Confirmation Bias
Confirmation Bias
What is confirmation bias?
- A cognitive bias that makes you see only what you already believe
- Works in search, interpretation and memory alike
- Mitigate with disconfirming questions and a premortem
Confirmation bias is less a problem of 'holding wrong beliefs' than of 'a verification process that only runs in one direction.' So what actually changes in an organization that takes this concept on board is not people’s mindset but the order of the meeting. Instead of gathering evidence and then stating a conclusion, the team first writes the conclusion in one sentence, defines the conditions under which that sentence would fail, and only then opens the data. A mechanism that reverses the direction of the burden of proof (accountability) is what reduces bias; a resolution to 'be objective' rarely changes a judgment on its own.
In practice it is often lumped together with anchoring, the availability heuristic, and groupthink, but each needs a different fix. Anchoring is judgment being pulled toward the first number presented; the availability heuristic is overestimating recent, vivid cases; groupthink is swallowing dissent to preserve harmony. The drive to defend a conclusion already held is the core of confirmation bias, so it tends to get worse as information increases. More data also means more evidence that happens to suit your view.
Bias gets especially strong under three conditions: someone has already publicly backed the idea in an executive meeting, so much budget and staff are committed that backing out is hard, or a tight deadline pushes everyone to reach a conclusion fast. Safeguards, in turn, only work if they are set up at the right moment. Disconfirmation criteria must be put in writing before anyone looks at the data, and the person playing the opposing role must face no penalty in their evaluation. Criteria first, data later: once that order breaks, the disconfirmation criteria themselves get loosened to fit the conclusion.
The most common failure in the field is setting up a devil’s advocate in form only. If one person is assigned the opposing role, is heard for 20 minutes, and the meeting moves on, the minutes become an alibi that says 'we even considered the opposing view,' which actually strengthens conviction. Another is the illusion that 'I’ve been trained, so I’m fine.' People who have learned about bias tend to spot it in others while staying blind to their own. Training alone does not reduce bias; the effect lasts only when it is built into the decision process, as in interview question sheets or investment review checklists.
Ara LimExpert in global leadership, AI, and business communication
John KimInstructor John Kim, author of B2B Marketing Design and CEO of Nanum ConsultingThe concept was established when British psychologist Peter Wason, in a 1960 rule-discovery experiment (the '2-4-6 task'), found that people tend to confirm their hypotheses rather than try to disprove them. It has since become one of the best-known cognitive biases through research on judgment and decision-making and behavioral economics.
Consider a hypothetical case. Company A, an industrial filter manufacturer (240 employees, KRW 78 billion (≈ US$58,357,000) in annual revenue), spent six weeks evaluating a move into a new maintenance rental business. Three years earlier it had put KRW 1.8 billion (≈ US$1,346,700) into a similar service business and then pulled out. This time, again, the only market data reaching the meeting was prepared by the people pushing for the project. To reduce the risk of repeating the same mistake, management started by reworking the review process.
- In the first meeting, the head of strategy wrote 'This business will generate KRW 15 billion (≈ US$11,222,500) in revenue within three years' on the whiteboard, then proposed deciding first which numbers would prove that sentence wrong.
- The finance lead wrote two lines into the meeting minutes before opening any market data: 'We stop if customer acquisition cost exceeds KRW 400,000 (≈ US$300) or 12-month retention falls below 70%.'
- The business development team designated 2 section managers as dedicated devil’s advocates and had them collect only 18 churned-customer interviews over 2 weeks. The CEO drew the line: 'this team is judged only on the quality of its counter-evidence.'
- In week 4, the 11 attendees imagined that 'this business has failed one year from now' and each wrote down three reasons why. Of the 33 responses, 14 clustered around a single issue: the supply of maintenance technicians.
- When the measured customer acquisition cost came in at KRW 520,000, (≈ US$390) Company A dropped the rental model and narrowed its scope to maintenance contracts with existing customers, recovering KRW 1.7 billion (≈ US$1,271,900) of the KRW 2.4 billion (≈ US$1,795,600) budget it had planned to invest.