← Wonderland

Kahneman and Tversky on Heuristics and Biases

How two Israeli psychologists overturned the assumption that humans reason like calculators — and quietly redrew the map of economics, medicine, and law.

In the 1960s, economics was built on a flattering picture of human beings. People were modelled as rational agents who weigh costs and benefits, update beliefs by Bayes' rule, and choose to maximise expected utility. The picture was elegant. It was also wrong in specific, replicable, almost embarrassing ways. The people who pointed this out were not economists. They were two psychologists at the Hebrew University in Jerusalem — Daniel Kahneman and Amos Tversky — who began, almost as a hobby, asking colleagues unusual questions over coffee and finding that even statisticians answered them like amateurs.

What emerged over the next thirty years was a new picture of the mind. Not a calculator, but two systems running in parallel, one fast and intuitive, the other slow and deliberate. Each makes different mistakes. The cataloguing of those mistakes — the heuristics-and-biases programme — earned Kahneman the 2002 Nobel Prize in Economics. Tversky had died six years earlier; the prize is not awarded posthumously.

Two minds in one head

The framework Kahneman would later call System 1 and System 2 was not a brain scan. It was a way of organising the data.

System 1 is fast. It recognises a face, completes “bread and…”, senses hostility in a voice, finishes “2 + 2 =”. It runs all the time, effortlessly, without your permission. System 2 is slow. It computes 17 × 24, checks the validity of a complex argument, fills in your tax form. It runs when you ask it to, costs effort, and shuts off when tired.

Most of life is System 1. This is efficient — we would starve if we had to consciously deliberate every step across a room — but it has a price. System 1 substitutes. When you face a hard question (“How happy am I with my life?”) it quietly answers an easier one (“How is my mood this morning?”) and hands the answer up to consciousness. You experience this as having an opinion. You do not notice the substitution.

The result is that intuition feels confident exactly when it should not. This is the discovery underneath everything else they did.

The biases zoo

Kahneman and Tversky's papers describe a menagerie of these substitutions. Three are foundational.

Availability. We estimate the frequency of an event by how easily examples come to mind. Plane crashes are on the news; bathtub drownings are not. Ask people which kills more, and they confidently rank crashes higher. The bathtubs win, by a wide margin.

Representativeness. We judge probability by how much something resembles a stereotype. The famous Linda problem: Linda is 31, single, outspoken, and bright. She majored in philosophy and was concerned with discrimination. Which is more probable — that Linda is a bank teller, or that Linda is a bank teller and active in the feminist movement? Most people pick the second. But the conjunction of two things can never be more probable than one of them alone. It cannot. Yet the second story fits the stereotype better, and the answer slides out before logic catches it.

Bank tellers Feminist activists Linda goes here… but she is still a bank teller

The conjunction fallacy. The shaded set is a subset of “bank tellers,” so it can never be more probable. System 1 does not see sets.

Anchoring. Spin a wheel marked 0 to 100. Ask people: is the percentage of African countries in the UN higher or lower than the number you just spun? Now ask them to estimate the percentage. Their estimates are dragged toward the spun number, which they know is random, which they have just watched a wheel produce. The anchor sticks anyway.

These are not laboratory curiosities. They show up in expert judgement too. Doctors estimating disease prevalence, judges setting sentence lengths, intelligence analysts projecting probabilities — all anchored, all swayed by availability, all reading stereotypes as probabilities.

Prospect theory and the value function

The deepest result came in 1979. Classical economics assumed people choose between gambles by computing expected utility — multiplying outcomes by probabilities and adding. Kahneman and Tversky showed real people do not. Instead, they evaluate outcomes as gains and losses relative to a reference point, and they hate losses about twice as much as they like equivalent gains.

This is prospect theory. Plotted, it produces a kinked, S-shaped curve through the origin: concave for gains (a second £1000 feels less than the first), convex and steeper for losses (losing £1000 stings more than gaining £1000 thrills).

gain → ← loss value felt reference point losing £100 hurts a lot gaining £100 pleases… …but about half as much

The prospect theory value function. Losses loom larger than gains, and both flatten as they grow.

From this kink falls a great deal. The endowment effect — that you demand more to sell something than you would pay to buy it. The status quo bias — that defaults stick. The disposition effect — that investors hold losing stocks too long and sell winners too early. The framing effect — that “90% survival” and “10% mortality” pull different choices from the same patients. All of it is the same kink, viewed from different angles.

Why this matters

The science of judgement under uncertainty replaced the assumption of rational man with something stranger and more useful: a mind whose intuitions are mostly serviceable, occasionally catastrophic, and reliably wrong in predictable directions. Once you can predict the error, you can engineer around it. Organ donation rates skyrocket when consent is the default instead of an opt-in — same population, same preferences, different frame. Pension enrolment doubles when employees must opt out rather than in. The architecture of choice is not neutral. Knowing this gives you a lever.

“We can be blind to the obvious, and we are also blind to our blindness.” — Daniel Kahneman, Thinking, Fast and Slow

It also gives an epistemic humility. The most uncomfortable lesson of the research is not that people are biased. It is that knowing about a bias does not protect you from it. The anchoring effect works on Kahneman as he describes it. System 1 is not a defect to be debugged; it is the kind of thing you are. The best one can do is to recognise the situations where it fails — irreversible decisions, high stakes, statistical reasoning, evaluating people — and force System 2 to take over.

That is the practical residue of forty years of work: not a list of fallacies to memorise, but a habit of pausing in the right places.


Further reading

  1. Tversky, A. & Kahneman, D. (1974). Judgment under Uncertainty: Heuristics and Biases. Science, 185.
  2. Kahneman, D. & Tversky, A. (1979). Prospect Theory: An Analysis of Decision under Risk. Econometrica, 47.
  3. Tversky, A. & Kahneman, D. (1981). The Framing of Decisions and the Psychology of Choice. Science, 211.
  4. Kahneman, D. (2011). Thinking, Fast and Slow. Farrar, Straus and Giroux.
  5. Lewis, M. (2016). The Undoing Project. W. W. Norton.