Date of Award

Winter 2014

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Psychology

Committee Director

J. Christopher Brill

Committee Member

James P. Bliss

Committee Member

Poornima Madhavan

Abstract

Automated signaling systems are frequently used to direct operator attention to potential hazards. Although these automated systems can lead to enhanced human performance, factors such as degraded alarm signal reliability and lack of trust can undermine the potential benefits of automation (Breznitz, 1984; Rice, 2009, Wickens & 2007). Additionally, work by Gilson, Mouloua, Graft, and McDonald (2001), as well as Keller and Rice (2009), suggest that an alarm contained within a larger array of alarms should not be evaluated individually. Due to the increasing use of multiple alarms in complex environments such as operating rooms and cockpits (Konkani, Oakley, & Bauld, 2012; Woods, Suter, & Billings, 1997), it is important to identify reaction strategies that may and should be used when an unreliable alarm is in the presence of other alarms. Accordingly, the influence of reliability level and the number of additional activated alarms on objective trust, reaction time, and acceptance rate with a 12-alarm array was evaluated using a 2 x 12 split-plot factorial design. Overall a significant linear trend was observed in objective trust measures as the number of additional activated alarms (p < .001). This finding indicates the number of additional activated alarms, instead of the given alarm reliability, was used to calibrate objective trust. Reaction time was found to be quadratic (p < .001). Acceptance rate followed a cubic trend (p < .001), with significant quadratic (p = .02) and significant linear (p < .001) derivative trends. This suggests participant response changed from alarm dismissal to acceptance near 50% of alarm array activation. Finally, there was a significant effect of reliability level (p < .001) on acceptance rate, although no significance differences were found between the 50% and 75% groups. Overall, the results constitute evidence for an extension of probability matching theory based on percent system activation and indicate the need to minimize alarms in display design.

DOI

10.25777/r37q-hw71

ISBN

9781321564662

Share

COinS