The five members of a neighborhood library committee meet to decide whether to close the reading room on Saturdays. Attendance has fallen, and the budget is tight.
The chair thinks the closure is premature, but expects everyone else to be more pragmatic. One member knows several families who rely on Saturday access, yet hesitates to make the meeting more difficult. Another is doubtful about the projected savings. The remaining two are uneasy with the plan but suspect the decision has already been made.
“Unless anyone has a serious objection, shall we proceed?” the chair asks. Nobody speaks. A few people nod, each reassured by the composure of the others. The vote is unanimous.
This is an invented committee, not a documented case. Its members might have approved a decision none of them would have recommended alone. How can a group do something that the people inside it would privately reject?
A unanimous vote tells us what people agreed to do. It may reveal very little about what they knew or wanted.
Social psychology
When Everyone Thinks Everyone Else Agrees
At the library, the chair reads silence as approval. Each member reads the others' restraint in the same way. Nobody notices that the apparent consensus rests partly on mistaken estimates of what others believe.
Social psychologists call one version pluralistic ignorance. People privately hold doubts yet mistakenly believe that others support the prevailing norm; by concealing their discomfort, they reinforce that belief.
In four studies published in 1993, Deborah Prentice and Dale Miller examined college students' attitudes toward campus drinking. Students widely believed they were more uncomfortable with prevailing practices than their peers were. In one study over a semester, male students shifted their attitudes toward what they mistakenly believed to be normal; female students did not show the same change.
That is evidence for a mechanism in a particular social environment, not proof that library committees usually harbor silent majorities. To establish the mechanism in our hypothetical committee, we would need independently collected private views, rather than inferring them from the eventual vote.
Prentice & Miller (1993), Pluralistic Ignorance and Alcohol Use on Campus · Princeton University
Information and reputation
The First Voice Changes the Meaning of the Second
Suppose the chair speaks first and endorses the closure. The next member assumes the chair has studied the finances more thoroughly and agrees. A third member sees two confident supporters. Her own uncertainty suddenly seems less worthy of attention.
An informational cascade develops when earlier choices persuade later decision-makers to discount private evidence. A reputational cascade can look similar from outside, but the motive is different: someone withholds disagreement to protect relationships, standing, or opportunity.
Cass Sunstein and Reid Hastie identify both mechanisms among the ways deliberation can go wrong. They also describe how discussion may amplify earlier errors, polarize views, and give shared information disproportionate influence.
Still, following another person's lead can be sensible when that person has genuine expertise. What matters is whether the apparent confirmation comes from independent judgments or from several people repeating one source's signal. Five agreeing voices may contain only one original inference.
Sunstein & Hastie (2008), Four Failures of Deliberating Groups · Harvard University
Experimental group research
The Fact That Never Reached the Table
Perhaps everyone has seen the report showing lower Saturday attendance. Only one member knows the entry counter failed for several weeks. Another knows a nearby branch may close. A third has heard about a grant that could fund weekend staffing.
Researchers call a related experimental problem a hidden profile: participants hold different pieces of information, and no individual can find the best answer until the group combines what each person knows.
A 2012 meta-analysis by Li Lu, Y. Connie Yuan, and Poppy Lauretta McLeod synthesized 65 studies involving 3,189 groups. The groups discussed common facts far more readily than unique ones. Under the compared experimental conditions, hidden-profile groups were reported as eight times less likely to identify the correct solution than groups whose members had full information.
The strength of this evidence comes from repeated laboratory studies with defined answers. Real-world budget choices differ: priorities conflict, facts can remain uncertain, and people may reasonably disagree about which cost matters most.
But the research makes one point hard to dismiss. A meeting can contain enough knowledge to change a decision and still end without that knowledge ever being assembled.
Disagreement Can Help Even When the Dissenter Is Wrong
An objection does not have to be correct to improve a discussion. It can make the group work harder at testing the answer it was about to accept.
In an experiment involving 135 three-person groups, Stefan Schulz-Hardt and colleagues distributed information about job candidates so that the best answer could be found only through pooling evidence. Groups whose members began with different preferences solved the problem more often than groups sharing the same initially suboptimal preference. Even dissenters who supported the wrong option improved the chance of reaching the right one.
The researchers linked the benefit mainly to more intensive discussion, with some reduction in discussion bias. This is experimental evidence for a causal effect in a designed task. It does not establish that every contrarian improves every meeting. Repeated bad-faith objections can consume time and obstruct legitimate decisions.
A member who asks, “How was the attendance estimate calculated?” may nevertheless contribute more than someone who agrees politely, even when the original estimate proves sound.
Schulz-Hardt et al. (2006), Dissent as a Facilitator for Decision Quality · PubMed
Silence Has a Price
It is easy to tell quiet members that they should have spoken. Sometimes they should. An institution cannot consider evidence that its participants refuse to disclose.
But the cost of speaking is unequal. The most junior employee may be judged difficult. A contractor may worry about renewal. A community representative may lack the technical language used by the people who designed the proposal. To them, asking an uncertain question can carry a very certain social cost.
A 2024 systematic review by Carla Capuano and Peggy Chekroun surveyed 48 studies of conformity and emphasized the importance of context and moderating factors. Conformity is a documented tendency, not an inevitable human response.
An organization that expects people to protect the decision from error cannot rely entirely on their courage. It has to make useful dissent less dangerous.
Institutional design
Sometimes the Agenda Makes the Choice
Suppose the committee is offered exactly two options: close Saturdays or cut weekday services. It cannot vote for a six-month trial, a shorter Saturday schedule, or postponement until the grant is reviewed.
Members might approve the closure because the permitted alternative seems worse. That vote reflects a real preference among the available options. It says less about the alternatives excluded before voting began.
Agenda design determines which facts count as timely, who must justify a delay, and whether disagreement can take the form of a workable alternative. This is a conceptual institutional analysis, not a claim that one voting rule always produces a specific mistake.
Institutions need closure as well as debate. Unlimited objections can paralyze services; absolute unanimity can reward obstruction. The challenge is giving relevant information a fair chance without removing the ability to decide.
What If Nobody Likes the Best Decision?
The strongest objection to our opening question is that an outcome can disappoint everyone and still be sensible. Five neighbors might want a library open every day and nevertheless support restricted hours after learning how the costs fall on other services.
In that case the collective decision accounts for shared constraints that individual wishes did not. Nobody's unhappiness proves manipulation.
Yet even a genuinely popular decision might be unfair to a minority or to people not represented in the room. Aggregating preferences alone does not settle questions of rights, affected outsiders, or the credibility of technical evidence.
The better standard asks whether a group considered relevant facts, constraints, competing interests, and reasonable alternatives—and whether it can correct the decision when those premises change.
Where the Fields Collide
Psychology studies how people misread the norms around them. Information theory and economics examine what one person can infer from others' visible choices. Group experiments investigate whether distributed evidence is actually pooled. Institutional design determines which kinds of disagreement the process permits.
These mechanisms can feed one another. A member stays quiet because she expects agreement. Someone else takes that silence as evidence. Unique information is withheld. The chair interprets the visible convergence as a reason to end discussion. By the time the vote occurs, everyone seems more certain than the underlying evidence justifies.
None of this requires a conspiracy or an incompetent member. Nor does it mean groups cannot outperform individuals. Independent evidence, varied expertise, and trusted disagreement can make collaboration extraordinarily useful.
The critical moment is when one person's behavior becomes another person's evidence. That is how private doubt can acquire the public appearance of certainty.
How a Meeting Could Be Different
Before discussion, members could record a preliminary view and the fact that would most change it. The chair could speak after hearing independent assessments rather than before them. The agenda could reserve time for information known to only one participant, and the final record could distinguish objections resolved from uncertainties still open.
Such safeguards follow logically from the mechanisms in the research, but they are not guaranteed cures. Anonymous input can hide responsibility; too many formal challenges can exhaust a meeting; a leader may sometimes need to disclose expertise early because delay is dangerous.
What they can do is preserve information that a unanimous show of hands might otherwise erase. Agreement should be the result of examining alternatives, rather than the condition imposed on a conversation at its start.
What We Know — and What We Don't
Evidence from the campus studies documents misperceived norms in a specific population. Hidden-profile experiments and their meta-analysis show a repeated problem with information pooling. An experiment manipulating prediscussion preferences shows that dissent can improve solutions in a particular task. Research on numerical estimates shows that social influence can increase agreement without increasing accuracy.
These studies concern different settings and outcomes. Their findings cannot be combined into a universal estimate of how frequently organizations make unwanted decisions.
We cannot infer private preferences from the minutes of a unanimous meeting. A correlation between silence and a bad outcome would not by itself demonstrate that conformity caused the outcome. Our library scene remains a thought experiment; applying research mechanisms to its imagined members is interpretation, not direct observation.
The practical uncertainty is therefore a question an institution should investigate: does its apparent agreement reflect independent reasoning and shared knowledge, or have the conditions of discussion made some evidence too costly to reveal?
Back to the Reading Room
After the vote, the chair thanks everyone for their agreement. Outside, two members admit that they had wanted more time. Another wonders whether the savings were verified. The committee's decision remains binding, and the families who use Saturdays will feel its effect.
The members cannot undo the consequence merely by saying they never truly supported the plan. Still, a new question emerges: what would they have decided if their doubts had been heard before anyone interpreted the silence of others?
They might have reached the same painful compromise. They might have found an alternative. What they missed was the chance to discover the difference.
A group can complete every formal step of a decision without completing the conversation its members needed. That is why we should examine how the agreement was produced before using unanimity as evidence that the choice was wise.
The Next Question
Now suppose a model could correctly predict the committee's vote from past meetings, the order of speakers, and each member's incentives. It could forecast the unanimous closure before the room even gathered.
Would that mean it understood the decision? Would it know which private doubt mattered, or which unheard fact would have changed the result?
A model may predict an outcome accurately while missing the mechanism behind it. Is prediction the same as understanding?
Sources & Further Reading
- Prentice & Miller (1993), Pluralistic Ignorance and Alcohol Use on Campus · Princeton University
- Sunstein & Hastie (2008), Four Failures of Deliberating Groups · Harvard University
- Lu, Yuan & McLeod (2012), Twenty-Five Years of Hidden Profiles in Group Decision Making: A Meta-Analysis · SAGE
- Schulz-Hardt et al. (2006), Dissent as a Facilitator for Decision Quality · PubMed
- Lorenz et al. (2011), How Social Influence Can Undermine the Wisdom of Crowd Effect · ETH Zürich
- Capuano & Chekroun (2024), A Systematic Review of Research on Conformity · International Review of Social Psychology
Beyond the Question is an interdisciplinary series by Arin Vale.
Read the editorial approachTHE NEXT QUESTION
Is Prediction the Same as Understanding?
Coming next in Beyond the Question.
Explore the series