What is a cognitive bias?
A systematic deviation of judgment from logic or probability that repeats predictably under the same conditions.
A cognitive bias is a systematic deviation of judgment from logic or probability: not a random error that cancels out over time, but a gap that repeats in the same direction whenever the same conditions appear. The term originates in 1974 with Amos Tversky and Daniel Kahneman, who published "Judgment under Uncertainty: Heuristics and Biases" in Science (185, 4157, pp. 1124-1131): they showed that people estimating probabilities and quantities do not run a formal calculation but a small set of heuristics, representativeness, availability, anchoring, that work well most of the time and fail predictably under specific conditions. It is therefore not a matter of intelligence or ignorance: the same mental shortcuts that allow fast decisions are what produce the error, which is why a bias resists the person who knows about it as much as the person who does not.
See also: Anchoring bias · The tendency to let the first number that appears in a discussion drive every estimate made after it.
The families of biases
The glossary's entries on individual biases, grouped by the ground they act on rather than by their field of origin.
| Family | Entries |
|---|---|
| Judgment and estimation | Confirmation bias, Anchoring bias, base rate fallacy, planning fallacy, apophenia, survivorship bias, Dunning-Kruger effect |
| Commitment and ownership | Sunk cost fallacy, status quo bias, IKEA effect, cognitive dissonance |
| Groups and organizations | Information cascade, bystander effect, Ringelmann effect, normalization of deviance |
| Choice architecture | Decoy effect, casino effect |
| Humans and machines | Automation bias, algorithm aversion, AI anthropomorphism |
An enterprise example
An investment committee evaluates three vendors for a data project. The first to present sets a reference budget that anchors every negotiation that follows; whoever already picked a partner in an earlier phase tends to defend that choice rather than assess it; and if the final meeting includes a deliberately expensive baseline, the vendor in the middle suddenly looks reasonable. Three biases from three different families act in the same meeting, and nobody in the room notices any of them happening.
Why it matters for decision makers
The countermeasures that work cut across families instead of being specific to each bias. Set evaluation criteria before seeing the options, so the order of presentation does not become the criterion. Have someone who did not build the technology evaluate it, because whoever invested time and credibility in a choice becomes its least reliable judge. Use reference classes, meaning data from comparable projects already closed, instead of intuitive estimates. And run a pre-mortem: imagine the project has failed and ask why, before starting, instead of finding out along the way. None of these practices removes the biases, which stay part of how the mind judges quickly; all of them shift where and when the bias gets a chance to act, which is the one lever actually available.
Related terms
- Anchoring bias · The tendency to let the first number that appears in a discussion drive every estimate made after it.
- Sunk cost fallacy · The sunk cost fallacy is the error of funding a project for what was already spent and cannot be recovered, rather than for the result it can still produce.
- Status quo bias · Preferring the current state of affairs simply because it is current, regardless of its actual value.
- Confirmation bias · The tendency to seek, interpret, favor and recall evidence in ways that confirm a belief already formed.
- Dunning-Kruger effect · The tendency to overestimate one's own competence when the skill needed to judge it is itself missing.
If it is a technical decision you keep postponing, that is the decision I help you make.
SEE CUSTOM TOOLS AND GOVERNANCE