Date of Award

Summer 2019

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Psychology

Committee Director

James P. Bliss

Committee Member

Jing Chen

Committee Member

Matt R. Judah

Abstract

Because automation use is common in many domains, understanding how to design it to optimize human-automation system performance is vital. Well-calibrated trust ensures good performance when using imperfect automation. Two factors that may jointly affect trust calibration are automation transparency and perceived reliability. Transparency information that explains automated processes and analyses to the operator may help the operator choose appropriate times to shed task control to automation. Because operator trust is positively correlated with automation use, behaviors such as task shedding to automation can indicate the presence of trust. This study used a 2 (reliability; between) × 3 (transparency; within) split-plot design to study the effects that reliability and amount of transparency information have on operators’ subjective trust and task shedding behaviors. Results showed a significant effect of reliability on trust, in which high reliability resulted in more trust. There was no effect of transparency on trust. There was no effect of either reliability or transparency on task shedding frequency or time to task shed. This may be due to high workload of the primary task, restricting participants’ ability to utilize transparency information beyond the automation recommendation. Another influence on these findings was participant hesitance to task shed which could have influenced behavior regardless of automation reliability. These findings contribute to the understanding of automation trust and operator task shedding behavior. Consistent with literature, reliability increased trust. However, there was no effect of transparency, demonstrating the complexity of the relationship between transparency and trust. Participants demonstrated a bias to retain personal control, even with highly reliable automation and at the cost of time-out errors. Future research should examine the relationship between workload and transparency and the influence of task importance on task shedding.

DOI

10.25777/6qtr-rm13

ORCID

0000-0002-6711-3179

Share

COinS