Why Do We Defend Our Beliefs Even When We Are Wrong?
Why can contrary evidence trigger stronger defense? A synthesis of confirmation bias, motivated reasoning, cognitive dissonance, self-protection and group identity.
A person can defend a belief firmly for good reasons. If the evidence strongly supports an explanation, it would not be rational to abandon it at every objection. The problem begins elsewhere: when we change the standard of evidence depending on whether we like the conclusion, scrutinize contrary information more harshly than supportive information, or experience correction as a threat to who we are.
This article therefore does not search for a single 'bias' that explains everything. It synthesizes several research traditions—confirmation bias, motivated reasoning, cognitive dissonance, self-protective and identity-protective cognition—and asks why admitting error can sometimes become psychologically, socially and morally more costly than preserving the old answer.
The central rule is simple: the strength of a belief does not tell us whether it is true; the way we handle contrary evidence tells us a great deal about the quality of our reasoning.
A belief is not only a sentence in the head
Beliefs are models with which we predict the world, choose actions and interpret events. Some are small and cheap to revise: which route to the shop is faster or when a bus leaves. Others are tied to our moral self-image, family, group, profession, life decisions or the story of what kind of person we are.
The same correction can therefore carry very different weight. If I am wrong about the weather forecast, little is lost. But if new information would mean that I defended a bad decision for years, hurt someone close to me, trusted the wrong institution or publicly committed myself to a claim, correction may carry additional costs: shame, status loss, a sense of betraying a group, or the need to reinterpret my own past.
This does not show that a deeply personal belief is false. It explains why the epistemic question 'what does the evidence show?' can become, at the same time, the psychological question 'what will it mean for me if this evidence is right?'
In his classic review, Raymond Nickerson showed that confirmation bias covers several related phenomena: searching for information that supports a current hypothesis, interpreting ambiguous data selectively, retrieving supportive examples more readily and neglecting possibilities that could falsify the belief.
In ordinary life, a person can sincerely feel that they are 'just looking at the facts' while the query they typed into a search engine, the people they chose to ask and the examples that came to mind were already tilted in one direction. Bias can begin before the evidence is explicitly evaluated.
From Belief to Verification already showed that belief is not evidence. This article adds a further question: gathering a large amount of material is not enough. We must ask whether our search was designed so that we could actually encounter something that contradicts us.
The classic experiment by Lord, Ross and Lepper demonstrated biased assimilation: participants with strong prior attitudes emphasized methodological weaknesses more strongly in a study that challenged their view than in a study that supported it. Mixed evidence could therefore strengthen initial positions.
Related work on motivated skepticism shows that people may accept congenial arguments with relatively little scrutiny while generating more counterarguments against uncongenial ones. The problem is not skepticism itself. Skepticism is essential. The problem is a double standard: 'Can I believe this?' for a desired conclusion and 'Must I believe this?' for an undesired one.
This creates one of the best practical tests of our own judgment: would I use the same methodological standard if the result supported the opposing side?
Motivated reasoning does not mean consciously lying to yourself
Ziva Kunda described motivated reasoning as a process in which goals influence which memories, rules and explanations are accessed and how information is evaluated. An important constraint in her account is often forgotten: people generally cannot believe anything they simply wish to believe. A desired conclusion must be supportable by reasons that appear plausible enough to them.
Motivated reasoning therefore need not sound internally like 'I know this is wrong, but I will lie anyway.' It is often more like a lawyer who is genuinely skilled at building the strongest case for one side: from many real details, the mind selects those that support the preferred explanation and treats the rest as less important or less reliable.
The alternative is not reasoning without motivation. Kunda also distinguishes accuracy motivation: when arriving at the best answer truly matters, motivation can increase care, information search and checking. The crucial question is which goal is steering the process.
Cognitive dissonance theory, developed by Leon Festinger and repeatedly refined in later research, describes the aversive tension that can arise when important beliefs, actions, decisions or aspects of self-concept conflict. If I regard myself as honest but discover that I behaved dishonestly, there is a discrepancy that demands resolution.
Dissonance can be reduced in several ways. I can change my behaviour, revise the belief, admit the mistake and try to repair it. Or I can minimize the event, add a justification, reinterpret what happened, or redirect doubt toward the source of the uncomfortable information. The same psychological tension therefore does not inevitably produce denial; it can also produce learning.
An important boundary follows: 'they are experiencing dissonance' is not a diagnosis we can assign from a distance to everyone who disagrees with us. The theory describes mechanisms that must be connected to specific conditions and behaviour.
Admitting error can threaten the self
Some beliefs are intertwined with feeling competent, moral or consistent. If a correction sounds like 'you were not merely wrong about a datum; you are stupid, bad or worthless,' the exchange shifts from evaluating a claim to defending the self.
Research on self-affirmation has found that affirming broader personal values can, under some conditions, reduce defensive processing of threatening information. The point is not to reassure ourselves that we are always right. It is to prevent one error from becoming a referendum on our entire worth.
Self-affirmation effects are not universal and depend on timing, context and emotional state. But the broader lesson is useful: belief revision becomes easier when 'my claim was wrong' can be separated from 'I failed as a person.'
A belief can signal belonging. In that case, changing one's mind is not only an intellectual revision; it may create conflict with people on whom we depend or feel like leaving a community. Research on cultural or identity-protective cognition describes how people can selectively credit information in ways that protect important group identities.
This mechanism matters especially on polarized issues, but it should not become a universal label. A religious, political, scientific or anti-establishment position is not, by itself, evidence of identity protection. The question is whether group attachment systematically changes the evidential standard.
When the social cost of conceding error is high, simply adding more facts may not solve the problem. A person is no longer evaluating only what is probably true, but also what may happen to them if they acknowledge it publicly.
Bias does not belong to one ideology or group
This article deliberately avoids the convenient story in which 'biased people' are always the others. A meta-analysis of 51 experimental studies found comparable average effects among liberal and conservative U.S. participants for one specific form of partisan bias. This does not imply that every ideology and every bias is always symmetrical; it does show why assuming our own immunity is dangerous.
The same logic applies outside politics. A scientist may defend a favored model, an entrepreneur an investment, a parent a parenting decision, a partner an interpretation of a conflict, and a researcher of alternative topics a theory into which years of work have been invested. The mechanism is human; the content changes.
The safest starting assumption is therefore not 'both sides are equally wrong' but: every side, including mine, needs a procedure capable of detecting its own error.
Good reasoning, expertise and numerical skill are valuable, but they are not magical protection. In work on myside bias, Stanovich and West found that the tendency to evaluate arguments from one's own side could be relatively independent of measured cognitive ability on the tasks they studied.
This does not establish the popular slogan that 'smarter people are more biased.' Higher ability often improves problem solving. But the same ability can provide more tools for constructing a sophisticated defense when the goal is to protect a preferred conclusion rather than test it.
That is why intellectual stance matters alongside ability: willingness to search for contrary reasons, postpone closure and state what evidence would change one's estimate.
Prior belief is not automatically bias
All sensible reasoning uses prior knowledge. If decades of high-quality evidence support a method, it is rational not to give one weak contrary report the same weight as the larger evidence base. Maintaining a well-supported belief is not the same as confirmation bias.
The distinction lies in the updating rules. Do we evaluate evidence by quality, independence, effect size, possible error and coherence with other data? Or does methodological rigor suddenly become important only when we dislike the result? How Do We Know What We Know? Evidence, Belief and Certainty framed this as calibration of evidential strength, while Where Does Fact End and Interpretation Begin? separated observations and sources from interpretation.
A good belief can therefore be very stable while still knowing its conditions for revision. Dogmatism is not simply high confidence; it begins when virtually no possible outcome is allowed to count against the explanation.
Lord, Ross and Lepper produced a famous example in which people became more confident in their initial views after reading mixed research. The finding became an icon of 'polarization through evidence.' But a single experimental pattern should not be turned into a law of nature.
The outcome depends on prior attitudes, the issue, evidence quality, presentation and measurement. Faced with contrary information, people sometimes polarize, sometimes barely move, and sometimes shift toward the better evidence.
This matters methodologically: this article does not use psychology as an excuse to declare every resistance irrational in advance. We have to examine what the person actually did with the evidence.
Corrections are not usually doomed to 'backfire'
A popular idea holds that correcting a false belief often produces the opposite result and makes people even more committed. This backfire effect was observed in some early studies, but later, larger work has not supported it as a general or common outcome.
Wood and Porter tested 52 potential cases in five experiments with more than ten thousand participants and found no corrections that systematically produced factual backfire. More recent reviews of misinformation similarly conclude that corrections usually improve belief accuracy, even though they are not perfectly effective.
A more realistic problem is often the continued influence of misinformation: a person may accept the correction while the original story still partly shapes later reasoning. Effective correction therefore does more than stamp a claim 'false'; it clearly states what is correct and, where possible, fills the causal gap with a better explanation.
In a romantic, family or friendship conflict, a belief is often not merely a description of an event but also a defense of intention: 'I am not that kind of person,' 'you are always the one who starts it,' 'if I admit this, everything becomes my fault.' Here research on self-concept meets the practical relationship psychology developed in the project Ljubiti brez samouničenja.
If admitting one error means total defeat in the relationship, everyone has a strong incentive to persist. A safer relationship makes room for the sentence 'I was wrong about this' without the automatic conclusion 'therefore I have no voice, worth or right to a boundary.'
This is not relativism. Facts remain facts and harm remains harm. It means that an environment in which correction is possible without total identity destruction increases the chance of an honest admission of error.
How to build a belief that can be revised
The first protection is an advance criterion. Before we know the result, write down what evidence would increase confidence, what would reduce it and what would seriously threaten the theory. This makes it harder to move the goalposts after the outcome is known.
The second protection is deliberately searching for the strongest contrary case. 'Consider the opposite' experiments found that asking people to seriously construct an alternative explanation can reduce some judgment biases. This is not a ritual of 'both sides.' The opposing case should be the strongest reasonable one, not a caricature.
The third protection is separating identity from proposition: 'my current estimate is X' is epistemically safer than 'I am a person who believes X.' We can add probabilities instead of absolute certainty, check source independence, ask someone to look for holes, and keep a record of changes in view. The goal is not to have no beliefs, but to have beliefs with an updating mechanism.
If our aim is to represent reality as accurately as possible, changing a belief in response to better evidence is a success of the process. The error was in the old model; the capacity to correct it shows that the model has not become a prison.
This becomes harder when years of public advocacy, money, reputation, group membership or moral self-image are at stake. That is why intellectual honesty is more than an argumentative technique. It includes the capacity to bear the psychological cost of saying: 'I understood this wrongly.'
The most dangerous belief is therefore not necessarily the strongest one. It becomes dangerous when it has an answer for every possible observation, an explanation for every contradiction, and a social system in which revision equals betrayal. A good belief can defend itself; more importantly, it can change.
Sources and further reading
- Nickerson, R. S. (1998). Confirmation Bias: A Ubiquitous Phenomenon in Many Guises. Review of General Psychology 2(2), 175–220 — broad review of selective search, interpretation and evidence testing that favor prior beliefs.
- Kunda, Z. (1990). The case for motivated reasoning. Psychological Bulletin 108(3), 480–498 — classic account distinguishing accuracy goals from directional goals and emphasizing that desired conclusions remain constrained by plausible justification.
- Lord, C. G., Ross, L. & Lepper, M. R. (1979). Biased assimilation and attitude polarization: The effects of prior theories on subsequently considered evidence. Journal of Personality and Social Psychology 37(11), 2098–2109 — classic experiment on asymmetric evaluation of congenial and uncongenial evidence.
- Taber, C. S. & Lodge, M. (2006). Motivated Skepticism in the Evaluation of Political Beliefs. American Journal of Political Science 50(3), 755–769 — experimental evidence for prior-attitude, disconfirmation and selective-exposure effects in polarized political judgments; used as a domain-specific example, not a universal model of politics.
- Festinger, L. (1957). A Theory of Cognitive Dissonance. Stanford University Press — foundational formulation of dissonance arising from psychologically relevant inconsistency.
- Harmon-Jones, E. (Ed.). (2019). Cognitive Dissonance: Reexamining a Pivotal Theory in Psychology, 2nd ed. American Psychological Association — modern overview of dissonance research, competing refinements and links to self, action and affect.
- Sherman, D. K. & Cohen, G. L. (2006). The Psychology of Self-Defense: Self-Affirmation Theory. Advances in Experimental Social Psychology 38, 183–242 — review of self-integrity threat, defensive information processing and conditions under which self-affirmation can reduce defensiveness.
- Critcher, C. R., Dunning, D. & Armor, D. A. (2010). When Self-Affirmations Reduce Defensiveness: Timing Is Key. Personality and Social Psychology Bulletin 36(7), 947–959 — evidence that self-affirmation effects depend on timing relative to defensive responding.
- Kahan, D. M., Braman, D., Gastil, J., Slovic, P. & Mertz, C. K. (2007). Culture and Identity-Protective Cognition: Explaining the White-Male Effect in Risk Perception. Journal of Empirical Legal Studies 4(3), 465–505 — influential empirical formulation of identity-protective cognition in risk perception; treated as a specific research program, not a universal diagnosis.
- Ditto, P. H. et al. (2019). At Least Bias Is Bipartisan: A Meta-Analytic Comparison of Partisan Bias in Liberals and Conservatives. Perspectives on Psychological Science 14(2), 273–291 — meta-analysis of 51 experimental studies finding comparable average levels for one operationalized form of partisan bias; not evidence that all ideological biases are symmetric.
- Stanovich, K. E. & West, R. F. (2007). Natural myside bias is independent of cognitive ability. Thinking & Reasoning 13(3), 225–247 — evidence that natural myside bias on studied tasks can persist independently of measured cognitive ability.
- Lewandowsky, S., Ecker, U. K. H., Seifert, C. M., Schwarz, N. & Cook, J. (2012). Misinformation and Its Correction: Continued Influence and Successful Debiasing. Psychological Science in the Public Interest 13(3), 106–131 — review of why misinformation can continue to influence reasoning even after correction and how corrections can be improved.
- Ecker, U. K. H. et al. (2022). The psychological drivers of misinformation belief and its resistance to correction. Nature Reviews Psychology 1, 13–29 — contemporary review emphasizing cognitive, social and affective drivers and evidence-based correction strategies.
- Wood, T. & Porter, E. (2019). The Elusive Backfire Effect: Mass Attitudes' Steadfast Factual Adherence. Political Behavior 41, 135–163 — five experiments testing 52 issues and finding no systematic factual backfire, challenging claims that corrections commonly worsen false beliefs.
- Lord, C. G., Lepper, M. R. & Preston, E. (1984). Considering the opposite: A corrective strategy for social judgment. Journal of Personality and Social Psychology 47(6), 1231–1243 — experimental evidence that deliberately generating alternative possibilities can reduce some forms of biased social judgment.
- Stanovich, K. E. & Toplak, M. E. (2023). Actively Open-Minded Thinking and Its Measurement. Journal of Intelligence 11(2), 27 — review and measurement framework for willingness to consider alternatives, attend to contradictory evidence and postpone closure.