Bertrand Russell told a story about a chicken. Every morning the farmer comes and feeds it. Day after day, the evidence mounts: the farmer is a source of food. The chicken’s confidence grows with each feeding, right up to the morning the farmer arrives and wrings its neck.
The chicken was not stupid. It reasoned exactly the way we all do — from the past to the future. It just discovered, too late, that the reasoning has no guarantee behind it.
The problem, named
This is the problem of induction, and David Hume set it out two centuries before Russell’s chicken. We believe the future will resemble the past. We believe cause reliably follows effect. But, Hume argued, we have no rational proof of either. Our certainty is a habit of mind — a psychological expectation built from repetition, not a logical truth we can defend.
No number of sunrises proves the sun must rise tomorrow. It only makes us feel it must.
Why a philosopher’s puzzle belongs in a risk meeting
I spend my working life around models and systems that are, at bottom, inductive. A risk model backtests against history and projects it forward. A capacity plan assumes next quarter looks like last quarter. A legacy system is trusted because it has never failed before.
Each of these is the chicken’s reasoning in a spreadsheet. And each carries the chicken’s blind spot: the pattern holds until the regime changes, and the change is invisible from inside the pattern.
The most dangerous sentence in technology and finance is not “this might fail.” It is “it has always worked.” The first invites scrutiny. The second forecloses it.
What to do with an argument you cannot win
Hume’s point is not that induction is useless — we could not function without it. The point is that it is provisional, and should be held that way. A few working habits follow from taking it seriously:
- Name the assumption. Behind every confident forecast is a claim that the future resembles the past. Say it out loud. It is easier to challenge once it is visible.
- Hunt for the regime change. Ask what would have to be true for the pattern to break — new regulation, a different rate environment, a dependency you don’t control — and watch those things, not the pattern itself.
- Distrust the longest-running success. The system that has never failed has simply never yet met the conditions that break it. Long reliability is comfort, not proof.
- Keep a margin for the unprecedented. The failure that matters is usually the one your history contains no example of.
The takeaway
Russell’s chicken is not a warning against reasoning from experience. It is a warning against mistaking experience for certainty. The farmer’s schedule was real and the evidence was genuine — and it was still fatally incomplete.
Good judgement under uncertainty is not about having more history. It is about remembering that history, however long, is always a sample — and that you are somewhere inside it, unable to see the edge.