The meeting is almost over.
Everyone agrees.
The presentation looks clean.
The senior person has already said the plan makes sense.
Nobody wants to be the one who slows the room down.
One person has doubts.
But the doubt feels small compared with the confidence around the table.
So they stay quiet.
The decision becomes unanimous.
Months later, after the failure, everyone asks the same question:
How did a room full of intelligent people miss something so obvious?
There is a word people reach for immediately.
Groupthink.
It is one of the most famous ideas in the psychology of decision making.
It is also one of the most overused.
Because groups absolutely can suppress dissent, copy each other, polarize and make terrible decisions.
But the classic scientific package called groupthink is much less secure than popular culture makes it sound.
The theory that explained famous disasters
Psychologist Irving Janis popularized groupthink in the early 1970s.
He examined major policy failures and argued that highly cohesive decision-making groups could become so motivated to preserve unity that they damaged their own reasoning.
The classic symptoms included things like:
Illusions of invulnerability.
Pressure on dissenters.
Self-censorship.
Stereotyping opponents.
Rationalizing warnings.
An illusion of unanimity.
The model was compelling because it felt instantly recognizable.
Anyone who has sat inside a bad meeting can see the picture.
The problem came later.
Researchers started asking whether the full causal theory worked as cleanly as the story.
The famous examples became less certain
Janis used historical cases such as the Bay of Pigs invasion to illustrate the model.
But later scholars gained access to declassified documents, oral histories and additional records.
A 1998 reanalysis by Stanford psychologist Roderick Kramer concluded that much of the new evidence did not support Janis's original characterization of the Bay of Pigs and Vietnam decision processes.
Kramer argued that political considerations, presidential framing and reputational concerns were more important than Janis had allowed.
That does not mean the decisions were good.
It means a famous explanation for why they failed may have been too tidy.
This is a useful warning.
Historical disasters are perfect environments for hindsight.
Once we know the outcome, it is easy to search backward for psychological symptoms that make the failure feel inevitable.
Is groupthink a myth?
Not exactly.
A major 1993 Psychological Bulletin review concluded that research did not provide convincing support for the groupthink phenomenon as Janis had specified it or for the idea that its proposed characteristics reliably produce bad outcomes.
A 1998 review was more sympathetic, arguing that the concept remained heuristically valuable and had stimulated useful research.
So the fair conclusion is not:
Groupthink is fake.
It is:
The famous Janis model should not be treated as a settled law of group failure.
Some ingredients around it are strongly supported.
Conformity is real.
Majority influence is real.
Norms can suppress dissent.
Groups can overweight shared information.
People can align with leaders.
Opinions can polarize.
The weak point is the temptation to wrap all of these processes into one simple diagnosis and then announce that every bad consensus was "groupthink."
Conformity is real even if the master theory is messy
A 2024 systematic review examined decades of conformity research and concluded that majority influence remains a robust research domain, while also emphasizing cultural, methodological and contextual variation.
That is important.
You do not need the full groupthink theory to explain why a person might stay quiet.
They may doubt themselves.
They may fear social punishment.
They may think everyone else has information they do not have.
They may believe the leader has already decided.
They may want the meeting to end.
They may care more about belonging than the issue.
They may simply update rationally because ten people appear to know something.
Conformity is not automatically irrational.
If everyone around you independently knows more than you, changing your mind can be smart.
The danger begins when the signals are not independent.
Ten people can look like ten sources when they are really one
Imagine ten people agree with a claim.
That feels like ten pieces of evidence.
But what if nine of them copied the first person?
Now you do not have ten independent judgments.
You have one judgment echoed nine times.
This is one reason social information can become deceptive.
People observe other people's conclusions without seeing how those conclusions were produced.
A chain forms.
The first few decisions influence the next.
Later people interpret the growing consensus as additional evidence.
The group can become confident while the independent evidence has barely changed.
This logic appears in research on informational cascades, conformity and collective decision making.
The consensus is real.
The informational diversity may not be.
Groups often discuss what everyone already knows
There is another trap.
Before a meeting, different people may hold different pieces of information.
Some facts are shared by everyone.
Other facts are unique to one member.
When discussion begins, shared information has a built-in advantage.
More people can mention it.
More people can confirm it.
It feels familiar.
The unique fact may appear only once.
If the group does not deliberately surface unshared information, the conversation can spend most of its time repeating what everybody already knew.
A meeting can therefore increase confidence without increasing the quality of the information pool.
This is one of the strangest failures of collective intelligence:
Putting knowledgeable people in the same room does not guarantee their knowledge will actually be combined.
The group norm can matter more than cohesion
Janis made cohesion central.
But later experiments complicated that story.
In two studies published in 2001, researchers manipulated group norms toward either consensus or critical thought in decision tasks.
Groups with norms encouraging critical thought made higher-quality decisions.
Consensus norms did not.
The effect was linked to how shared and unshared information was valued.
This suggests a more useful question than:
Is this group too cohesive?
Ask:
What kind of behavior does this group treat as loyalty?
If loyalty means agreeing, dissent becomes costly.
If loyalty means helping the group avoid error, dissent can become a service.
The same cohesive group can behave differently under different norms.
Dissent is uncomfortable because it changes more than the decision
A dissenter does not merely add information.
They change the social situation.
Now the group must confront uncertainty.
The leader may lose face.
Other members may have to reopen work they thought was finished.
People who already supported the plan may experience the dissent as criticism of their judgment.
This is why groups often punish internal deviants more strongly than outsiders.
The outsider was never expected to agree.
The insider violated the norm.
A major review of deviance and dissent argued that dissent can be both functional and dysfunctional, and that groups vary in when they encourage or punish it.
The important point is not that dissent is always correct.
It is that a group which cannot tolerate dissent loses a major error-correction mechanism.
Consensus can be excellent evidence
There is a fashionable mistake at the opposite extreme.
Once people learn about groupthink, they begin distrusting consensus itself.
Experts agree?
Groupthink.
Scientists agree?
Groupthink.
The team agrees?
Groupthink.
That is not critical thinking.
It is reverse conformity.
Consensus can be powerful evidence when people reached their conclusions through sufficiently independent processes, when expertise is relevant, when dissent had real opportunity to survive and when the result can be challenged by new evidence.
Aviation works because teams coordinate.
Medicine relies on standards.
Science builds consensus because repeated findings constrain what can plausibly be believed.
The presence of agreement does not tell you whether reasoning failed.
You need to know how the agreement was produced.
Why social media makes groupthink language irresistible
Online, people rarely see the full decision process.
They see the final opinion.
Then they see thousands of people repeating it.
This creates two opposite illusions.
One side sees consensus and thinks:
Everyone knows this is true.
The other side sees consensus and thinks:
Everyone has been brainwashed.
Both conclusions can be premature.
You need to know where the information came from.
How many independent sources exist?
What incentives shape public expression?
What happens to people who disagree?
Is uncertainty allowed?
Does the group update after failed predictions?
Without those questions, "groupthink" becomes an insult rather than an analysis.
A better warning system
Instead of diagnosing groupthink, look for concrete failure conditions.
Is the leader's preference announced before evidence is discussed?
Are dissenters punished socially or professionally?
Does everyone rely on the same information source?
Are opposing arguments described instead of represented by someone who actually holds them?
Are uncertainties removed from presentations because they look weak?
Does the group repeatedly discuss shared facts while unique evidence goes unheard?
Does status determine whose information counts?
Is the decision framed as proof of loyalty?
Can the plan be reversed after new evidence?
These are observable questions.
They are more useful than a psychological label.
The DarkBrain lens
A smart group is not automatically smarter than its members.
A cohesive group is not automatically irrational.
A dissenter is not automatically brave.
A consensus is not automatically corrupt.
The real variable is the decision architecture.
How independent were the judgments?
How expensive was disagreement?
How visible was uncertainty?
What information never entered the room?
And perhaps the most important question:
If the group's preferred answer were wrong, what mechanism inside the group would be capable of proving it?
If there is no good answer, the danger is not simply that everyone agrees.
The danger is that agreement has replaced the ability to detect error.