What is a cognitive bias? How the mind takes shortcuts

A cognitive bias is a systematic error in judgment: a way of thinking that leads people wrong in a predictable direction, not at random. Most come from heuristics, the mental shortcuts the mind uses to judge quickly with little information. The shortcuts usually work well enough, which is why people have them. Biases are where they fail, and because they fail in the same way for most people, they can be studied, measured and sometimes corrected.

In short

• A cognitive bias is a predictable tilt in judgment, usually the side effect of a shortcut that works most of the time.
• The modern study of biases began with Amos Tversky and Daniel Kahneman in the 1970s, and anchoring, availability, framing and hindsight are among the best documented.
• Some famous findings have replicated strongly; others, such as ego depletion and several priming effects, largely failed when repeated with large samples.

What makes an error a bias?

The word "systematic" does the work. If people guessing the weight of an ox are off by random amounts, some too high and some too low, the errors cancel out in the average. A bias pushes most people the same way, so averaging doesn't help. A bathroom scale that always reads 2 kg heavy is biased; one that jitters around the true weight is merely noisy.

A heuristic is a rule of thumb, such as "if I can think of many examples, it must be common". That one, called the availability heuristic, is often right: things that happen often are usually easy to recall. It goes wrong when something is memorable for another reason. Plane crashes are vivid and heavily reported, so people tend to overestimate how dangerous flying is compared with driving. The heuristic isn't stupid. It is a fast estimate using the information that comes to mind first, and the bias is the gap between that estimate and the facts.

A cognitive bias is also different from a logical fallacy. A fallacy is a flaw in an argument, whoever makes it; a bias is a tilt in the person doing the judging. What is a logical fallacy? covers the first kind.

Where did the idea of cognitive biases come from?

In 1974, the Israeli psychologists Amos Tversky and Daniel Kahneman published Judgment under Uncertainty: Heuristics and Biases in Science. It described three shortcuts people use when judging likelihoods: representativeness (judging by how much something resembles a type), availability (judging by how easily examples come to mind) and anchoring (starting from a number and adjusting too little). For each, they showed the predictable errors it produces.

The work challenged the assumption in economics that people make decisions as rational calculators. With prospect theory in 1979, Tversky and Kahneman showed that people weigh losses more heavily than equal gains. Kahneman received the 2002 Nobel Memorial Prize in Economic Sciences for this research; Tversky had died in 1996, and the prize is not awarded after death (Nobel Prize).

What are the most common cognitive biases?

Lists online run to more than 100 named biases, many overlapping. These are among the best established, each with the kind of demonstration that first showed it.

Bias	What it is	The classic demonstration
Anchoring	Estimates stay too close to a number seen first, even an irrelevant one	People who saw a wheel of fortune stop at 10 guessed, on median, that 25% of UN member countries were African; those who saw 65 guessed 45% (Tversky and Kahneman, 1974)
Availability	Judging how common something is by how easily examples come to mind	Asked whether more English words start with K or have K as the third letter, most people say "start with", though by Tversky and Kahneman's count the reverse is true: words starting with a letter are easier to recall
Representativeness	Judging likelihood by resemblance, ignoring base rates	In the "Linda problem" (1983), most people rated "Linda is a bank teller and a feminist" as more likely than "Linda is a bank teller", which can't be, since the first is a subset of the second
Framing	The same choice, described as a gain or a loss, gets different answers	In the 1981 "disease problem", most chose the sure option when outcomes were described as lives saved, and the gamble when the identical outcomes were described as lives lost
Hindsight bias	After learning an outcome, believing it was more predictable than it was	In studies starting with Baruch Fischhoff's in 1975, people told how a historical event ended rated that ending as more likely than people not told
[[what-is-confirmation-bias	Confirmation bias]]	Seeking and favoring evidence that fits what one already believes

The Linda problem breaks a basic rule of probability: two things together can never be more likely than either one alone. Related errors have notes of their own: the sunk cost fallacy, survivorship bias and the Dunning-Kruger effect.

Which findings about biases have held up?

Not all of them. When psychology faced its replication crisis in the 2010s, biases research was tested hard, and the results split clearly.

In the first Many Labs project, published in 2014, researchers repeated 13 classic and recent effects in 36 separate samples, 6,344 participants in all, using the same methods (Klein and colleagues). Ten replicated consistently, among them anchoring and the gain-versus-loss framing effect. Two priming effects did not replicate at all: that a glimpse of the US flag makes people more conservative, and that images of money make people more accepting of the social system.

Ego depletion, the claim that self-control is a fuel that running down on one task leaves people weaker on the next, fared worse still. A preregistered replication in 23 laboratories with 2,141 participants, published in 2016, found an effect close to zero (Hagger and colleagues). The broader lesson, set out in what psychology is, is that the most famous findings were not always the most solid.

The core heuristics and biases, anchoring, framing, availability and hindsight, have been reproduced many times; many striking "subtle cue changes behavior" priming results have not. Even anchoring has limits: some studies using anchors that participants could plainly see were random, such as a die roll, found little or no effect.

Still argued

How irrational biases make people is debated. Kahneman and Tversky's tradition treats biases as departures from the rules of logic and probability. The psychologist Gerd Gigerenzer and colleagues argue that many "biases" are sensible shortcuts that work well in the real environments people live in, and that some look like errors only because laboratory puzzles strip away the context, or present probabilities in a form people rarely meet; stated as natural frequencies, such as "10 out of 1,000", some errors shrink a good deal. Both sides agree that heuristics are usually useful and sometimes fail; they disagree on how often, and on whether the failures say more about the mind or about the tests.

Can you avoid cognitive biases?

Knowing about a bias does little to switch it off. People readily spot biases in others while rating themselves as less affected, a pattern psychologists call the bias blind spot. What helps more is changing the procedure rather than relying on willpower.

A few habits have good support. Write down a prediction or an estimate before seeing anyone else's, so there is no anchor to stick to. Ask for the base rate before judging an individual case: how often does this happen in general? Restate a choice in the opposite frame, "90% survive" as "10% die", and see if the answer changes. Before deciding, imagine the decision has failed and list the likeliest reasons. And for important judgments, use a checklist and an outside opinion, because other people's biases point in other directions. Good habits of weighing evidence are, in large part, defenses against biases. A different mechanism, the discomfort of holding beliefs that clash with actions, is cognitive dissonance.

Check yourself

• Without looking back, explain the difference between a biased judgment and a merely noisy one.
• Why does the availability heuristic usually work, and when does it fail?
• Which famous findings replicated in the Many Labs project, and which didn't?

Sources

• Judgment under Uncertainty: Heuristics and Biases — Amos Tversky and Daniel Kahneman, Science, 1974
• The Prize in Economic Sciences 2002, press release — Royal Swedish Academy of Sciences, 2002
• Investigating Variation in Replicability: A "Many Labs" Replication Project — Richard Klein and colleagues, Social Psychology, 2014
• A Multilab Preregistered Replication of the Ego-Depletion Effect — Martin Hagger and colleagues, Perspectives on Psychological Science, 2016
• Cognitive bias — Wikipedia
• Anchoring effect — Wikipedia