We have more ways to communicate than any generation in history. So why can it sometimes feel as though we understand one another less? An intriguing—and contested—new line of research suggests that the answer may lie not simply in what we believe, but in the kinds of connections we create.

There was a time when avoiding someone's opinions was relatively straightforward.

You didn't invite them to dinner.

Today, they may be in your family WhatsApp group, your neighborhood Facebook page, your LinkedIn feed and, somehow, arguing underneath a photograph of your dog.

We are connected to one another on a scale that would have been almost unimaginable a generation ago.

A friend who moved overseas once gradually disappeared from everyday life. Today, we can know what she ate for breakfast.

Former classmates remain visible decades after graduation. Colleagues follow one another home electronically. Strangers debate strangers on subjects ranging from politics to parenting.

We have gained an extraordinary ability to stay connected.

But has all this connection actually brought us closer?

A provocative study published in the Proceedings of the National Academy of Sciences suggests that the answer may be more complicated than we imagine.

Its central idea is surprisingly simple:

Under certain circumstances, increasing the number of social connections in a community may not reduce division.

It may increase it.

The great connection experiment

Researchers Stefan Thurner, Markus Hofer and Jan Korbel began with an observation.

Over roughly the same period that smartphones and social media transformed communication, political polarization also increased substantially in the United States.

The researchers assembled survey data that they interpreted as suggesting another change: the number of close friendships people reported had increased considerably.

Could these developments somehow be connected?

To investigate the question, the researchers didn't conduct an experiment on actual people. Instead, they created an agent-based computational model—essentially a simplified mathematical society populated by simulated individuals.

Those virtual people had opinions.

They had friends.

They had disagreements.

And, rather like real people, they preferred their social world to make some psychological sense.

Two familiar features of human behavior drove the model: homophily and social balance.

We like people who are like us

Homophily sounds complicated.

The basic idea isn't.

People tend to associate with people who resemble them.

That similarity might involve education, interests, culture, age, religion, lifestyle, politics—or simply a shared way of looking at the world.

Friendship can then reinforce the similarity.

If everyone around the dinner table broadly agrees about something, repeatedly discussing it may strengthen the group's confidence that its position is sensible.

And if everyone you respect thinks something is obvious, opposing views can gradually begin to seem not merely different, but rather peculiar.

This tendency to cluster with similar people has been observed in schools, workplaces and online communities.

Then there is the awkward triangle

The second ingredient in the model is something called social balance.

Imagine three women: Anna, Beth and Claire.

Anna likes Beth.

Beth likes Claire.

But Anna and Claire can't stand one another.

Anyone who has experienced friendship-group politics will recognize the difficulty.

Beth is now caught in an uncomfortable triangle.

Social-balance theory suggests that relationships tend to move toward configurations that reduce this kind of tension.

Perhaps Anna eventually warms to Claire.

But there is another stable possibility.

Anna and Beth remain friends—and both become hostile toward Claire.

Suddenly there is an “us” and a “her.”

Scale that elementary social dynamic across thousands or millions of interactions and something interesting can happen.

A tipping point

In the researchers' simulated society, relatively sparse networks could accommodate considerable diversity.

But as the average number of social connections increased, the model eventually reached a critical threshold.

Then something dramatic happened.

Instead of becoming increasingly integrated, the simulated society divided into two strongly opposed camps.

Physicists call this kind of abrupt change a phase transition.

Water provides the familiar analogy. Gradually lowering its temperature produces relatively little visible change—until a threshold is crossed and liquid becomes ice.

The researchers found something mathematically analogous in their simulated social network.

Above a certain level of connectivity, the system could abruptly reorganize itself into polarized groups.

And when the researchers introduced a small minority of rigid individuals with extreme views—described in the model as radical influencers—polarization began at lower levels of connectivity.

More exposure isn't necessarily the answer

This creates an uncomfortable challenge to one of our favorite assumptions about disagreement.

We often say that people simply need to hear the other side.

Sometimes they do.

But exposure is not the same thing as understanding.

A well-known field experiment published in PNAS in 2018 deliberately exposed frequent Twitter users to political messages from the opposing side.

The researchers expected that encountering different perspectives might moderate people's views.

That isn't what consistently happened.

Republican participants exposed to liberal messages subsequently expressed more conservative positions. Democrats moved slightly in the opposite direction after exposure to conservative messages, although that change was not statistically significant.

The researchers were careful not to claim that exposure to opposing views always increases polarization. Their experiment involved a particular population and a particular social-media environment.

But it demonstrated something important:

Simply showing people opinions they disagree with does not guarantee greater tolerance.

Sometimes people defend their existing identity more strongly instead.

But friendship can be different

There is another body of research that complicates the picture again.

Psychologists Thomas Pettigrew and Linda Tropp examined hundreds of studies involving contact between different social groups.

Their large meta-analysis found that intergroup contact generally reduced prejudice, rather than increasing it. Later research suggested that reduced anxiety and greater empathy and perspective-taking help explain why such contact can work.

Research specifically examining friendships across group boundaries has also found evidence that these relationships can be particularly valuable forms of intergroup contact.

So how can both things be true?

Perhaps because contact and relationship aren't the same thing.

Seeing an angry stranger's political comment is contact.

Knowing someone well enough to understand why she thinks differently is a relationship.

One gives us information.

The other gives us context.

A scientific disagreement worth knowing about

There is an important qualification to the new research.

The claim that people's number of close friendships has dramatically increased is disputed.

After the original PNAS paper appeared, sociologists Beate Völker, Rense Corten and Gerald Mollenhorst challenged its interpretation of the survey evidence.

They argued that different surveys asked about friendships in different ways, that comparable data do not consistently show the proposed increase, and that the evidence isn't strong enough to conclude that contemporary societies became more polarized because people acquired more close friends.

Their conclusion was considerably more cautious: friendship networks may indeed have changed, but we do not yet know confidently how much—or even always in which direction.

The original researchers subsequently defended their analysis.

This isn't a flaw in science.

It is science doing what science is supposed to do.

Someone proposes an explanation.

Others test its assumptions.

Evidence accumulates.

The story becomes more complicated.

Maybe we're counting the wrong connections

And here the story becomes particularly interesting.

Another group of researchers writing in PNAS in 2026 proposed that modern technology may indeed have expanded our social networks—but that much of the expansion could involve weak ties and acquaintances rather than emotionally close friendships.

That distinction matters.

Your closest friend and someone whose posts occasionally appear in your social-media feed are technically both social connections.

Psychologically, they are very different things.

The researchers argued that technological change has made long-distance and low-intensity interactions easier and more frequent. Their modelling and empirical work suggest that the growing importance of weak ties could provide another pathway through which increased connectivity contributes to polarization.

Perhaps, then, modern life has not simply given us “more friends.”

It has given us something historically unusual:

an enormous outer circle of people we know just enough to react to, but perhaps not well enough to understand.

The difference between knowing someone and knowing about them

Think about your own relationships.

A close friend can say something you strongly disagree with and remain your friend.

Why?

Because the opinion isn't all you know about her.

You know that she looked after her mother when she was ill.

You remember the ridiculous holiday you took together.

You know she worries about her daughter.

You know when she's joking.

You know when she's frightened.

You have accumulated thousands of pieces of information about one another.

A stranger online has none of that protection.

You encounter perhaps twenty words representing one opinion.

And those twenty words can become the person.

She is no longer Susan who happens to disagree with you.

She becomes one of those people.

That may be one of the profound differences between human relationships and digital connectivity.

The problem may not be disagreement

The original mathematical model contains another intriguing result.

One factor that could delay polarization was greater tolerance for disagreement.

In the language of the model, people become less sensitive to the social discomfort produced by encountering conflicting opinions.

That doesn't require changing anyone's beliefs.

It requires becoming more comfortable with the fact that another person can disagree with us without becoming an enemy.

That distinction seems increasingly important.

Tolerance does not mean agreement.

It does not mean pretending every argument is equally persuasive.

And it certainly doesn't require abandoning strongly held principles.

It means preserving the distinction between:

“I think you're wrong.”

and

“There is something wrong with you because you think that.”

Civil society depends rather heavily upon that small piece of intellectual real estate.

Perhaps we need better connections, not more of them

For years, technology companies understandably treated connectivity as an achievement.

More followers.

More friends.

More sharing.

More engagement.

Bigger networks.

But human relationships may not obey the same logic as telecommunications networks.

More isn't automatically better.

A network optimized for engagement may be very different from a community optimized for understanding.

And this is where the research offers something more useful than another argument about whether social media is good or bad.

The important question may be:

What kind of connection are we creating?

A friendship capable of surviving disagreement may be socially valuable in a way that another hundred followers are not.

A conversation between people who genuinely listen to one another may accomplish something that exposing millions of strangers to opposing opinions cannot.

The research on intergroup contact gives us reason for cautious optimism. Meaningful contact can reduce prejudice, particularly when it encourages empathy and reduces anxiety about people outside our familiar group.

The coffee table may still have something to teach the algorithm

There is something else that happens when people sit across from one another.

We see hesitation.

Humor.

Embarrassment.

Kindness.

Contradiction.

We discover that people are usually considerably more complicated than their opinions.

The woman who disagrees with us about politics may share our concerns about aging parents.

The colleague whose religious beliefs seem incomprehensible may be the person who quietly helps when everyone else has gone home.

The neighbor whose worldview seems completely different may make us laugh.

Real relationships continually interfere with our attempts to reduce people to categories.

Perhaps that interference is healthy.

Lydia's™ Take

We don't yet know whether greater social connectivity itself is causing societies to become more polarized.

The new mathematical research offers an intriguing mechanism, not a final verdict—and other researchers have raised legitimate questions about some of the empirical assumptions behind it.

But the larger question deserves our attention.

Technology has solved the problem of connecting people remarkably well.

It has been less successful at teaching us how to live with all those connections.

Perhaps the next challenge isn't connecting everyone to everyone else.

Perhaps it is rediscovering something considerably older:

how to know people well enough that disagreement doesn't automatically become dislike.

That is a smaller ambition than connecting the world.

It may also be the more important one.

Research & Trusted Sources

Thurner, S., Hofer, M., & Korbel, J. (2025). Why more social interactions lead to more polarization in societies. Proceedings of the National Academy of Sciences, 122(44). Read the original PNAS study

Petrova, K. (2026). A computational model explains the sudden rise in societal division. PsyPost. Read the PsyPost overview

Völker, B., Corten, R., & Mollenhorst, G. (2026). More social interactions, more polarization? The evidence is not there. Proceedings of the National Academy of Sciences. Read the PNAS critique

Thurner, S., Hofer, M., & Korbel, J. (2026). Reply to Völker et al.: On the evidence of increased close friendships in societies. Proceedings of the National Academy of Sciences. Read the authors' response

Górski, P. J., Vaccario, G., Ho?yst, J. A., & Macy, M. W. (2026). More connections, higher polarization: The role of weak ties and status-driven dynamics. Proceedings of the National Academy of Sciences. Read the PNAS paper on weak ties

Bail, C. A., et al. (2018). Exposure to opposing views on social media can increase political polarization. Proceedings of the National Academy of Sciences. Read the study

Pettigrew, T. F., & Tropp, L. R. (2006). A meta-analytic test of intergroup contact theory. Journal of Personality and Social Psychology. Read the research summary