Citation

Nikolic MI, Sarter NB. Hum. Factors 2007; 49(4): 553-563.

Affiliation

The Boeing Company, PO Box 3707, Seattle, WA 98124-2207, USA. mark.i.nikolic@boeing.com

Copyright

(Copyright © 2007, Human Factors and Ergonomics Society, Publisher SAGE Publishing)

DOI

unavailable

PMID

17702208

Abstract

OBJECTIVE: To examine operator strategies for diagnosing and recovering from errors and disturbances as well as the impact of automation design and time pressure on these processes. BACKGROUND: Considerable efforts have been directed at error prevention through training and design. However, because errors cannot be eliminated completely, their detection, diagnosis, and recovery must also be supported. Research has focused almost exclusively on error detection. Little is known about error diagnosis and recovery, especially in the context of event-driven tasks and domains. METHOD: With a confederate pilot, 12 airline pilots flew a 1-hr simulator scenario that involved three challenging automation-related tasks and events that were likely to produce erroneous actions or assessments. Behavioral data were compared with a canonical path to examine pilots' error and disturbance management strategies. Debriefings were conducted to probe pilots' system knowledge. RESULTS: Pilots seldom followed the canonical path to cope with the scenario events. Detection of a disturbance was often delayed. Diagnostic episodes were rare because of pilots' knowledge gaps and time criticality. In many cases, generic inefficient recovery strategies were observed, and pilots relied on high levels of automation to manage the consequences of an error. CONCLUSION: Our findings describe and explain the nature and shortcomings of pilots' error management activities. They highlight the need for improved automation training and design to achieve more timely detection, accurate explanation, and effective recovery from errors and disturbances. APPLICATION: Our findings can inform the design of tools and techniques that support disturbance management in various complex, event-driven environments.

Language: en