emphasize the need for a PSM system to be simultaneously retrospective and prospective,, with incident investigation providing the vital bridge between the lessons of the past and safer designs and operation in the future (CCPS, 2003). This point is expanded upon in the following sections.

Relationship Between A Priori and Post-Facto Assessment

Although advance preparation is essential, incident investigations are conducted after the fact—that is, after a loss-producing event or a near-miss has occurred. In conducting a post-incident process assessment, it is important to avoid the problem of hindsight bias.. Hindsight bias, known commonly as “Monday morning quarterbacking” or “20-20 hindsight” is the tendency to view events as more foreseeable or more inevitable after the fact than they actually would have appeared at the time actions needed to be taken (Fischhoff, 1975; Blank et al., 2007; Louie et al., 2007). In particular, anyone who is judging the safety of a facility after an incident has information that was not available to those who conducted any pre-incident process assessment. Although most people recognize it would be unfair to use later information to second-guess earlier decisions, research on hindsight bias cited above has shown that cognitive biases can limit our ability to recognize the additional information that we have acquired after the event. While such new information should never be ignored, it is important to acknowledge that critical factors may not have been obvious before an incident, because this can help identify new opportunities for analysis, monitoring, communication, etc.

The chain of events that produced a chemical release is obvious in retrospect because it happened, even though it might not have been obvious in prospect because safety analysts failed to imagine that such an event chain could happen. In such cases, it is important to judge what the safety analysts could reasonably have been expected to anticipate by examining the safety analyses conducted in other facilities. If facilities with similar designs had also failed to anticipate that chain of events, then those conducting a pos-tincident process assessment should be wary of the effects of hindsight bias. However, if facilities with similar designs had anticipated that chain of events, then those conducting a post-incident process assessment should be less concerned that their analyses are being affected by hindsight bias.

Alternatively, it might be that the probability (rather than the possibility) of that chain of events might seem more likely in retrospect than in prospect because a pre-incident safety assessment underestimated the probability that such an event chain could happen. In this case, it is important to balance the possibility that the post-incident process assessment is being affected by hindsight bias against the possibility that any pre-incident process assessment was affected by optimistic bias (Weinstein, 1989), also known as comparative optimism (Klar and Ayal, 2004). In other words, the pre-incident process assessment might have

The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement