The Neglected Context of Risk Assessment A Mindset for Method Choice

摘要

During World War II, Lancaster bombers deployed by the Royal Air Force were being shot down increasingly often because German air defense aircraft were continuously being improved. When scientists were asked why the casualty rate was so high, they concluded that the Lancasters were vulnerable because they lacked speed and maneuverability. It was recommended that gun turrets be removed to make the aircraft lighter. Military authorities, however, thought that guns were good and more guns were better, so they added guns and gunners, which slowed the aircraft even more, which led to even more casualties. A bomber without gun turrets was inconceivable (Sagan, 1993). A bomber without gun turrets is like a risk assessment without formal modeling; it is inconceivable to insiders. When I was asked to comment on methods of risk assessment in maritime risk mitigation, the invitation came with the stipulation that it would help if my remarks encouraged people to think "outside the box" on the question of how to do risk assessments. The formal methods of risk assessment that are now common in the maritime industry appeal to the heart of the engineer that lurks in many maritime personnel. But those formal methods are also blunt instruments. They give the misleading impression that risk is well understood, fully mapped, and that, if it weren't for operator error, the maritime system would function reliably. What formal methods miss is the situated nature of risk taking. Formal methods are less sensitive to local contingencies, subunit norms, informal agreements, idiosyncratic labels and language, tricks of the trade, strong local cultures, emergent changes, unintended consequences, sudden opportunities, resourceful improvisation, and unexpected setbacks. Local variations such as these shape most risk scenarios even though these determinants go undetected. Their presence is not just noise. It is often more patterned and more predictable than people imagine and more tied to personal and organizational factors than people are willing to admit. To incorporate more of these factors requires a return to some of the basics in inquiry with the question, How has this issue been handled by people currently doing maritime risk assessments? In many cases the answer is, Not very well. Those lapses in procedures of inquiry stand in the way of more effective risk mitigation. If modelers make more of an effort to address problems such as those I will mention, then the adequacy of their database will improve, as will the lessons that practitioners are able to draw from these data. Modeling doesn't need more analytics any more than the Lancasters need more guns. It needs different analytics and nonanalytics. To think outside the box is to take that diagnosis seriously.