What is confirmation bias?

Confirmation bias is a cognitive bias in which a person favours evidence that confirms an existing belief or hypothesis. It operates in three places at once:

• Search. You look in places likely to agree with you — the bullish thread, the analyst who already liked the stock, the customer who already loves the product.
• Interpretation. Ambiguous evidence gets read as supportive. A flat quarter becomes "resilient"; a bad one becomes "an outlier".
• Recall. Confirming episodes come to mind faster, so your track record feels better than it was.

In plain terms: it is not that you ignore the truth on purpose. It is that the version of the world reaching your desk has already been filtered before you start reasoning about it.

The term comes from cognitive psychologist Peter Wason, whose rule-discovery experiments in the 1960s found that people test a hypothesis by generating cases they expect to confirm it, rather than cases that could disprove it — the opposite of how falsification works in science.

Psychologists classify it alongside the other systematic judgement errors documented by Amos Tversky and Daniel Kahneman: not random mistakes, but predictable deviations that show up reliably across people and situations. That predictability is the useful part. A bias you can predict is a bias you can design a process around.

Further reading: overview of the research literature.

The bias is easiest to recognise in someone else's reasoning. These are the shapes it takes most often:

• Trading

  You hold a losing position and read only the bullish threads about it. Each one feels like new evidence, but you selected for it. The bearish earnings note gets labelled 'noise' rather than data.

• Investing

  After buying, you follow the company's own announcements and the analysts who already rated it a buy. Disconfirming coverage never enters the feed, so the thesis looks stronger every week without being tested once.

• Hiring

  You form an impression of a candidate in the first two minutes, then spend the interview asking questions that would confirm it. Their strengths in other areas go unprobed because you were not looking there.

• Building a product

  You show the prototype to friendly users, count their enthusiasm as validation, and treat churned users as a bad fit rather than a signal. The roadmap gets built on the survivors' opinions.

• Medicine and science

  A clinician anchors on an initial diagnosis and orders tests likely to confirm it. In research, the equivalent is p-hacking: analysing data repeatedly until it agrees with the hypothesis.

• News and politics

  You read the outlet whose framing you already share. Ranking algorithms then amplify the pattern, so the feed becomes a record of your prior beliefs rather than a source of new information.

Three forces keep it in place. Efficiency: re-deriving a belief from scratch is expensive, so the brain defaults to updating cheaply. Identity: once a view has been stated out loud — to a partner, a group chat, or a board — abandoning it costs status, so the mind quietly looks for reasons not to. Memory: retrieval favours belief-consistent examples, so the evidence base you feel you have is already skewed.

This is why "try to be more objective" fails as a remedy. The bias operates before the part of you that intends to be objective gets involved.

In a position, confirmation bias rarely shows up as a single bad call. It shows up as a thesis that never gets re-tested while the facts underneath it change: you keep reading the same sources, the position keeps looking justified, and the stop keeps moving.

It also compounds with other biases. Confirmation bias supplies the evidence that makes loss aversion feel like patience and sunk cost feel like conviction. That is why catching the language early matters more than diagnosing it afterwards.

You cannot remove the bias, but you can make it harder for it to reach the decision. Each of these works by putting structure where willpower would otherwise be:

1. 1

   Write the thesis before the research

   Put your view and your reasoning in writing first. Once it exists on paper you can check what you actually believed, rather than the tidied-up version memory supplies afterwards.

2. 2

   Name your falsifiers in advance

   Specify the price, metric, or date at which you would conclude you were wrong. A thesis with no exit condition cannot be disproven — only defended.

3. 3

   Argue the other side properly

   Write the strongest version of the opposing case, not a strawman. If you cannot state why an intelligent person disagrees with you, you have not finished researching.

4. 4

   Change your inputs, not just your effort

   Deliberately add sources that disagree with you. Confirmation bias operates on what reaches you, so the fix is upstream of your judgement.

5. 5

   Ask a real disconfirmer

   Give one person the explicit job of finding the flaw. Assigning the role removes the social cost of raising it.

6. 6

   Check your own language

   Phrases like 'as expected', 'this proves I was right', or 'just like I said' are markers that you are scoring evidence rather than weighing it.

• Why day traders lose money

  The hub piece: how confirmation bias keeps a deteriorating position feeling justified, alongside the six other biases behind the failure rate.

• Anchoring bias

  Judging value against an irrelevant reference point, usually a past price or the first number you saw.

• Loss aversion

  Losses feel roughly twice as heavy as equivalent gains, which makes exits disproportionately painful.

• Sunk cost fallacy

  Continuing because of what you have already spent, rather than what lies ahead.

• Recency bias

  Over-weighting the most recent event and treating a single data point as a pattern.

The Bias Checker screens for all of them at once, and the articles go deeper on each.