Some Challenges in the Design of Human-Automation Interaction for Safety-Critical Systems

Increasing amounts of automation are being introduced to safety-critical domains. While the introduction of automation has led to an overall increase in reliability and improved safety, it has also introduced a class of failure modes, and new challenges in risk assessment for the new systems, particularly in the assessment of rare events resulting from complex inter-related factors. Designing successful human-automation systems is challenging, and the challenges go beyond good interface development (e.g., Roth, Malin, & Schreckenghost 1997; Christoffersen & Woods, 2002). Human-automation design is particularly challenging when the underlying automation technology generates behavior that is difficult for the user to anticipate or understand. These challenges have been recognized in several safety-critical domains, and have resulted in increased efforts to develop training, procedures, regulations and guidance material (CAST, 2008, IAEA, 2001, FAA, 2013, ICAO, 2012). This paper points to the continuing need for new methods to describe and characterize the operational environment within which new automation concepts are being presented. We will describe challenges to the successful development and evaluation of human–automation systems in safety-critical domains, and describe some approaches that could be used to address these challenges. We will draw from experience with the aviation, spaceflight and nuclear power domains.