The Premortem

What it is

The inversion is the whole trick. "What could go wrong?" is answered defensively by a group that has already converged on a decision. "It failed, and here is why" is answered as a matter of narrative competence, and the same people who would not raise an objection will happily explain a fictional catastrophe.

Daniel Kahneman's account of why it works gives two mechanisms, and both are social rather than cognitive. It legitimises doubt, because in a team that has reached agreement dissent reads as disloyalty, and the exercise makes doubt the assigned task so that raising one costs nothing. And it conscripts the supporters, since the people who back the plan know it best, and the exercise turns them into searchers for threats to it rather than defenders of it. The leverage is therefore not in making an individual think better. It is in changing what a group is permitted to say, which is the same shape as the general finding that organisational debiasing works where individual debiasing barely does.

The lineage behind it can be checked end to end, which is unusual. Mitchell, Russo and Pennington found that asking people to imagine an event has already happened improves their ability to generate explanations for it. That is the mechanism the exercise runs on, and Klein names it.

In effect

Kahneman's hedges must travel with the procedure, because they keep getting stripped off in secondary coverage. He calls it not a panacea, says it does not provide complete protection against nasty surprises, and on training people out of overconfidence in general writes that he is not optimistic and reports only a few cases of modest success.

The concept is also a clean instance of an uncredited ancestor. The book Risk presents the premortem as one of its own eleven exercises, and Gary Klein is not named anywhere in it. An omission with a short, clean, findable lineage behind it is worse than an obscure one, because nothing made it difficult.

What it does not say

It does not say the premortem has been shown to improve decisions. Klein's article is a practitioner's proposal in a management magazine, not a trial, and neither it nor Kahneman's account reports a controlled test with a control condition, a measured outcome or a sample.

It does not inherit the evidence behind prospective hindsight. Mitchell, Russo and Pennington tested prospective hindsight. They did not test the premortem, and those are two different claims.

It does not work without the conditions it assumes. The exercise manufactures psychological safety artificially for an hour, and in a room where doubt is still punished afterwards it is a form and not a procedure.

---