Title

Training for vigilance: Using predictive power to evaluate feedback effectiveness

Authors

Authors

J. L. Szalma; P. A. Hancock; J. S. Warm; W. N. Dember;K. S. Parsons

Comments

Authors: contact us about adding a copy of your work at STARS@ucf.edu

Abbreviated Journal Title

Hum. Factors

Keywords

SIGNAL-DETECTION-THEORY; DECREMENT; TIME; Behavioral Sciences; Engineering, Industrial; Ergonomics; Psychology, ; Applied; Psychology

Abstract

Objective: We examined the effects of knowledge of results (KR) on vigilance accuracy and report the first use of positive and negative predictive power (PPP and NPP) to assess vigilance training effectiveness. Background: Training individuals to detect infrequent signals among a plethora of nonsignals is critical to success in many failure-intolerant monitoring technologies. KR has been widely used for vigilance training, but the effect of the schedule of KR presentation on accuracy has been neglected. Previous research on training for vigilance has used signal detection metrics or hits and false alarms. In this study diagnosticity measures were applied to augment traditional analytic methods. Method: We examined the effects of continuous KR and a partial-KR regimen versus a no-KR control on decision diagnosticity. Results: Signal detection theory (SDT) analysis indicated that KR induced conservatism in responding but did not enhance sensitivity. However, KR in both forms equally enhanced PPP while selectively impairing NPP. Conclusion: There is a trade-off in the effectiveness of KR in reducing false alarms and misses. Together, SDT and PPP/NPP measures provide a more complete portrait of performance effects. Application: PPP and NPP together provide another assessment technique for vigilance performance, and as additional diagnostic tools, these measures are potentially useful to the human factors community.

Journal Title

Human Factors

Volume

48

Issue/Number

4

Publication Date

1-1-2006

Document Type

Article

Language

English

First Page

682

Last Page

692

WOS Identifier

WOS:000243125300007

ISSN

0018-7208

Share

COinS