Prospect theory described how a person values a gamble once its odds are laid before him. But the odds are rarely laid before him. In the ordinary run of life the probabilities are not given but guessed — how likely is this venture to pay, this patient to recover, this applicant to succeed, this road to be clear — and the guessing is the prior and harder task on which all valuing depends. Here the same two investigators, Amos Tversky and Daniel Kahneman, in work that preceded prospect theory by half a decade, found a second country of systematic departure. Their thesis was that when the mind is asked a hard question about likelihood or quantity, it does not labor toward the true answer by the rules of probability; it quietly swaps the hard question for an easier one it can answer at a glance, and reports the answer to the easy question as though it were the answer to the hard. The substitutions are quick, effortless, and most of the time good enough. But they are not the thing asked for, and so they fail — not at random, like a hand that slips, but lawfully, in the same direction every time, for the same reasons. Tversky and Kahneman called the shortcuts heuristics and the lawful failures biases, and they built a science out of reading the second to expose the first.
The first shortcut is REPRESENTATIVENESS: to judge how probable it is that something belongs to a category, the mind asks instead how much it RESEMBLES the category, how well it fits the type. Often resemblance and probability run together, and the shortcut serves. But they can be pulled apart, and when they are, the judgment follows resemblance off a cliff. Describe a quiet, tidy, detail-loving soul who likes order and small careful tasks, and ask whether he is more likely a librarian or a farmer, and most people answer librarian — the description fits the type — though farmers so outnumber librarians that the tidy soul is, on the bare arithmetic, far more likely to be found behind a plow than a desk. The mind has fastened on the resemblance and ignored the BASE RATE, the sheer prior commonness of the categories, which probability theory says must weigh in any sound estimate. This insensitivity to base rates is the signature error of representativeness, and it is everywhere a vivid description tempts us to forget how rare the vivid case really is.1
Representativeness produces an error stranger than base-rate neglect, one that violates not merely good statistical sense but the bare logic of probability. Sketch a woman, single and outspoken and deeply engaged as a student with questions of justice and discrimination, and then ask which is more probable: that she is now a bank teller, or that she is a bank teller AND active in a movement for women's rights. A large majority choose the second. Yet the second cannot possibly be more probable than the first, for it is a part of the first — every feminist bank teller is a bank teller, so the conjunction can be no likelier than its conjunct, and is almost always less. People rank it higher because the fuller portrait, the teller-and-activist, RESEMBLES the sketch better than the bare teller does, and resemblance is the question the mind has answered. This is the conjunction fallacy, and its force is that adding detail — which can only narrow a possibility and lower its probability — makes a story feel MORE likely, because detail makes it fit the type. It is the engine behind every plausible, specific, and overconfident forecast, every vivid scenario believed precisely because its richness of incident makes it ring true.2
The second shortcut is AVAILABILITY: to judge how frequent or how probable a kind of event is, the mind asks how EASILY instances of it come to mind. Things that are common usually are easy to recall, so again the shortcut often serves. But ease of recall answers to more than frequency — it answers to vividness, to recency, to how much an event frightened or amused or was talked about — and wherever those pull apart from true frequency, the estimate is dragged off with them. People judge violent death by accident or homicide to be likelier than death by stroke or diabetes, though the quiet diseases kill far more, because the violent deaths are the ones the newspapers carry and the mind retrieves. They overrate the dangers that are dramatic and dread-soaked — the plane, the shark, the act of terror — and underrate the dangers that are dull and frequent — the car, the stairs, the slow disease — because the dramatic hazard is the one that springs to mind. The availability of an instance, its readiness to be recalled, masquerades as evidence of how often the thing occurs, and a world mediated by stories and screens, which serve up the rare and the vivid precisely because they are rare and vivid, keeps the masquerade well supplied.3
The third shortcut is ANCHORING AND ADJUSTMENT: asked to estimate some unknown quantity, the mind seizes on a starting number — an anchor, whatever happens to be at hand — and adjusts away from it toward its true estimate, but adjusts too little, so the final figure stays pulled toward where it began. What gives this its uncanny edge is that the anchor need not be relevant, need not even be meaningful; it works when it is plainly arbitrary. In the experiment that fixed the phenomenon, a wheel of fortune was spun to land on a number patently random, and people were then asked what percentage of the nations of Africa belonged to the United Nations; those whose wheel had stopped high guessed high, those whose wheel had stopped low guessed low, though everyone had watched the number arrive by chance and no one supposed it carried any information at all. The arbitrary anchor dragged the estimate toward itself regardless. This is why the first figure named in a negotiation shapes the whole, why a high sticker price makes a mere discount feel like generosity, why any number placed before us — a suggested donation, an initial offer, a stray statistic — bends the judgments that follow toward it. The mind cannot help starting somewhere, and where it starts taints where it ends.4
Gather the three, and a picture of the judging mind emerges that is the very opposite of the calculating agent the older theory assumed. Judgment under uncertainty does not run on the rules of probability; it runs on resemblance, on recall, on whatever number lies near to hand — fast machinery, evolved or learned for a world that demanded speed, and serviceable enough that we survive by it daily. The biases are not the malfunctions of this machinery but its fingerprints: each systematic error reveals the shortcut that produced it, exactly as a visual illusion, by fooling the eye in a regular way, lays bare the rules by which the eye constructs what it sees. By cataloguing the illusions of judgment, Tversky and Kahneman reverse-engineered its workings, and together with prospect theory — that account of how the prospects, once judged, are valued — they mapped both halves of decision under uncertainty, the forming of beliefs and the weighing of outcomes, and found each one departing from the rational model in its own lawful way.5 Yet notice the word they chose, and that this volume has borrowed for its title: BIAS. A bias is a leaning-away from something, and the something is the correct answer, the verdict of logic and probability against which the heuristic is measured and found wanting. Built into the very name is a judgment — that the shortcut is a defect, the departure an error, the rational ideal the rightful standard. That judgment is the one thing in this whole science that its next and final plate will refuse to grant. The flagship asks whether the mind, in reaching for its shortcuts, is failing a test of reason, or passing a different test that the test of reason never thought to set.