2 Heuristics in Estimation and Prediction

Why quick judgments can go wrong

When estimating probabilities, frequencies, or future values, people often rely on heuristics: quick mental shortcuts that can make judgments efficient, especially when information is incomplete. The same shortcuts can also produce predictable errors when an easy-to-use cue is mistaken for the quantity being judged.

Three influential heuristics use different cues: ease of recall, resemblance, and an initial number. None is guaranteed to produce an error; each can be useful when its cue is relevant. The important check is whether the cue provides enough evidence for the judgment.

Availability and ease of recall

The availability heuristic uses ease of recall to estimate how frequent or likely something is. Ease of recall can be informative because common events may generate more examples, but recall is also shaped by vividness, recency, personal experience, and media attention. These influences mean that what comes readily to mind does not always match actual frequency.

After hearing several reports about house fires, for example, someone may estimate that a house fire is more likely than it really is. Dramatic, memorable reports are easy to recall, while less vivid risks may come to mind less readily. The judgment substitutes “How easily can I think of an example?” for “How often does this happen?”

Availability can shape predictions, too. A recent product launch or memorable success story may be easy to recall, leading a forecaster to overestimate the chance that a similar launch will succeed. To check such a judgment, seek relevant records or a representative set of cases rather than relying only on memorable examples.

Representativeness and category judgments

The representativeness heuristic judges the probability that a person, event, or outcome belongs to a category by how closely it resembles a typical example. Resemblance can be useful, but it may draw attention away from other evidence, particularly base rates: how common the category is in the first place.

A person who is quiet, orderly, and fond of books may seem likely to be a librarian. That description alone does not establish the person’s occupation. A sound estimate also considers how common librarians are relative to other occupations and how diagnostic the description actually is.

A useful check is to ask: What is the relevant base rate, and how strongly does this evidence distinguish among the possibilities?

Resemblance and sequences

Representativeness can also distort predictions about sequences. A short run of outcomes that resembles a familiar pattern may seem especially likely to continue or unlikely to recur, even when the process is random.

For example, after several coin flips land heads, someone may expect tails because the sequence “feels due.” If the flips are independent and the coin is fair, however, previous results do not change the probability of the next flip.

Anchoring and numerical estimates

The anchoring-and-adjustment heuristic begins with an initial value—the anchor—and uses adjustments from that starting point to make a numerical estimate. Estimates often remain too close to the starting value, even when it is arbitrary or only weakly related to the answer.

For example, after hearing a suggested repair price, someone may use it as a starting point and adjust upward or downward. Even an unreliable suggestion can influence the final estimate if the person does not adjust far enough away from it.

Anchors can affect forecasts as well as estimates. An early sales projection, an initial deadline, or the first price mentioned in a negotiation may shape later judgments. To reduce unwanted anchoring, form an independent estimate before seeing others’ numbers, compare several relevant reference values, and make adjustments explicit. A reliable benchmark can provide useful context, but its relevance should be checked.

Compare the three heuristics

The three heuristics direct attention to different cues:

• Availability: How easily examples or scenarios come to mind. Memorable or recent examples may seem more frequent or likely than they are.

• Representativeness: How much an instance resembles a category or pattern. Resemblance may overshadow base rates or the process generating the data.

• Anchoring: An initial number or starting point. An estimate may stay too close to an irrelevant or weak starting value.

Tversky and Kahneman described heuristics as economical strategies that can be effective but can also lead to systematic errors. More reliable judgments check records and frequencies, consider base rates, and compare estimates against independent benchmarks rather than accepting the first intuitive answer.