Version classiqueVersion mobile

Proceedings of the fourth Resilience Engineering Symposium

Erik Hollnagel
Éric Rigaud
Denis Besnard

Safety In Numbers? Using Graphs Or Gut Feelings When It Comes To Safety

Kyla Zimmermann (née Steele)


This paper will discuss the concept of quantifiable safety metrics and indicators, and draws empirical examples from the aviation industry.

Texte intégral

1Arguments questioning the value of quantitative safety measurement and accident classification using taxonomies are not new, however the practices persist in the aviation industry. During interviews conducted with aviation professionals in the context of a research project, interviewees made a distinction between safety statistics based on past events and “real-time”, dynamic safety (for an overview of the study, see Steele and Zimmermann publications in the 2006 and 2008 Resilience Conference proceedings). Several of the pilots, aircraft mechanics, and air traffic controllers interviewed explained that intellectually they know aviation is safe because the statistics say so, yet they discussed their qualitative perception of safety and risk-taking on the job as something entirely separate and unrelated to the quantitative, post hoc numbers. The numbers tell them that flying is safe, yet they feel anxious about some of their operational practices, regulations, and management decisions. These practitioners unwittingly illustrated a major issue facing safety experts today: Should we believe the graphs or our gut feelings when it comes to safety?

2One simple answer may be that safety is, in fact, decreasing, and the numbers just need to catch up in order to show this trend. Or there may in fact be a paradox here: For ultra-safe systems, such as scheduled air transport, society takes the safety for granted and thus the more sensationalised and reactive the response to an accident. This may increase the feelings of risk and fear among passengers as well as practitioners, for example, making them less tolerant of risks than in decades past.

3I will also present empirical examples from the arena of accident investigation and data collection which demonstrate the problems of trying to count and calculate “safety”. The issues encountered in practice range from the practical (e.g. the problems of gathering data from different sources in different formats) to the conceptual (e.g. how do we define “rate” in order to compare transportation modes). This discussion illustrates some fundamental resilience and human factors concepts such as “WYLFIWYF”. These dilemmas continue to plague accident investigators and safety regulators in an increasingly data-driven, quantitative world where “hard” data is required to justify changes.


Senior Human Factors Analyst, Transportation Safety Board of Canada

© Presses des Mines, 2011

Licence OpenEdition Books

Rechercher dans OpenEdition Search

Vous allez être redirigé vers OpenEdition Search