Fast-and-frugal heuristics
Simple rules that use little information yet often decide as well as complex models.
What it means
Fast-and-frugal heuristics are simple decision rules that ignore most available information and forgo heavy computation, yet under the right conditions match or beat far more complex statistical models. Central to Gerd Gigerenzer's 'ecological rationality' program, they are studied as part of an 'adaptive toolbox' from which the mind selects a rule fitted to the structure of the environment, in deliberate contrast to the heuristics-and-biases tradition that frames mental shortcuts mainly as sources of error. A canonical example is take-the-best, which decides between options by consulting cues one at a time in order of validity and choosing as soon as one cue discriminates, ignoring all remaining cues. Their surprising accuracy is explained by the bias–variance tradeoff: by being simple they incur some bias but drastically cut variance, so they generalize robustly when data are scarce, noisy, or uncertain, where complex models overfit—a 'less-is-more' effect. They are 'fast' because they stop searching quickly and 'frugal' because they need little information. It matters because it reframes simplicity as a feature, not a failing, of good judgment.
The adaptive toolbox
The program treats the mind not as one general-purpose calculator but as a repertoire of specialized rules, each assembled from three building blocks: a search rule that says where to look, a stopping rule that says when to stop looking, and a decision rule that says what to do with what has been found. Take-the-best searches cues in order of validity, stops at the first cue that discriminates between the options, and chooses the option that cue favors. Which rule the mind should reach for depends on the structure of the environment, the idea of ecological rationality. A heuristic is neither smart nor foolish in the abstract; it is well or badly matched to a particular world, and the same rule can be either.
Why less can be more
The counterintuitive accuracy comes from splitting error into bias and variance. A flexible model with many free parameters bends to fit the data it is trained on, but each parameter is estimated from a limited, noisy sample, so the fitted values wander from one sample to the next, which is high variance. When the model then meets fresh cases, that wandering resurfaces as error: it has partly memorized yesterday's noise. A frugal rule with few or no parameters cannot bend that far, so it carries more bias but far less variance. When the world is uncertain and samples are small, the variance term dominates total error, and the simple rule generalizes better. Give it abundant, stable, well-measured data and the advantage reverses toward the complex model.
What the evidence shows
In head-to-head competitions across real datasets, from which of two cities is larger to which firm will fail, take-the-best has matched or beaten multiple regression on out-of-sample accuracy while reading only a fraction of the cues. In finance, the naive 1/N rule that spreads money equally across assets was not reliably beaten by any of fourteen optimizing models across seven datasets, because the optimizers' estimation error swamped their theoretical edge. In an emergency room, a three-question fast-and-frugal tree sorted chest-pain patients into coronary care more accurately than physicians' own judgments and a statistical instrument. The recurring pattern is that simplicity pays where predictability is low and data are scarce.
The dispute over description
Two claims must be kept apart: that simple rules can predict well, and that people actually use them. The first is largely a mathematical result and holds under stated conditions; the second is contested. When Bröder tested take-the-best directly, only a minority of participants behaved as the rule predicts, and that share rose when acquiring information was costly and the cue structure favored it, suggesting people switch strategies rather than run one fixed algorithm. Critics also note that take-the-best's choices frequently coincide with those of weighting models, so fitting the observed data does not by itself prove the underlying mechanism. The program's strongest claims are prescriptive, about how to decide well under uncertainty, more than descriptive of how minds always work.
Examples
To predict which of two cities is larger, simply choosing the one you recognize—ignoring all other facts—can outperform weighting many cues, an instance of a fast-and-frugal heuristic.
Splitting savings equally across the available funds, the 1/N rule, holds its own against elaborate optimised portfolios, because the clever models fit yesterday's noise and the simple rule cannot.
A fielder catches a high ball without computing its arc: he runs so the ball stays at a constant angle in his gaze. One cue, no equations, and he arrives in time.
To catch a ball dropping out of the sky, an outfielder does not compute its trajectory; fixing the gaze on the ball and adjusting running speed so the angle of gaze stays constant steers the runner to the landing spot on a single cue.
To guess which lapsed customers will buy again, a manager can rank them by how long since their last purchase alone; this one-cue hiatus rule holds its own against a fitted statistical churn model.
First described in Gerd Gigerenzer and the ABC Research Group.
Key references
- Gigerenzer, G., & Gaissmaier, W. (2011). Heuristic decision making. Annual Review of Psychology, 62, 451-482. doi.org/10.1146/annurev-psych-120709-145346
- Hilbig, B. E. (2010). Reconsidering 'evidence' for fast-and-frugal heuristics. Psychonomic Bulletin & Review, 17(6), 923-930. doi.org/10.3758/PBR.17.6.923
- DeMiguel, V., Garlappi, L., & Uppal, R. (2009). Optimal versus naive diversification: How inefficient is the 1/N portfolio strategy? Review of Financial Studies, 22(5), 1915-1953. doi.org/10.1093/rfs/hhm075
- Gigerenzer, G., & Brighton, H. (2009). Homo heuristicus: Why biased minds make better inferences. Topics in Cognitive Science, 1(1), 107-143. doi.org/10.1111/j.1756-8765.2008.01006.x
- Broeder, A. (2000). Assessing the empirical validity of the 'take-the-best' heuristic as a model of human probabilistic inference. Journal of Experimental Psychology: Learning, Memory, and Cognition, 26(5), 1332-1346. doi.org/10.1037/0278-7393.26.5.1332
- Gigerenzer, G., & Goldstein, D. G. (1996). Reasoning the fast and frugal way: Models of bounded rationality. Psychological Review, 103(4), 650-669. doi.org/10.1037/0033-295X.103.4.650