Adaptive Task Allocation in Human-Machine Teams with Trust and Workload Cognitive Models

In mixed-initiative systems where teams of humans and automated agents collaborate to perform decision-making tasks, determining factors of joint performance include human cognitive workload and the level of trust placed by the operators in the automation. Both workload and trust are dynamic variables that change over time based on current task allocation and on the result of past interactions. In this paper, we propose a methodology leveraging quantitative models of trust and workload to automatically and dynamically suggest efficient task allocations in mixed human-machine systems. Our approach is based on a Markov decision process framework and is presented for concreteness in the context of a human-machine team performing repeated binary decision-making tasks. Simulation results show the emergence of interesting automation behaviors such as seeking trust, attempting to repair trust after an error and adjusting human workload for optimal performance. Overall, the human-aware dynamic task allocation strategy shows the potential of significant team performance improvement compared to a static task distribution, even in the presence of significant errors in the trust and workload models used.