The Field Guide › Paper
Havinga, De Boer, Rae and Dekker take the aviation success story of crew resource management (CRM), the team and non-technical-skills training credited with helping make flying safe, and ask a deceptively simple question: what actually happened when other high-hazard industries adopted it? Their systematic review examines how CRM has been conceptualised and evaluated in maritime, nuclear power, oil and gas, and air-traffic-control settings, and the answer is unflattering to the assumption that CRM is a portable, proven intervention. Across the industries there is broad agreement on the goals (safety and efficiency) but wide disagreement, as much within each industry as between them, about what CRM actually is and how it is supposed to work, so that the same three-letter label covers a grab-bag of quite different programmes. Worse, the evaluation is weak: most studies fail basic methodological tests, and, tellingly, the way CRM is evaluated does not line up with how it is conceptualised, leaving most of what people claim about it untested. Of the handful of evaluations that meet a reasonable quality bar, CRM shows a clear effect less than half the time. For a corpus about psychological safety two things make this paper valuable beyond its safety-training subject. First, it is a clean worked example of the measurement-critique this map returns to repeatedly: an intervention whose evaluation measures are misaligned with its own theory of change, so that the evidence cannot actually answer whether it works. And second, in a striking aside the authors suggest that where CRM does change behaviour, the mechanism may not be the specific attitudes and skills it teaches at all but a single overarching variable, a psychological safety climate, raising the possibility that the active ingredient in a celebrated training programme is the very thing this map is about. Its limits are those of a review of a messy literature: it can only work with what was published and described, and it diagnoses the confusion more sharply than it resolves it, closing with a call for researchers to say plainly what they think CRM changes and to measure that. (Text drawn from the 2017 Safety paper, 3(4), art. 26.)