Date of Award

4-14-2021

Degree Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Computer Information Systems

First Advisor

JJ Po-an Hsieh

Second Advisor

Mark Keil

Abstract

Overreliance on automation often leads to problems such as human skill decay, a loss of situation awareness, and even casualties. Human-centered automation (HCA) is proposed to keep humans in the loop so that both humans and automation work on the same task. Effective task allocation between humans and automation has attracted a great deal of attention since the introduction of automation technologies, which assist humans in task performance. Yet, the rise of intelligent systems (IntelSys) that are capable of making task-allocation judgments has inspired an urgent inquiry: who (humans or IntelSys) should have the authority to make the decision to allocate tasks between humans and automation?

The current literature proposes three decision-making approaches (DMAs) to allocate tasks between humans and automation in HCA: IntelSys-Decides, Human-Decides, and Intel-Advises approaches. In the IntelSys-Decides approach, IntelSys has full decision-making authority to allocate tasks between humans and automation. In the Human-Decides approach, humans have the sole authority to make task allocation decisions. In the IntelSys-Advises approach, humans have the freedom to make task allocation decisions at any time; meanwhile, IntelSys generates task allocation advice to humans, who can either accept or reject the advice by IntelSys. Given the fact that there are no consistent findings about which approach is the best, it is necessary to identify the best approach, and if necessary, to identify the boundary conditions for the three DMAs.

Drawing on the perspective of team-based decision-making, this study proposes four hypotheses that compare the impacts of three decision-making approaches (DMAs) on human-automation team performance. To test the hypotheses, we conducted a large-scale experiment with 881 participants playing on a gaming platform. The results suggest that the effectiveness of DMAs is contingent on the task uncertainty and human expertise. The findings reveal insights on whether humans or IntelSys should assume the role of decision-maker in human-automation teams under different scenarios.

DOI

https://doi.org/10.57709/22867099

File Upload Confirmation

1

COinS