The Field Guide › Paper
One of the earliest and most widely taught applications of Edmondson's psychological safety construct beyond conventional organisational settings, and among the first to set it alongside behavioural decision theory and normal accident theory in a single integrated analysis. Roberto examines the 1996 Everest disaster, in which expedition leaders Rob Hall and Scott Fischer and three clients died during descent, through three levels he insists are complementary rather than rival explanations. At the individual level, cognitive biases impaired judgement: sunk cost escalation (clients who had spent up to $70,000 and weeks of suffering found it nearly impossible to turn around short of the summit; only four did), overconfidence in both expedition leaders, and a recency effect by which several seasons of benign weather led experienced guides to underestimate a storm that was, historically, entirely normal for the mountain. At the group level, the antecedents Edmondson identified for team psychological safety were systematically absent: steep status differences (a guide-client protocol under which clients had been conditioned not to question guides' judgement, and a $15,000 pay differential between guides that Beidleman later identified as part of why he held back his serious reservations about climbing past midday), leader behaviour (Hall announced at base camp that 'my word will be absolute law, beyond appeal'), and near-total unfamiliarity between team members, most of whom had never met before the expedition and so had no basis for the trust that disagreement requires. At the system level, Perrow's complex interactions and tight coupling: a customs delay in Russia, another expedition's misplaced fixed ropes, a failed negotiation with Outside magazine, and a rigid eighteen-hour oxygen-limited summit schedule interacted so that individually trivial failures cascaded. The integrative claim is the paper's lasting contribution: an absence of psychological safety makes cognitive bias harder to catch, because nobody tests anyone's assumptions; unchallenged bias becomes most dangerous in tightly coupled systems, where one error triggers others; and complexity in turn raises the cost of every silence. The three levels are mutually reinforcing concepts, not competing explanations, and Roberto closes by warning that attributing such failures to individual human error is itself a bias, one that conveniently persuades ambitious observers they would have done better.