Confirmation bias

Confirmation bias is the tendency to treat evidence as a lawyer for what we already believe rather than a test of whether we are right. It shapes what we search for, notice, remember, and count as convincing — often without feeling like bias at all.

WikipediaKnowledge quiz Self assessment Microcourse

Mechanism

How it works

A prior belief makes supporting evidence feel relevant and contradictory evidence feel flawed, exceptional, or beside the point. The bias enters before a conclusion is reached: in the questions we ask, sources we select, tests we run, and examples we remember. That makes it especially hard to catch through introspection alone.

Examples

Where it shows up

A believer remembers the three horoscope predictions that seemed accurate and never records the many that failed.
A manager searches for evidence that a launch will work, but never asks what result would make the team cancel it.
A researcher keeps adjusting an analysis until the expected pattern appears, then treats the successful analysis as if it were the only one tried.

Consequences

What it can distort

Weak ideas can survive repeated contact with evidence because the evidence is filtered before it can update the belief.
Teams become overconfident in a plan precisely because their research process has quietly excluded the information that would have changed it.

Countermeasures

How to work around it

Write down what evidence would change your mind before gathering evidence; if nothing would, you're not investigating, you're advocating.
Search for disconfirmation directly: phrase queries and experiments so a negative result is informative ('what would kill this idea fastest?').
Assign someone the explicit job of building the case against, with real stakes for finding flaws.
Track your prediction hit-rate; calibration data is the only reliable mirror for this bias.

Caveats

Critiques and limits

Seeking confirming evidence is not always irrational: a mature theory may deserve more tests of its implications. The error is treating support as diagnostic while failing to seek evidence that could distinguish the theory from its alternatives.

Taxonomy

Fields of impact

Political Science & Public Policy

Psychology & Social Science

Economics & Finance

Evidence

How solid is the research?

Robust — replicates reliably

Documented across hypothesis testing, evidence evaluation, and memory search since Wason's 1960s studies; Nickerson's 1998 review catalogs its many replicated guises.

Research

Relevant papers

Confirmation bias: A ubiquitous phenomenon in many guises

Nickerson, R. S. (1998)

Review of General Psychology, 2(2), 175-220

Bias in Human Reasoning: Causes and Consequences

Evans, J. St. B. T. (1989)

Psychology Press

Confirmation, disconfirmation, and information in hypothesis testing

Klayman, J., & Ha, Y.-W. (1987)

Psychological Review, 94(2), 211-228

Case studies

Real-world patterns.

Real-world examples showing how Confirmation bias manifests in practice

Case study

SmartSync and the Launch Everyone Wanted to Believe

A real-world example of Confirmation bias in action

Context

A mid-size SaaS company built a new feature, SmartSync, marketed internally as a game-changer for customer retention after an evangelizing product demo and positive early feedback from a small pilot. Leadership grew confident the feature would reduce churn and greenlit a full rollout without broad, segmented analysis. The company was already under pressure to show growth and reduce churn to hit quarterly targets.

Situation

Product, marketing, and executive teams pointed to the pilot's qualitative feedback and a short-term rise in engagement as evidence SmartSync would improve retention across the board. Because the initial pilot users were early-adopter power users, the team equated their behavior with the whole customer base. Engineers and analysts were asked to prepare launch materials quickly, and the feature was released to all customers three weeks after the pilot.

The bias in action

Team members selectively highlighted metrics that fit the desired narrative (short-term session length and NPS responses from pilot users) while downplaying or ignoring cohort-level retention analyses that showed deterioration among mid-tier customers. Analysts ran exploratory queries but prioritized charts that showed positive trends in the pilot segment; contradictory subgroup results were placed in appendices or dismissed as 'noise.' When a junior analyst raised concerns about a rising churn signal in newer customers, the comment was framed as an outlier rather than investigated. The prevailing belief that SmartSync would reduce churn shaped the interpretation of every subsequent data pull.

Outcome

Within two quarters of the full rollout, overall monthly churn rose from 6.2% to 8.0% — a 1.8 percentage-point increase — concentrated in mid-tier and new customers who constituted 28% of the user base. Customer support volumes for sync-related issues increased by 30% in the first quarter post-launch. The company estimated an ARR impact of approximately $420,000 and had to delay two planned marketing initiatives to cover remediation costs.

What's inside the full case study

Unlock the deeper breakdown with real-world impact, measurable effects, lessons learned, better-approach recommendations, and relevant fields.

Real-world impact

Affected groups, timeframe, and measurable outcomes.

Lessons learned

Practical takeaways and a better path forward.

Full case breakdownEmail access

Want the full analysis?

Request access to the complete case study, including measurable impact, lessons learned, and the recommended better approach.

Recommended books

Mistakes Were Made (But Not by Me)

Carol Tavris, Elliot Aronson, 2007

The Scout Mindset

Julia Galef, 2021

Judgment under Uncertainty: Heuristics and Biases

Daniel Kahneman, Paul Slovic, Amos Tversky, 1982

Thinking, Fast and Slow

Daniel Kahneman, 2011

The Righteous Mind

Jonathan Haidt, 2012

Related biases

Nearby patterns.

Motivated reasoning

Motivated reasoning supplies the goal (reach the desired conclusion); confirmation bias is the search-and-evaluation machinery that gets it there.

Belief perseverance

Belief perseverance is the tendency to maintain a belief even after the evidence that originally supported it has been completely discredited.

Congruence bias

Congruence bias is confirmation bias in experiment design: testing only the hypothesis you have.

Selective perception

Selective perception is noticing, interpreting, and remembering information through a filter shaped by what you already expect or believe..

Backfire effect

The backfire effect is the proposed possibility that a correction makes a false belief stronger rather than weaker.

Illusory correlation

Illusory correlation is seeing a meaningful relationship between two things because their co-occurrences are vivid, even when the full data do not support one..

Study on Microcourse

Learn the wider pattern.

Dive deeper into Confirmation bias and related biases in Reasoning and Logical Fallacieswith structured lessons, examples, and practice exercises.

Study

Practice

Test your knowledge.

Apply what you have learned and reinforce your understanding of Confirmation bias with a short quiz or self-assessment.

Understanding quiz Self-assessment

Entry last reviewed 2026-07-16 · sources verified against the published literature — methodology