Strength in Differences: How Diversity Can Improve Collective Judgment

A group becomes more diverse when its members differ in background, knowledge, experience, identity or ways of approaching a problem.

That difference can be valuable because no individual sees every relevant part of a complex situation. An engineer, nurse, economist and community organiser examining the same public-health problem may notice different constraints. A product team containing people with different technical specialties, ages, languages or life experiences may recognise needs that another group overlooks. A scientific collaboration spanning disciplines may combine ideas that would rarely meet inside one field.

But none of this means diversity automatically produces better decisions.

Difference can expand the information available to a group. It can also increase misunderstanding, coordination costs, subgroup formation and conflict. Research on work-group diversity has therefore produced mixed results rather than a universal performance law. A major Annual Review of Psychology assessment concluded that diversity can influence group processes and performance both positively and negatively and argued that the important scientific question is not simply whether diversity “works,” but which processes and conditions determine its effects.

That distinction is the key to understanding the strength in differences.

Diversity is not performance.

It is a source of variation.

Whether that variation becomes broader knowledge, better scrutiny and more creative problem solving depends on what the group does with it.

Different people can expand what a group knows and how it searches for solutions

The most straightforward advantage of diversity is informational.

People who have worked in different functions, industries, communities or disciplines often possess different knowledge. A finance specialist notices cash-flow risk. An operations manager sees a bottleneck. A frontline employee understands how customers actually behave. A lawyer notices regulatory exposure. A designer recognises that the proposed interface will be difficult for certain users.

No single person needs to represent an entire demographic group for this mechanism to work.

The advantage comes from adding information that would otherwise be absent.

Difference can also exist in problem-solving approaches, not merely in stored knowledge. Two equally competent people may search for solutions differently. One decomposes a problem analytically. Another looks for analogous cases. Another experiments rapidly. Another questions the assumptions defining the problem in the first place.

This kind of functional or cognitive diversity has an important theoretical foundation. In a 2004 mathematical model, Lu Hong and Scott Page showed that under specified conditions, a collection of problem solvers using sufficiently different approaches can outperform a group composed only of individually higher-performing but more similar problem solvers. The logic is that similar experts may become trapped in similar parts of the solution space, while different heuristics give a group additional routes toward a better answer.

That finding is often simplified into:

“Diverse groups are smarter than groups of experts.”

That is too broad.

Hong and Page developed a formal model with particular assumptions about the problem, the population and the diversity of problem-solving approaches. It is not evidence that any demographically diverse collection of people will outperform every homogeneous expert team.

The useful insight is narrower:

when a problem benefits from several genuinely different and competent ways of searching for solutions, similarity among individually strong problem solvers can become a limitation.

Task type therefore matters enormously.

If ten people are performing a routine procedure with one clearly established method, additional perspectives may contribute little. Coordination costs may dominate. But if the task involves uncertainty, incomplete information, innovation, forecasting or competing interpretations, a wider range of relevant knowledge can become much more valuable.

Research on team innovation makes essentially this distinction. An Annual Review of Organizational Psychology and Organizational Behavior review identified two major requirements for innovation: teams need access to useful informational resources, and they need a team climate that allows those resources to be exchanged and integrated. Diversity can help with the first requirement, but it does not automatically deliver the second.

Demographic characteristics complicate the picture further.

Age, gender, nationality, ethnicity, disability or socioeconomic background may be associated with different experiences, opportunities and constraints. Those experiences can matter enormously when organisations design products, policies and services for populations that are themselves diverse.

But demographic identity is an imperfect proxy for viewpoint.

Two people from different ethnic groups may have attended the same universities, trained in the same profession, worked at the same companies and developed very similar approaches to decisions.

Two people who appear demographically similar may have radically different expertise, values and life histories.

A diverse organisation should therefore avoid the assumption that one employee can “represent” everyone who shares a visible characteristic.

Representation can broaden whose experiences are present.

It cannot reduce people to those categories.

Difference can improve scrutiny—but it can also create friction

Diversity can affect group thinking even when members do not bring completely different technical information.

Sometimes the presence of social difference changes how carefully people think.

A particularly interesting example comes from experimental financial markets. Researchers randomly assigned participants in North America and Southeast Asia to ethnically homogeneous or diverse trading markets. Prices in the diverse experimental markets tracked fundamental values substantially more accurately; across the experiments, pricing accuracy was reported as 58% higher in the diverse markets. The researchers argued that participants in homogeneous markets appeared more willing to trust and imitate one another's judgments, while social difference created friction that encouraged greater scrutiny.

That result should not be converted into a universal claim that ethnic diversity improves every financial decision.

The experiments involved particular artificial markets with known underlying asset values. Real organisations and markets are more complicated.

But the mechanism is useful.

Comfort can sometimes reduce scrutiny.

When everybody in a room assumes that everybody else sees the problem in roughly the same way, questionable assumptions may pass without challenge.

Difference can interrupt that process.

The person who asks:

“Why are we assuming the customer can read English?”

or:

“Why do we think every household owns a car?”

or:

“What happens if this process is used by someone with limited mobility?”

may reveal that what looked like a universal assumption was really the experience of one particular group.

Dissent can perform a similar function.

Groups naturally create pressures toward conformity. Once a preferred solution begins to emerge, members may stop searching for alternatives or avoid raising inconvenient evidence.

Research on dissent treats disagreement more carefully than popular “devil's advocate” advice. Dissent can be functional because it disrupts uniformity and forces groups to reconsider assumptions, but it can also be punished, interpreted as disloyalty or become dysfunctional depending on how the group responds.

This is why the best groups are not necessarily the groups with the least disagreement.

They are groups capable of distinguishing conflict over ideas from hostility toward people.

A disagreement such as:

“I don't think the evidence supports this forecast.”

can improve a decision.

A conflict that becomes:

“You never understand the business.”

creates a different problem.

Research on diversity and team performance has long tried to explain this tension. The categorisation–elaboration model proposed by van Knippenberg, De Dreu and Homan argues that diversity can create informational benefits when groups deeply process different task-relevant perspectives, while social categorisation and intergroup bias can interfere with that process. In other words, the same differences that provide valuable information can also create “us versus them” dynamics that prevent the information from being used.

That makes friction potentially productive but never automatically so.

A diverse group may initially take longer to develop shared language.

Members may disagree more frequently.

People may need to explain assumptions that another group would take for granted.

Those costs can be worth paying if the task benefits from deeper scrutiny.

But if disagreement becomes personal, status differences silence contributors or members stop trusting one another, the informational advantage can disappear.

The goal is therefore not to eliminate friction.

It is to create conditions under which friction produces better inquiry rather than damaged relationships.

Representation creates possibility; inclusion determines whether the difference reaches the decision

A team can look diverse in a photograph and still make decisions exactly as it did before.

That happens when people are present but do not have meaningful access to information, important assignments, informal networks or decision-making influence.

The distinction between representation and participation is therefore crucial.

Imagine a meeting containing ten people from varied backgrounds.

Three senior members speak for most of the hour.

Junior employees rarely interrupt.

One specialist repeatedly raises technical concerns but is ignored until a senior colleague repeats them.

Several people know that the proposed schedule is unrealistic but believe challenging the project leader will damage their careers.

The group is diverse in composition.

Its decision process is not.

Psychological safety helps explain why.

Amy Edmondson originally defined team psychological safety as a shared belief that the team is safe for interpersonal risk-taking. In her classic study of 51 work teams, psychological safety was associated with learning behaviour such as discussing errors, seeking feedback and experimenting rather than simply protecting one's image.

The concept is sometimes misunderstood as meaning that everyone should feel comfortable all the time.

That is not the useful interpretation.

A psychologically safe team can contain difficult disagreement.

The important condition is that members do not expect humiliation or punishment merely for raising a relevant concern, admitting uncertainty or challenging an assumption in good faith.

This is particularly important for diversity because unusual information often sits with people who have less formal power.

A junior engineer may notice the safety problem.

A nurse may recognise that a procedure will fail in practice.

A disabled employee may identify an accessibility barrier.

A regional salesperson may understand why a national campaign will not work locally.

A new employee may see a strange process that long-serving staff have stopped questioning.

If rank determines which information counts, the organisation can hire all of those people and still waste most of the potential advantage.

Leaders therefore influence whether diversity becomes informational.

They decide whose contributions receive attention, whether disagreement is interpreted as useful or troublesome and how rapidly a team closes discussion.

Inclusive leadership does not require accepting every idea.

Many suggestions will be wrong.

The standard should remain evidence and relevance.

What matters is that the quality of an idea has a genuine opportunity to matter before the status of the person presenting it determines the outcome.

Practical systems can help because informal group behaviour often favours the quickest, most confident or highest-status speakers.

Written input before a meeting can give people time to formulate views.

Anonymous idea generation can reduce some status effects during early stages.

Structured turn-taking can prevent the same two people from dominating every discussion.

Asking the person closest to the operational problem to speak before senior leadership gives them a chance to provide information without merely reacting to the executive position.

Shared vocabulary matters too.

Interdisciplinary teams often appear to disagree when they are actually using the same words differently. A marketer's definition of “customer,” a software engineer's definition of “user” and a finance team's definition of “account” may overlap without being identical.

Spending five minutes agreeing on terms can prevent an hour of false conflict.

Teams also need decision rules.

Diversity should not mean that every disagreement continues until everybody becomes exhausted.

Groups need to know who has final authority, what evidence matters, when consultation ends and how unresolved disagreement will be recorded.

Clear decision rights make it possible to gather broad input without turning participation into paralysis.

The process can therefore be summarised as:

variation → voice → scrutiny → integration → decision

Recruiting diverse people supplies only the first stage.

The strongest argument for inclusion is not that every diverse team earns more money

Diversity discussions often become trapped inside the “business case.”

The argument goes like this:

Companies should become more diverse because diverse companies make more money, innovate more and outperform competitors.

That claim is attractive because it appears to align fairness with profit.

The evidence is much less universal.

The OECD's review of diversity and economic outcomes concluded that the firm-level business case is not especially clear-cut. Effects vary across sectors and outcomes. Positive relationships appear more often in knowledge-intensive, high-skilled and innovation-driven environments, but effects can be small, insignificant or different in other contexts.

That finding should not be treated as an argument against diversity.

It reveals a problem with the premise.

Fair access to employment, education, authority and participation should not depend on proving that the participation of historically excluded groups increases quarterly profits.

An organisation should not have to demonstrate that hiring a qualified woman increases revenue before concluding that excluding women from opportunity would be wrong.

A society does not need an innovation study to justify equal civic rights.

People have claims to fair treatment and opportunity independently of whether their presence improves a performance metric.

The OECD makes a related distinction. While it describes firm-level performance evidence as mixed, it identifies a strong economic and social argument against discrimination and exclusion because failing to use the skills and potential of large population groups carries substantial costs.

This gives diversity two different justifications that should not be confused.

The first is instrumental:

difference can improve information, creativity, error detection, market understanding and decision quality under the right conditions.

The second is normative:

fairness, equal opportunity and legitimate participation matter even when no measurable performance bonus can be demonstrated.

Keeping these arguments separate actually strengthens both.

It allows organisations to evaluate performance evidence honestly rather than exaggerating every correlation.

It also prevents human equality from being treated as a corporate productivity programme.

Representation matters for another reason: legitimacy.

Institutions make decisions affecting people who may have very different experiences of those institutions.

A transport authority designs systems for elderly passengers, disabled passengers, parents with young children, tourists and daily commuters.

A hospital serves people from different languages, religions and income groups.

A government creates policy for citizens whose lives may be quite unlike those of the officials drafting it.

No committee member perfectly represents a category.

But institutions in which authority is drawn from an extremely narrow population are more likely to miss information and can also face legitimate questions about whose experiences are considered normal.

This is especially important where exclusion has been historical rather than accidental.

Participation can therefore matter both because it improves what an institution knows and because people affected by decisions have a reasonable claim to be represented within the systems exercising authority over them.

Diversity becomes strength when institutions convert difference into disciplined inquiry

The most productive way to manage diversity is not to assume that different people will spontaneously produce better outcomes simply by occupying the same room.

The work begins after recruitment.

First, the diversity has to be relevant enough to the problem to matter.

Not every difference improves every task.

If a team is diagnosing a complicated manufacturing failure, diversity of engineering specialties, production experience, maintenance knowledge and operational perspective may be especially important.

If it is designing public services, language, disability, regional and user experience may become highly relevant alongside technical expertise.

The question should not be:

“How many kinds of difference can we collect?”

It should be:

“Which relevant knowledge, experience and problem-solving approaches are currently missing?”

Second, the team needs a common purpose.

Difference without a shared objective can become parallel identity rather than cooperation.

A strong group understands what everyone is jointly responsible for achieving, even when members disagree strongly about how to achieve it.

Shared purpose does not erase differences.

It creates a reason to work through them.

Third, groups need norms that permit good conflict.

Members should be able to challenge assumptions without treating disagreement as personal disloyalty.

Leaders can model this directly by rewarding useful correction.

When someone identifies a flaw, the first reaction should not be:

“Why are you being negative?”

It might be:

“Show us the evidence. If you're right, we need to know now.”

Fourth, information has to be integrated.

Listening to six viewpoints and then making the decision exactly as though none had been heard is not inclusion.

Teams need explicit ways to compare evidence, identify contradictions and determine whether new information actually changes the recommendation.

This is the distinction between consultation and influence.

A feedback loop helps.

After major decisions, organisations can ask:

Which perspectives changed the decision?

Which warnings proved correct?

Who was repeatedly ignored?

Were certain teams represented but rarely assigned meaningful responsibility?

Do promotion, speaking time and high-value assignments remain concentrated even though headcount looks diverse?

Measurement should therefore move beyond simple representation statistics.

Headcount tells you who entered the organisation.

It does not tell you who participates in important work.

It does not tell you who is promoted.

It does not tell you whose ideas survive contact with hierarchy.

Time must also be taken seriously.

A newly assembled interdisciplinary or culturally diverse team may initially communicate less efficiently than a long-established homogeneous group because members do not yet share terminology, assumptions or trust.

Expecting instant harmony can lead organisations to interpret ordinary coordination costs as evidence that diversity “failed.”

Teams often need routines before different knowledge becomes easy to combine.

This is why the strongest diversity research increasingly points toward contingency rather than slogans.

Van Knippenberg's review of team innovation emphasises the need to combine diverse informational resources with the climate and processes needed to exchange and integrate them.

The logic can be stated simply:

Diversity expands possible inputs.

Inclusion determines whether those inputs are heard.

Psychological safety affects whether people risk contributing them.

Shared goals determine whether disagreement remains cooperative.

Decision processes determine whether the information changes anything.

That is how difference becomes collective intelligence.

The strength of a diverse group is therefore not that every member thinks differently about everything.

Nor is it that demographic difference automatically creates superior judgment.

The advantage is that complex problems usually contain more relevant information than one person, profession or social experience can possess.

A homogeneous group may still perform brilliantly.

A diverse group may perform badly.

But when a task benefits from multiple forms of relevant knowledge and when institutions genuinely allow that knowledge to enter the decision, difference creates possibilities that narrower groups may not possess.

That is the most defensible meaning of strength in differences.

Not:

“Diversity always makes groups better.”

But:

“No one perspective contains the whole problem, and good institutions are capable of converting relevant differences into better questions, stronger scrutiny and broader judgment.”

Diversity supplies the variation.

Structure determines whether it becomes learning.

Trust determines whether people speak.

Fairness determines whether they get the opportunity.

And disciplined decision-making determines whether difference becomes an asset rather than unresolved friction.

That is a more demanding idea than celebrating diversity as a slogan.

It is also a more useful one.