The Inside View and the Outside View

What it is

The inside view is the natural way to think about a project. This team, this plan, this set of obstacles, this run of good or bad luck. It uses everything you know and it is where almost all forecasts come from.

The outside view feels perverse, because it deliberately sets aside detailed knowledge that is genuinely relevant, and it is usually more accurate. It asks what happened to comparable cases and starts from that distribution.

The pair is a diagnosis of the planning fallacy rather than a separate finding. The fallacy is what the inside view produces; the outside view is the proposed correction. Attribution needs care. Kahneman and Lovallo introduced the terms in Management Science in 1993, and an earlier corrective procedure using reference classes comes from Kahneman and Amos Tversky in 1979. Bent Flyvbjerg built the formal method on top of them, reference-class forecasting, using databases covering hundreds of projects, which is what turned a way of thinking into a procedure a public body can be required to follow. Kahneman records in the endnotes to Thinking, Fast and Slow at page 466 that numerous authors have mislabelled his terms, so any secondary source using the phrases differently should be checked.

In effect

The illustration Kahneman uses is a curriculum-writing team, seen from both sides. From the inside, the team's estimate was about two years: they knew their own competence, their material and their schedule. From the outside, a colleague's recollection of comparable teams gave a very different picture, with roughly forty per cent never finishing and the finishers taking between seven and ten years. The project took eight.

The outside-view number was, in the event, almost exactly right, and this is where the argument has to slow down, because Kahneman refuses the obvious conclusion. He states that the accuracy of that forecast should not count as evidence for the validity of the outside view, since a single case landing inside a predicted range demonstrates nothing about the method. He also admits that looking for the outside view will never be the natural thing to do, including for him, after a career spent arguing for it. Both admissions are his, in the same chapter that recommends the technique.

The case for the technique rests elsewhere. Reference-class forecasting is a formal method with data behind it and has been adopted by public bodies as a required step rather than an exhortation, which fits the broader pattern that individual debiasing barely works and organisational debiasing does. The outside view is effective largely because it can be made compulsory, which is a different kind of claim from telling someone to be less optimistic.

What it does not say

It does not say the book's examples prove the method. Kahneman disqualifies his best illustration himself, and writing that presents the curriculum story as proof is doing something the author explicitly refused to do.

It does not say the supporting evidence is uncontested. The accuracy gains for reference-class forecasting rest mainly on Flyvbjerg's own group's data, and some construction-management researchers read the cost-overrun evidence differently. The planning fallacy the method is meant to correct is on firmer ground, with Roger Buehler, Dale Griffin and Michael Ross establishing it experimentally in 1994.

It does not say any assembled comparison set is an outside view. A reference class built from a handful of remembered cases is the law of small numbers wearing the remedy's clothes, and the method's value depends entirely on the class being large and genuinely comparable.

It does not say the habit will stick without a structure to enforce it. A method that requires a person to override an instinct every time is fragile outside an institution that makes it a step in a process, and that limit is Kahneman's own.

---