynamic Task Allocation: Issues for Implementing Adaptive Intelligent Automation

Abstract

How should tasks be allocated dynamically between people and intelligent machines? What are the initial issues? Previous work on static task allocation and work on human performance when multi-tasking and when interrupted provides suggestions on how to dynamically allocate tasks between humans and machines. We use these results to explore previous theories of task allocation. Some of these theories have direct suggestions for dynamic task allocation and some have indirect implications. We use both types to provide a list of suggestions for creating systems that do dynamic task allocation. The context we will be working with is a type of pilot’s associate that has a description of the pilot’s tasks and flight mission built within a cognitive architecture. The proposed associate has an additional component that can match the pilot’s performance to these tasks, predict the flight phase and pilot’s current tasks, and use this information to dynamically allocate these tasks between the pilot and the automation. These suggestions are to inform the design of a high-level intelligent controller.