A cognitive bias is a systematic error in judgement — one that leans in a predictable direction instead of scattering at random. The word that matters is systematic. Rolling a wrong answer once is a mistake; getting it wrong in the same direction, reliably, across thousands of people, is a bias. The idea comes from research on how people judge uncertain things when they cannot calculate the answer.
What a cognitive bias is — and what it is not
Three different things are called a bias in everyday use, and only one of them is the subject of this page.
- A cognitive bias is a predictable tilt in how anyone judges, remembers or decides. It is a description of normal human thinking, not a fault in a particular person.
- A cognitive distortion is a term from cognitive behavioural therapy for a habitual negative thought pattern — such as all-or-nothing thinking or catastrophising — that a client works on with a therapist. Several languages use nearly the same words for both ideas, which is why the two get mixed up.
- A prejudice is an attitude towards a group of people. Some cognitive biases feed prejudices, but the words are not interchangeable.
The research tradition behind the first meaning began with a 1974 paper in Science by Amos Tversky and Daniel Kahneman, Judgment under Uncertainty: Heuristics and Biases. Their argument was that people faced with an uncertain quantity do not compute a probability; they substitute a small number of mental shortcuts, or heuristics. The shortcuts are usually efficient, and the errors they produce are regular enough to be predicted in advance. That predictability is what turned a list of quirks into a research programme.
Types of cognitive bias, with examples
There is no agreed count. Popular articles quote a dozen; reference lists run past a hundred, mostly because the same tendency is given different names in different fields. These are the ones people look up most often.
| Bias | What it is | An everyday example |
|---|---|---|
| Confirmation bias | Seeking, noticing and remembering evidence that fits what you already believe | Reading the one review that agrees with your choice and skimming the three that do not |
| Anchoring | Letting the first number you encounter pull your final estimate towards it | A high asking price making a smaller discount feel like a bargain |
| Availability heuristic | Judging how common something is by how easily an example comes to mind | Overestimating rare risks that were recently in the news |
| Hindsight bias | Believing, after the fact, that the outcome was obvious all along | “Everyone could see that result coming” — said by people who did not predict it |
| Bias blind spot | Detecting bias more readily in other people than in yourself | Agreeing that people overrate their own driving, and rating your own driving above average |
| Dunning-Kruger effect | The weak link between how well people perform and how well they think they performed | Leaving an exam confident, having misread the instructions |
The last one has a page of its own, because the famous version of it is not quite what the original study found: the Dunning-Kruger effect explains what the 1999 data actually showed and where the well-known curve came from.
The bias blind spot is worth a second look, because it is the one that makes the rest hard to fix. In three surveys reported by Emily Pronin, Daniel Lin and Lee Ross in 2002, people rated themselves as less subject to a range of biases than the average American, than their own classmates and than fellow travellers in an airport. In a follow-up, participants who had shown a better-than-average bias insisted their self-assessments were accurate and objective even after reading a description of how they could have been affected.
Does being intelligent protect you from cognitive bias?
This is the question an intelligence site should answer, and the honest answer is: far less than most people assume.
The most direct evidence is a 2012 paper by Richard West, Russell Meserve and Keith Stanovich in the Journal of Personality and Social Psychology. Its title states the result: “Cognitive sophistication does not attenuate the bias blind spot”. Being better at reasoning tasks did not make people better at seeing bias in their own judgement.
A separate line of work reaches the same place from the measurement side. When Irene Scopelliti, Carey Morewedge and colleagues built and validated a scale for the bias blind spot in 2015, they reported that it was distinct from measures of intelligence, decision-making ability and personality traits related to self-esteem and self-presentation — a separate characteristic, not a by-product of being less able. Their scale predicted concrete behaviour: people who scored higher on it were more likely to ignore advice from others.
Stanovich and West made the general version of this argument in 2008 in a paper called On the relative independence of thinking biases and cognitive ability. If the two were the same thing, an intelligence test would already measure rationality. The research tradition that follows from this treats them as different targets — which is the reason short measures of reflective thinking exist at all, such as the three-item test introduced by Shane Frederick in 2005, built from questions with an intuitive answer that is wrong.
None of this says intelligence is useless here. It says the protection is narrower than it feels. A reasoning score tells you about the machinery; whether you stop and use it on your own conclusions is a different habit. Our own test measures reasoning performance, and how IQ tests work sets out what that number does and does not cover.
How much of this research has actually replicated?
Cognitive-bias articles rarely mention that this literature was near the centre of psychology's replication crisis. The best single answer comes from Many Labs 2, published in 2018, which repeated 28 classic and contemporary findings with peer-reviewed protocols across 125 samples, 15,305 participants and 36 countries.
- 15 of the 28 replications (54%) produced a statistically significant effect in the same direction as the original; 14 (50%) survived a much stricter threshold.
- The effects shrank. The median comparable effect size was 0.60 in the original studies and 0.15 in the replications. In 16 of the replications (57%) the effect was small by the usual convention, and 9 (32%) pointed in the opposite direction to the original.
- A weak result was not explained by where it ran. Differences between settings showed up mainly in the findings with the largest effects; among the effects that came out near zero overall, only one varied significantly by setting — which argues against the common excuse that a failed replication was run on the wrong people.
So the right posture is neither “all of this is debunked” nor “here are the 12 biases that rule your life”. Some effects are robust and smaller than advertised; some did not survive careful testing at all. Any list that presents every entry with equal confidence is telling you less than it appears to. The same pattern shows up elsewhere in popular psychology — the learning styles claim failed direct testing, while growth mindset turned out to be real but much weaker than its reputation.
Can you reduce your own biases?
Early attempts to train bias away worked so poorly that researchers largely moved on to redesigning decisions instead — changing the default, the incentive or the way the options are presented. Then two longitudinal experiments reported by Carey Morewedge and colleagues in 2015 found a single session was enough to move the measurement. Participants either played one computer game or watched one instructional video covering biases relevant to intelligence analysis.
- Immediately afterwards, the games reduced measured bias by at least 31.9% and the videos by at least 18.6%.
- At least two months later the reduction was still there — at least 23.6% for games and 19.2% for videos. Games, which gave personalised feedback and practice, beat videos.
- The improvement was domain general: it appeared on problems in contexts and formats the training had not used.
Practical habits that follow from the research, rather than from motivational writing:
- Write the estimate down before you hear anyone else's. Anchoring works through the first number in the room.
- Ask what would have to be true for you to be wrong — a concrete alternative, not a general willingness to reconsider.
- Record predictions with dates. Hindsight bias needs your memory of what you expected; a written record removes the raw material.
- Assume the blind spot applies to you. It is the single best-supported finding on this page, and the one everybody exempts themselves from.
Trust and scope notes
This page is educational. It summarises published research on judgement and decision-making and is not psychological, medical or financial advice, and it is not a diagnosis of anyone. If you are looking for help with distressing thought patterns, that is the clinical idea of a cognitive distortion and belongs with a qualified professional, not with this page.
Our IQ test measures reasoning performance. It does not measure bias, rationality or self-insight, and a score does not tell you how well you judge your own judgement. Whether a test score is a good measure of anything is examined in are IQ tests accurate?
IQ Revealed is not affiliated with, endorsed by or connected to the American Psychological Association, the American Association for the Advancement of Science, the Institute for Operations Research and the Management Sciences, SAGE Publishing, or any of the universities, journals or researchers named on this page. They are named so that each claim can be traced to the work that made it. Quoted phrases are taken from the published titles and abstracts of the papers linked in the sources.