Why Conspiracy Believers Think Everyone Agrees With Them: The Psychology of Overconfidence Behind Conspiracy Theories

The most surprising thing about many conspiracy believers may not be what they believe.

It may be how certain they are that everybody else believes it too.

A major psychological study involving more than 4,000 U.S. adults found that people who endorsed conspiracy theories consistently overestimated their own cognitive performance.

They thought they were better at numerical reasoning than their actual scores suggested.

They overestimated how well they performed on difficult perception tasks.

And the more fringe the conspiracy belief, the stronger the relationship with this overconfidence tended to become.

Then the researchers asked another question.

How many other people do you think agree with you?

The answer revealed an even larger distortion.

Across three of the studies, the conspiracy claims tested were actually supported by a majority of participants only 12% of the time.

Yet people who believed those claims thought they themselves were on the majority side 93% of the time.

In other words, many believers were not thinking:

“Almost everyone disagrees with me, but I know a secret truth.”

They appeared to be thinking:

“Most people basically see this the way I do.”

That difference matters.

It suggests that conspiracy belief may sometimes involve something deeper than exposure to misleading Facebook posts, political anger or the desire to feel special.

It may involve a basic problem of cognitive calibration:

How accurately do I know what I know?

How accurately can I judge my own ability?

And do I recognize when the evidence should make me less confident?

The study, titled Overconfidently Conspiratorial: Conspiracy Believers are Dispositionally Overconfident and Massively Overestimate How Much Others Agree With Them, was written by Gordon Pennycook, Jabin Binnendyk and David G. Rand.

It was first published online in Personality and Social Psychology Bulletin on May 24, 2025 and appears in the journal's August 2026 issue.

Its conclusion is provocative.

But it requires one important correction before we go any further.

The study does not prove that social-media algorithms are irrelevant to conspiracy theories.

Nor does it demonstrate that overconfidence is the single cause of conspiratorial thinking.

What it does show is something arguably more interesting:

People who believe conspiracy theories appear unusually likely to be overconfident in their own judgments—and extraordinarily poor at recognizing how socially fringe some of those beliefs actually are.

That could help explain why correcting conspiracy theories is often so difficult.

The Study Included 4,181 U.S. Adults Across Eight Experiments

The researchers did not base their conclusions on one survey.

They conducted eight studies involving 4,181 adults in the United States.

The central idea was relatively simple.

Most conspiracy research had previously focused heavily on psychological needs and motivations.

Researchers had examined whether people are attracted to conspiracy theories because they:

• Want certainty
• Feel powerless
• Distrust institutions
• Want to feel unique
• Have narcissistic tendencies
• Need control
• Prefer intuitive rather than analytical thinking

Pennycook and colleagues proposed an additional possibility.

Perhaps some people are simply too confident in the accuracy of their own reasoning.

Not confident because they actually perform better.

Confident beyond what their performance justifies.

That distinction is the heart of the paper.

Confidence Is Not the Same as Ability

Imagine two people take a 10-question reasoning test.

Person A gets eight answers correct and predicts that they probably got seven or eight correct.

Person B gets four correct but believes they got nine correct.

Person B does not merely have weaker performance.

They are miscalibrated.

Their confidence does not accurately track their ability.

This is what psychologists mean when they discuss overconfidence in this context.

The important measurement is therefore not:

How smart are you?

It is:

How accurately can you estimate how well you performed?

That difference allowed the researchers to separate cognitive ability from confidence about cognitive ability.

Conspiracy Believers Overestimated Their Numeracy Performance

Participants completed numerical reasoning tasks.

The researchers then compared how well participants actually performed with how well they believed they had performed.

People who endorsed conspiracy theories were more likely to overrate their own performance.

Crucially, the association remained even after the researchers statistically accounted for actual performance.

That means the finding was not simply:

“People who perform poorly believe more conspiracies.”

The more interesting result was:

Conspiracy belief was associated with thinking you performed better than you actually did.

That is a calibration problem.

The Researchers Then Used a Perception Task

Numeracy creates an obvious complication.

Perhaps people who believe conspiracy theories simply have different mathematical abilities.

So the researchers also used a difficult perceptual task.

Participants viewed images that were intentionally difficult to identify and then estimated how well they had performed.

The task was designed so that actual performance depended heavily on guessing.

Yet people who believed conspiracy theories were still more likely to believe that they had performed unusually well.

This was important because it suggested that the phenomenon extended beyond one particular skill such as mathematics.

The researchers interpreted the pattern as evidence of dispositional overconfidence:

A broader tendency to overestimate one's abilities.

The Relationship Was Strongest for Fringe Conspiracies

One of the study's most revealing findings concerns the popularity of the conspiracy theory itself.

Overconfidence was especially strongly associated with belief in the least widely accepted conspiracy claims.

That makes intuitive sense.

If a belief is accepted by almost everyone, endorsing it requires little confidence in your own judgment against the crowd.

But imagine believing something rejected by nearly everyone around you.

To maintain that belief, you may need to believe that:

Experts are wrong.

Institutions are wrong.

Journalists are wrong.

Most researchers are wrong.

People challenging you are misinformed.

And somehow you have seen through what they have not.

That requires a substantial degree of confidence in one's own ability to distinguish truth from error.

The new research suggests that, for some believers, that confidence may be much greater than their demonstrated judgment warrants.

Then Came the 93% Finding

The researchers also examined what psychologists call the false consensus effect.

Humans generally tend to overestimate how common their own beliefs, preferences and behaviors are.

We use ourselves and the people around us as a rough model of society.

If most of our friends think something, it can begin to feel as though most people think it.

But among conspiracy believers, the distortion observed in this study was enormous.

Across the conspiracy claims examined in three studies, only 12% were genuinely majority beliefs within the participant samples.

Nevertheless, believers thought their own position represented the majority 93% of the time.

That does not mean believers thought precisely 93% of Americans agreed with every conspiracy.

The statistic means that in 93% of cases, believers estimated that their own position was on the majority side.

That is an important distinction.

They Overestimated Agreement by More Than Fourfold

The study's abstract describes believers—particularly the most overconfident ones—as overestimating agreement by more than four times.

That is not a small polling error.

Imagine supporting an idea accepted by 15% of people while believing 60% or 70% agree with you.

Your entire understanding of the social environment changes.

You stop thinking of yourself as holding an unusual position.

Instead, you may begin wondering why institutions supposedly refuse to acknowledge what “everyone knows.”

That can create fertile ground for further conspiratorial reasoning.

They Even Thought Political Opponents Agreed More Than They Did

Perhaps the effect could be explained simply by political echo chambers.

A participant thinks most people agree because their friends, family and political allies agree.

The researchers tested something close to this possibility.

Participants were asked to estimate agreement among people belonging to the opposing political camp.

The overestimation persisted.

That finding is particularly important.

It suggests that at least some believers were not merely saying:

“People like me believe this.”

They seemed to genuinely misunderstand the overall distribution of opinion.

Cornell's summary of the research described them as largely unaware that their beliefs were on the fringe.

This Is More Than Wanting to Feel Unique

An older theory proposed almost the opposite dynamic.

Perhaps conspiracy believers are attracted to fringe ideas precisely because being different feels good.

Knowing something that “the masses” do not know can create a sense of intellectual uniqueness.

Previous research has indeed linked conspiracy belief with a need for uniqueness, as well as narcissistic traits and other social motivations.

But the new study complicates that explanation.

If believers desperately wanted to perceive themselves as members of a tiny enlightened minority, we might expect them to recognize that their beliefs were unpopular.

Instead, they overwhelmingly believed they were in the majority.

And the association between conspiracy belief and overconfidence remained even after statistically controlling for:

• Need for uniqueness
• Narcissism
• Analytical thinking ability

That does not mean those earlier factors are irrelevant.

It means overconfidence appears to contribute something additional.

Does This Mean Narcissism Does Not Matter?

No.

A robust literature still links narcissism with conspiracy belief.

A 2022 review concluded that narcissistic traits can make conspiracy theories appealing through several mechanisms, including paranoia, gullibility, dominance motives, perceived superiority and the desire for uniqueness.

A 2026 study also found that narcissism and the need for uniqueness could weaken the normally protective relationship between education and conspiracy belief.

Human psychology rarely contains one cause.

A person can simultaneously:

Feel powerless.

Want certainty.

Distrust institutions.

Desire uniqueness.

Be surrounded by conspiratorial peers.

And be overconfident in their reasoning.

These forces can reinforce one another.

The new research adds another piece to that system.

It does not erase the others.

Earlier Research Identified Three Major Motivational Families

A highly influential review by Karen Douglas, Robbie Sutton and Aleksandra Cichocka grouped motivations for conspiracy belief into three broad categories.

Epistemic motives

People want to understand events.

Uncertainty is uncomfortable.

Conspiracy theories can provide clear explanations for complicated or frightening situations.

Existential motives

People want security and control.

A hidden enemy can sometimes feel psychologically easier to process than random chaos.

Social motives

People want themselves and their groups to feel competent and morally valuable.

Conspiracy explanations can shift blame toward supposedly powerful outsiders.

The authors also noted something paradoxical.

Conspiracy theories may promise to satisfy these needs without actually doing so.

Instead, they may deepen distrust and alienation.

The overconfidence findings fit surprisingly well into this framework.

If you desperately want an explanation and are unusually confident that your interpretation is correct, uncertainty disappears very quickly.

Possibly too quickly.

The Dangerous Question Is Not “How Much Do You Know?”

It is:

How good are you at recognizing when you might be wrong?

A person can possess substantial factual knowledge and still reason badly if confidence is poorly calibrated.

Likewise, someone with limited expertise can make sound decisions if they accurately recognize the limits of their knowledge and consult better sources when necessary.

This is the value of epistemic humility.

Epistemic humility does not mean:

“I know nothing.”

It means:

“My confidence should depend on the quality of the evidence.”

That is very different.

Overconfidence Can Make Counterevidence Look Like Proof

Suppose you believe a government agency is concealing something.

A scientist provides evidence contradicting your theory.

There are at least two possible reactions.

Calibrated reaction

“Perhaps I should reconsider my belief.”

Overconfident reaction

“Of course the scientist denies it. That proves how deep the cover-up goes.”

The second response creates a self-sealing belief.

Evidence against the conspiracy becomes evidence for the conspiracy.

Once that structure develops, factual correction becomes extraordinarily difficult.

Not because information is unavailable.

Because the person's confidence system no longer gives contradictory information sufficient weight.

Conspiracy Theories Can Become Epistemically Self-Protecting

Many fringe theories contain built-in mechanisms that explain away criticism.

No evidence?

The evidence was destroyed.

Experts disagree?

The experts are compromised.

Journalists debunk it?

The media are part of the conspiracy.

Most people reject it?

They are brainwashed.

A predicted event does not happen?

The conspirators changed their plan.

This makes the theory difficult to falsify.

Combine that structure with overconfidence and the psychological defense becomes extremely strong.

A person does not merely possess a questionable belief.

They possess a framework that protects the belief from correction.

Is This Just the Dunning-Kruger Effect?

The comparison is tempting.

The famous Dunning-Kruger literature concerns situations in which people with lower competence may lack the knowledge needed to recognize their own mistakes.

But it would be too simplistic to label the new conspiracy findings merely “the Dunning-Kruger effect.”

Pennycook and colleagues specifically measured overconfidence relative to actual performance and interpreted the pattern as dispositional overconfidence.

The study was not simply a replication of the classic Dunning-Kruger paradigm.

The broader family resemblance is obvious:

Both concern miscalibration between perceived and actual ability.

But the exact constructs and methods are not identical.

The Study Does Not Prove Overconfidence Causes Conspiracy Belief

This is one of the most important limitations.

The authors describe overconfidence as a possible “driver” of conspiracy belief.

But the evidence primarily demonstrates a strong relationship between the two.

Correlation does not tell us the direction of causation with certainty.

Several possibilities remain.

Overconfidence could encourage conspiracy belief

People who trust their judgment too much may more readily dismiss expert consensus.

Conspiracy belief could increase overconfidence

Becoming immersed in narratives about possessing hidden knowledge could strengthen perceived intellectual superiority.

A third factor could influence both

For example, certain reasoning styles, information environments or personality characteristics might simultaneously increase overconfidence and conspiracy belief.

The relationship could be reciprocal

Overconfidence encourages conspiratorial thinking.

Conspiratorial communities reinforce overconfidence.

The cycle strengthens.

The eight-study design makes the association compelling.

It does not turn psychology into a one-direction causal equation.

And No, the Study Does Not Show Social Media Is Innocent

This is where some interpretations go too far.

The research focused on individual cognitive characteristics.

It did not experimentally compare societies with and without recommendation algorithms.

It therefore cannot establish that social media played no role in the spread of conspiracy theories.

Research on social platforms shows that conspiracy narratives can circulate inside polarized communities and echo chambers, with recommendation systems and moderation structures potentially affecting their visibility.

A 2023 audit of Twitter's recommendation system found algorithmic amplification of content from users within similar communities and greater amplification of emotionally charged and toxic posts.

Search-engine audits have also found that some systems return considerably more conspiracy-promoting material than others.

Algorithms are therefore part of the information environment.

The new Cornell-led research addresses a different question:

Why might some people be especially receptive to conspiratorial claims once they encounter them?

People Also Select Their Own Information

At the same time, blaming algorithms for everything is equally simplistic.

Users are not passive objects whose beliefs are entirely manufactured by recommendation engines.

Research examining online news behavior has found substantial evidence that people actively choose unreliable or ideologically congenial information rather than merely being forced into it by search algorithms.

A 2026 analysis of approximately 15 million tweets also found that the composition of a person's social network predicted conspiracy-theory dissemination: even relatively small exposure to conspiracy-spreading neighbors was associated with substantially greater likelihood of spreading such material.

So the better model is not:

Algorithms versus psychology.

It is:

Psychology + social networks + selective exposure + algorithms + incentives + social reinforcement.

Social Media May Amplify What Human Psychology Already Supplies

Consider an overconfident person who believes a fringe theory.

Before social media, they might have had difficulty finding hundreds of people who agreed.

Today they can search instantly.

An online group appears.

Thousands of posts repeat the claim.

Videos explain it.

Comments applaud it.

People who disagree leave or are removed.

Now the believer experiences something psychologically powerful:

social proof.

The original belief did not necessarily originate with the algorithm.

But the platform can make the belief feel more normal.

That is particularly important given the study's finding that believers already tend to overestimate how much other people agree with them.

The digital environment can potentially supply exactly the evidence their false-consensus intuition is looking for.

A Small Online Community Can Feel Like Society

Imagine a conspiracy group containing 100,000 members.

That sounds enormous.

But in a country of hundreds of millions of people, it may represent a tiny minority.

Inside the group, however, nearly every visible post supports the belief.

The psychological experience becomes:

“Everyone is talking about this.”

What the person really means is:

“Everyone in the information environment I repeatedly enter is talking about this.”

Those are radically different statements.

The internet makes it easy to confuse community size with population prevalence.

The Cornell-led findings suggest conspiracy believers may already be unusually vulnerable to precisely that confusion.

Likes and Shares Can Intensify Perceived Consensus

Social-media engagement produces visible numbers.

10,000 likes.

50,000 shares.

Three million views.

Those metrics feel like evidence of widespread agreement.

But they rarely tell us what percentage of the population supports the underlying claim.

A video watched by five million people could still represent a belief rejected by 95% of a country's population.

Views are not endorsements.

Likes are not representative polling.

Comments are not random samples.

Algorithms do not construct scientifically representative crowds.

Yet human intuition can treat these numbers as social evidence.

For someone already predisposed to false consensus, the effect could be powerful.

People Sometimes Share Conspiracies They Do Not Even Believe

There is another complication.

Belief and sharing are not the same behavior.

Experiments involving more than 4,000 participants have found that people sometimes knowingly share conspiracy theories for social reasons, including obtaining engagement. Positive social feedback can encourage further sharing.

That means seeing a conspiracy repeatedly online does not necessarily tell you how many people truly believe it.

Some may believe it.

Some may be joking.

Some may be outraged.

Some may be debunking it.

Some may simply want attention.

The visible volume of content can therefore exaggerate perceived social consensus even further.

The 93% Statistic Helps Explain Why Fact-Checking Can Fail

Suppose someone believes almost everybody agrees with them.

Then you show them a fact-check.

They may not interpret it as:

“I must have misunderstood the evidence.”

They may interpret it as:

“Why is this institution trying so hard to deny what everyone already knows?”

The correction itself becomes suspicious.

This can create a powerful asymmetry.

The fact-checker thinks the disagreement concerns one factual claim.

The believer may think the disagreement concerns the credibility of the entire information system.

Correcting the isolated fact therefore addresses only one layer of the problem.

Facts Still Matter

None of this means fact-checking is useless.

A large 2025 meta-analysis involving data from 194,438 participants across 40 countries found that people generally can distinguish true news from false news and actually tend to err more toward skepticism than indiscriminate credulity.

People are not hopelessly incapable of judging evidence.

And misinformation interventions can improve accuracy.

A major Nature Human Behaviour review identified multiple useful approaches, including:

• Accuracy prompts
• Debunking
• Friction before sharing
• Inoculation or prebunking
• Lateral reading
• Media-literacy training
• Source-credibility indicators
• Fact-check labels

The lesson from the overconfidence research is not:

Stop correcting misinformation.

It is:

Correction alone may be insufficient when confidence remains badly calibrated.

Accuracy Prompts Can Help

One promising intervention is remarkably simple.

Before someone shares information, briefly direct their attention toward whether it is accurate.

A 2022 meta-analysis of 20 experiments involving 26,863 participants found that accuracy prompts improved the quality of information people said they would share, primarily by reducing sharing intentions for false headlines.

The effect is not magical.

It does not cure conspiratorial thinking.

But it demonstrates an important principle:

People sometimes share misinformation not because they consciously prefer falsehood, but because accuracy is not the feature controlling their attention at that moment.

A platform optimized around engagement can make that problem worse.

A design optimized around reflection can potentially reduce it.

Reflection Itself May Reduce Conspiracy Belief

The Overconfidently Conspiratorial paper cites related experimental work showing that encouraging reflection can decrease conspiracy belief.

That matters because the proposed mechanism is not simply ignorance.

It involves confidence and reasoning style.

A useful intervention might therefore ask questions such as:

What evidence would change your mind?

How confident are you, from 0 to 100?

What would an expert on the opposing side say?

How many people do you think actually hold this belief?

What source would you trust if it contradicted your current view?

These questions do something ordinary fact corrections often do not.

They force the person to examine the process by which confidence was formed.

AI Can Even Reduce Some Conspiracy Beliefs

Recent experiments suggest that personalized conversations can sometimes reduce conspiracy belief.

A 2026 Scientific Reports study found that participants who discussed health-related conspiracy beliefs with an AI-based system reduced their confidence in those beliefs relative to a control condition.

That finding is especially interesting because conspiracy theories are often assumed to be completely resistant to argument.

They are not always.

People can revise beliefs.

But effective correction may require engagement with the believer's actual reasons rather than dropping a generic fact-check into their feed.

Humiliation Is Probably a Bad Strategy

Imagine telling someone:

“You believe this because you're overconfident and bad at reasoning.”

Even if the statistical research supports a population-level association, using it as an insult toward an individual is unlikely to produce intellectual humility.

It will probably produce defensiveness.

The study should therefore not become ammunition for calling conspiracy believers stupid.

The researchers did not find that every conspiracy believer lacks intelligence.

They found that, on average, believers were more likely to misjudge their own performance.

Smart people can be overconfident.

Educated people can be overconfident.

Experts can be overconfident outside their domains.

Nobody is fully immune.

The useful question is not:

“Which group is stupid?”

It is:

“What practices help all of us keep confidence aligned with evidence?”

Intelligence Does Not Guarantee Good Calibration

A person can be highly intelligent and still possess poor epistemic habits.

Intelligence can even make motivated reasoning more sophisticated.

If someone strongly wants a conclusion to be true, greater reasoning ability may sometimes help them construct more elaborate arguments defending it.

This is why education alone is not a complete solution.

The 2026 study on education and conspiracy belief found that the normally protective effect of education disappeared among participants scoring relatively high on traits such as narcissism and need for uniqueness.

Knowledge matters.

But so does the willingness to revise what knowledge seems to imply.

The Scientific Mindset Is Not “Trust Experts Blindly”

Good epistemic calibration does not require unquestioning obedience.

Experts make mistakes.

Governments sometimes lie.

Corporations sometimes conceal wrongdoing.

Real conspiracies have occurred throughout history.

Scientific consensus changes when better evidence appears.

Healthy skepticism is essential.

The difference is what happens after evidence arrives.

Skepticism says:

“Show me the evidence.”

Conspiratorial certainty can become:

“No evidence could ever change my mind because contrary evidence is itself part of the conspiracy.”

Those positions may sound superficially similar.

They are epistemically opposite.

Real Conspiracies Do Exist

This point deserves emphasis.

The word “conspiracy” refers to people secretly coordinating toward a goal.

That unquestionably happens.

Watergate involved conspiracy.

Corporate price-fixing conspiracies occur.

Organized criminal conspiracies occur.

Intelligence agencies conduct secret operations.

Therefore, believing that powerful groups sometimes coordinate secretly is not irrational.

Researchers studying conspiracy theories typically focus on claims that are unsupported, highly speculative or resistant to available evidence.

The psychological problem is not believing that conspiracies are possible.

It is using conspiratorial explanation as a default framework even when the evidence does not justify the confidence being placed in it.

The Most Fringe Claims May Reveal the Mechanism Best

Why was overconfidence most strongly associated with fringe conspiracy theories?

Because popular beliefs offer many reasons for confidence.

If almost everyone around you agrees, confidence may partly reflect legitimate social evidence.

But when nearly everyone disagrees and a person remains extremely certain, individual confidence becomes far more important.

The believer must effectively conclude:

“I am better positioned to understand this than nearly everyone who rejects it.”

That is exactly where miscalibration becomes consequential.

False Consensus Can Create Political Misunderstanding

The implications extend beyond conspiracy theories.

If people systematically overestimate how many citizens share their views, they can misinterpret:

Elections.

Polling.

Public-health compliance.

Protests.

Institutional decisions.

Scientific consensus.

Media coverage.

A person expecting 70% public support for their position may interpret a 20% election result not as evidence that they misjudged public opinion, but as proof that the election was manipulated.

False consensus can therefore become a bridge between ordinary cognitive bias and larger conspiracy narratives.

Overconfidence Can Make Losing Feel Impossible

This dynamic is particularly dangerous when people form strong predictions.

“My candidate will obviously win.”

“Everyone I know supports them.”

“My feed is full of supporters.”

“The rallies are enormous.”

“The polls must be fake.”

If the candidate loses, someone with calibrated confidence can update:

“My information environment was not representative.”

Someone with extreme overconfidence may instead reason:

“The outcome cannot be real.”

The psychological step from unexpected outcome to hidden manipulation becomes much smaller.

The new study does not specifically prove this sequence in elections.

But its false-consensus findings reveal the cognitive architecture that could make such interpretations attractive.

Conspiracy Believers May Not Feel Like Outsiders

Popular culture frequently depicts the conspiracy theorist as someone proudly standing alone.

Everybody else is asleep.

Only I can see the truth.

The Cornell-led study paints a subtler picture.

Many believers may not experience themselves as lonely dissidents.

They may believe society is largely on their side.

That is a fundamentally different psychological situation.

They are not necessarily asking:

“Why does nobody believe me?”

They may be asking:

“Why do institutions refuse to acknowledge what most people already know?”

If that perception is wrong, the institutional silence itself can look conspiratorial.

This Creates a Feedback Loop

The mechanism could work like this:

1. A person encounters a fringe claim.
2. They are unusually confident in their ability to judge it.
3. They accept it.
4. They overestimate how many others agree.
5. They join communities where agreement is highly visible.
6. Their perceived consensus increases.
7. Institutions contradict the belief.
8. Institutional disagreement feels suspicious because “everyone” supposedly knows the truth.
9. Confidence increases further.

At that point, correcting the original claim may be difficult because the belief is being supported by an entire social and cognitive structure.

The Internet Did Not Invent Conspiracy Thinking

Conspiracy theories existed centuries before recommendation algorithms.

People blamed secret groups for:

Disease.

Political revolutions.

Economic crises.

Religious conflict.

Assassinations.

Wars.

Conspiracy thinking clearly does not require TikTok, Facebook or YouTube.

Human beings already possess the cognitive ingredients.

Pattern detection.

Agency detection.

Group suspicion.

Motivated reasoning.

Fear.

Social identity.

Overconfidence.

The internet did not create these tendencies.

It gave them new infrastructure.

Social Media Changed Speed, Scale and Visibility

Before modern social networks, a fringe theory might take weeks or months to circulate nationally.

Today a claim can move across continents in hours.

A believer can find thousands of like-minded strangers instantly.

A recommendation engine can repeatedly surface similar material.

Engagement statistics can provide artificial cues of popularity.

Influencers can monetize attention.

Communities can form around narratives almost immediately.

So asking whether conspiracy theories come from psychology or algorithms creates a false choice.

Human psychology supplies vulnerability.

Media systems influence exposure and reinforcement.

Both can matter simultaneously.

The Cornell Study Is Really About Epistemic Self-Awareness

The most important question raised by the research may not concern conspiracy theories at all.

It concerns the ability to say:

“I might be wrong.”

That sentence sounds simple.

Psychologically, it is difficult.

Our beliefs become connected to:

Identity.

Friends.

Political groups.

Moral values.

Status.

Self-respect.

Once that happens, changing a belief can feel like losing part of oneself.

Overconfidence protects against that discomfort.

But it also prevents correction.

Good Reasoning Requires Confidence to Move

Imagine confidence as a dial rather than a switch.

Weak evidence:

20% confidence.

Moderate evidence:

60%.

Strong convergent evidence:

90%.

Contradictory evidence appears:

Confidence falls.

New evidence appears:

Confidence rises again.

That is healthy calibration.

Conspiracy thinking often becomes problematic when the dial gets stuck near 100%.

No new evidence can move it.

At that point, the belief stops functioning like a hypothesis.

It becomes identity.

The Best Antidote May Be Better Questions

Fact-checking usually asks:

Is this claim true?

The new research suggests another layer of questions may matter:

How confident are you?

Why are you that confident?

How well do you usually perform on questions like this?

What evidence would lower your confidence?

How common do you believe your position is?

What data supports that estimate?

These questions target calibration rather than merely content.

That may be crucial because the content of conspiracy theories constantly changes.

The cognitive habit can remain.

Teaching Epistemic Humility May Scale Better Than Debunking Every Claim

A new conspiracy can appear tomorrow.

Fact-checkers must investigate it.

Evidence must be collected.

Articles must be written.

By then, millions may already have seen it.

Teaching general reasoning habits attacks the problem earlier.

Useful habits include:

• Estimating confidence rather than declaring certainty
• Comparing independent sources
• Seeking evidence against one's preferred conclusion
• Distinguishing expertise from confidence
• Checking base rates
• Recognizing nonrepresentative social samples
• Using lateral reading
• Asking what evidence could falsify a claim
• Waiting before sharing emotionally provocative material

These skills remain useful even when the particular conspiracy changes.

Platforms Can Also Design for Calibration

Social-media companies do not need to diagnose users psychologically.

They can alter information environments.

Possible interventions include:

• Prompts asking whether users have read an article before sharing
• Accuracy reminders
• Friction before reposting viral claims
• Context from independent sources
• Visible corrections
• Reduced algorithmic rewards for repeated misinformation
• Better exposure to credible opposing information

Research already suggests that accuracy prompts and related interventions can improve sharing discernment.

The goal is not to eliminate disagreement.

It is to make confidence more responsive to evidence.

The 12% and 93% Numbers Should Not Be Misread

These statistics are powerful enough that they will inevitably become simplified online.

So the exact meaning matters.

The study did not find that only 12% of Americans believe conspiracy theories.

It found that across the conspiracy claims examined in three of the studies, the belief position was actually held by a majority only 12% of the time.

Yet believers estimated themselves to be in the majority 93% of the time.

Those numbers describe the mismatch between actual and perceived consensus across the items studied.

They are not national prevalence estimates for conspiracy belief generally.

The Sample Was American

All 4,181 participants were U.S. adults.

That creates another limitation.

The relationship between overconfidence, social consensus and conspiracy belief may differ across cultures.

Countries vary in:

Institutional trust.

Media systems.

Political polarization.

Education.

Historical experience.

Religious traditions.

Internet use.

The study provides strong evidence in the American context.

Replication elsewhere would strengthen claims about universality.

Conspiracy Theories Are Not Explained by One Personality Type

There is also no single “conspiracy theorist personality.”

Different people can reach similar beliefs for different reasons.

One person may feel powerless.

Another may distrust government because of genuine historical abuse.

Another may want group belonging.

Another may be politically motivated.

Another may misunderstand statistics.

Another may be overconfident.

Another may simply encounter persuasive misinformation repeatedly.

Psychological averages describe tendencies.

They do not allow us to diagnose strangers.

That distinction is essential if this research is to improve public understanding rather than simply create another stereotype.

What the Study Changes

For years, much public discussion framed conspiracy theories primarily as an information-delivery problem.

Bad information enters an algorithm.

The algorithm amplifies it.

People believe it.

The new research highlights a missing layer.

Two people can encounter exactly the same false claim.

One thinks:

“That sounds interesting. I should check.”

The other thinks:

“I immediately know this is true.”

The difference exists before either person clicks the share button.

Platforms matter.

Information matters.

But so does the mind interpreting the information.

The Real Problem May Be Confidence Without Correction

Knowledge is never perfect.

Experts are wrong sometimes.

Institutions fail.

Science revises itself.

Ordinary people misunderstand things.

The ability to navigate that uncertainty depends on calibration.

Someone who knows relatively little but knows that they know relatively little can seek help.

Someone who knows relatively little but believes they understand everything has no reason to seek correction.

That may be the most important implication of the Pennycook, Binnendyk and Rand study.

The problem is not simply ignorance.

It is ignorance that feels like expertise.

Frequently Asked Questions About Overconfidence and Conspiracy Beliefs

What did the new conspiracy-theory study find?

Researchers found that conspiracy believers consistently overestimated their performance on numeracy and perception tasks and dramatically overestimated how much other people agreed with their conspiracy beliefs.

Who conducted the study?

The authors were Gordon Pennycook, Jabin Binnendyk and David G. Rand.

Pennycook and Binnendyk were affiliated with Cornell University, while Rand was affiliated with MIT.

When was the study published?

The paper was first published online on May 24, 2025.

It appears in Volume 52, Issue 8 of Personality and Social Psychology Bulletin, published in August 2026.

How many people participated?

The research included 4,181 U.S. adults across eight studies.

What does dispositional overconfidence mean?

It refers to a general tendency to evaluate one's own performance or abilities more positively than actual performance justifies.

Did conspiracy believers perform worse on cognitive tests?

The crucial finding was not merely performance.

Believers overestimated their performance even after actual performance was statistically taken into account.

What kinds of tests did participants take?

The studies included measures involving numerical reasoning and difficult visual-perception tasks.

Why use a perception task?

It helped test whether overconfidence extended beyond mathematics.

Participants could still overestimate how well they performed even on a task where genuine skill differences were limited.

What does the 12% figure mean?

Across conspiracy claims examined in three studies, the belief position was held by a true majority of participants only 12% of the time.

What does the 93% figure mean?

People who endorsed conspiracy beliefs thought their position was the majority position 93% of the time.

Does that mean conspiracy believers thought 93% of Americans agreed with them?

No.

The 93% statistic concerns how often believers thought they were on the majority side, not the exact percentage of people they believed supported every conspiracy.

How badly did believers overestimate agreement?

The study reports that believers—particularly overconfident believers—overestimated agreement by more than fourfold.

Did political partisanship explain the false-consensus effect?

Not completely.

The pattern persisted even when participants estimated agreement among people from the opposing political side.

Did narcissism explain the effect?

No.

The association between overconfidence and conspiracy belief remained after the researchers controlled statistically for narcissism.

Did need for uniqueness explain it?

No.

The relationship also remained after controlling for the need to feel unique.

Does that mean narcissism and uniqueness do not matter?

No.

Previous research shows both can be associated with conspiracy beliefs.

The new study suggests overconfidence contributes additional explanatory value.

Was overconfidence stronger for fringe conspiracy theories?

Yes.

The relationship was strongest for the least widely supported conspiracy claims.

Is this the Dunning-Kruger effect?

It is related conceptually because both involve miscalibration between perceived and actual ability.

But the study specifically examines dispositional overconfidence and should not simply be relabeled as a Dunning-Kruger study.

Does the study prove overconfidence causes conspiracy theories?

No.

It provides robust evidence of an association.

More work is required to establish the exact causal direction.

Did the study prove social media algorithms do not cause conspiracy theories?

No.

The research did not experimentally test whether removing social-media algorithms eliminates conspiracy belief.

It examined psychological characteristics associated with conspiracy endorsement.

Can social media still amplify conspiracy theories?

Yes.

Research indicates that social networks, echo chambers, recommendation systems and patterns of exposure can influence how conspiracy-related content spreads.

Did conspiracy theories exist before social media?

Yes.

Conspiratorial explanations are centuries older than modern digital platforms.

Social media changed their potential speed, scale and accessibility rather than inventing the psychological phenomenon.

Why do people believe conspiracy theories?

Research suggests multiple contributing factors, including:

• Need for certainty
• Need for control
• Social identity
• Narcissism
• Need for uniqueness
• Intuitive thinking
• Distrust
• Exposure to conspiratorial communities
• Overconfidence

Are conspiracy believers necessarily unintelligent?

No.

The study concerns confidence calibration, not a claim that all believers are unintelligent.

A person can be intelligent and still overestimate the reliability of particular judgments.

Can highly educated people believe conspiracies?

Yes.

Education can reduce susceptibility on average, but it does not make people immune.

Some personality and motivational factors can weaken education's protective relationship.

Are all conspiracy theories false?

No.

Actual conspiracies occur.

Scientific research on conspiracy belief usually concerns unsupported, weakly evidenced or highly speculative conspiratorial explanations.

Why is false consensus important?

If someone believes most people share their opinion, institutional disagreement can appear suspicious.

This may reinforce the belief that authorities or media organizations are ignoring an obvious truth.

Can echo chambers worsen false consensus?

Potentially.

Highly homogeneous online communities can expose users repeatedly to people who agree with them, making a minority belief appear far more widespread than it really is.

Do likes and views prove that an opinion is popular?

No.

Social-media engagement is not a representative public-opinion sample.

A post can receive millions of views while still representing a small minority position.

Do people ever share conspiracies they know are false?

Yes.

Experimental research has found that some people knowingly share conspiracy material to gain social engagement, and positive feedback can encourage further sharing.

Why might fact-checking fail?

A factual correction addresses the claim.

It may not address deeper beliefs about institutions, personal expertise or perceived social consensus.

Highly confident believers can sometimes reinterpret contradictory evidence as part of the conspiracy itself.

Does fact-checking ever work?

Yes.

Research supports several misinformation interventions, including debunking, fact-check labels, accuracy prompts, media-literacy approaches and source verification.

What are accuracy prompts?

They are small interventions encouraging people to consider whether information is accurate before sharing it.

A meta-analysis of 20 experiments found that such prompts improved sharing discernment and reduced intentions to share false headlines.

Can AI reduce conspiracy belief?

Some recent experimental research suggests personalized AI conversations can reduce confidence in certain conspiracy beliefs.

What is epistemic humility?

Epistemic humility is recognizing that your own knowledge can be incomplete or wrong and adjusting confidence according to evidence.

It does not require believing experts blindly.

How can someone become less overconfident?

Useful habits include estimating confidence numerically, actively searching for disconfirming evidence, comparing independent sources and asking what evidence would change one's mind.

What is the biggest takeaway from the study?

The biggest finding is not simply that conspiracy believers hold unusual ideas.

It is that many appear to misjudge both themselves and the social world around them.

They overestimate their own cognitive performance.

They underestimate the possibility that they are mistaken.

And they can dramatically overestimate how many other people share their views.

That changes how we should think about conspiracy theories.

The problem is not merely that bad information exists online.

Bad information has always existed.

Nor can everything be blamed on recommendation algorithms.

Algorithms can amplify, rank and reinforce ideas, but they still interact with human beings who bring their own cognitive habits to the screen.

The deeper vulnerability may begin before the feed loads.

It begins when confidence outruns evidence.

When:

“I think this is true”

quietly becomes:

“I know this is true.”

And then:

“Everyone else knows it too.”

That final step may be one of the most consequential.

Because once a person believes that almost everybody secretly agrees with them, disagreement stops looking like disagreement.

It starts looking like suppression.

Experts are no longer mistaken.

They are hiding something.

Journalists are not skeptical.

They are covering it up.

Polls are not surprising.

They are manipulated.

Evidence is not contradictory.

It is planted.

At that point, correcting one factual claim may barely touch the structure holding the belief together.

The Pennycook, Binnendyk and Rand study therefore offers a more useful lesson than simply mocking conspiracy theorists as overconfident.

It reminds us that one of the most important intellectual abilities is not knowing the answer.

It is knowing how certain we deserve to be that our answer is right.

And that is a skill everyone—not only conspiracy believers—can get wrong.