Date of Award

5-2019

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Psychology

Committee Member

Richard Pak, Committee Chair

Committee Member

Kelly Caine

Committee Member

Ewart de Visser

Abstract

Trust is a critical factor in successful and productive human-automation interactions. When automation malfunctions, trust is negatively affected. The development of increasingly complex multiple-component systems, or those with a several autonomous elements, introduces even more ways for a system to err. One example is in smart home control systems where different subsystems may be controlled by different autonomous routines or rules. Multiple studies suggest that one error-prone component can lower user trust in the remaining components (the “pull down” effect). Other research suggests that certain types of information, when presented to the user, can reduce the strength of the pull-down effect by promoting heterogeneity of agents. The current study investigated the effectiveness of increasing the number of voiced agents within a system as a strategy for decreasing the strength of the pull down effect. Participants interacted with either a single- or four-agent system. A simulated smart home task required participants to adjust the lighting for several rooms of a house. Participants first completed a block with all reliable room lightings, and then a block with all but one reliable room lighting. Inconsistent with the current literature, the results did not reveal any pull down effect. In both agent conditions the presence of the unreliable room lighting did not decrease trust in the reliable room lightings. In the single-agent condition trust in the reliable room lightings increased between both reliability blocks. However, this trend was not seen with the four-agent condition. Future studies should investigate the effects of anthropomorphism, automation domain, and task characteristics on trust.

Share

COinS