Polarization Is Not A Disagreement

Two things get called polarization and they behave completely differently.

The first is disagreement about policy. What should tax rates be, how should borders work, what should schools teach.

The second is how people feel about each other. Whether you would be upset if your child married one of them, whether you would hire one, whether you think they are dishonest and dangerous rather than mistaken.

Almost everything people worry about is the second one, and the two have not moved together.

The Measurements

On most specific policy questions, populations in democratic countries are less far apart than the argument sounds. Many positions have converged over decades. On a fair number of issues, majorities across the divide want roughly similar things and disagree about degree.

Feelings have gone the other way and the change is large. When researchers ask people to rate their warmth toward their own side and the other side, warmth toward one's own side has been fairly stable for decades. Warmth toward the other side has dropped a long way. Willingness to have a family member marry across the line, which almost nobody objected to in the mid twentieth century, is now a common objection, though the older survey wording makes exact comparison shaky.

The gap between those two trends is the finding. People do not increasingly disagree. They increasingly cannot stand each other.

Disagreement is normal, permanent, and mostly harmless. Believing the other side is a danger to your existence is none of those things, and only the second one has ever produced violence.

Where It Comes From

The second kind does not grow naturally out of the first. It is produced, and largely by three ordinary processes that nobody had to intend.

Identities stacked up. Over several decades, in many countries, political affiliation lined up with religion, region, education, race, and the kind of place you live. It used to be common for your neighbors, your congregation, and your union to contain people who voted differently. As those overlaps thinned out, a political disagreement stopped being one disagreement among many and became a disagreement with everything at once. Losing an argument now means losing on every identity you hold, which is why it feels like a threat rather than a debate.

Attention rewards hostility. Material about the other side outperforms material about your own by a wide margin, and material expressing contempt outperforms material expressing agreement. Nobody sat down and designed this. It is what happens when engagement is the measure, and it means that the most visible member of the other side, in your experience of them, is systematically their worst one.

Somebody is always better off. Not a conspiracy. A reliable incentive. Fear produces turnout, donations, subscriptions, and cohesion, and it produces them faster than persuasion does. A great many people whose living depends on it discovered this without needing to coordinate.

The Part Almost Nobody Knows

Now the finding that changes how this looks.

People are badly wrong about what the other side believes, and wrong in a specific direction. Asked to estimate how many on the other side hold extreme positions, people overestimate substantially, often by a factor of two or more.

The people who are most wrong are not the disengaged. They are the most politically engaged and the heaviest news consumers, and in some measures the most educated. Which follows directly from Lesson 6.1, and should stop anybody from treating information as the remedy.

The same pattern holds for something more dangerous. People substantially overestimate how much the other side supports political violence. That misperception is not harmless, because support for violence is largely defensive. People endorse it when they believe the other side already has. Studies that correct the misperception, by simply showing people accurate figures about what their opponents actually endorse, reduce the reader's own support for violence.

That is one of the more hopeful findings in this course and it deserves the caution that goes with it. The effects are modest, they fade, and nobody has shown they change behavior at scale. It is a real result, not a solution.

Contempt, At Population Size

Lesson 1.3 said that contempt arrives dressed as clarity. It never feels like an emotion. It feels like having finally understood what these people are.

Now put that at the scale of a country. What contempt does to one relationship is that it stops you predicting the other person, because you no longer bother modeling anyone you have written off. At population scale that same failure means a large group of people become genuinely unreadable to you. Their behavior looks senseless, so the only available explanations are stupidity or malice, which confirms the contempt and closes the circuit.

That is the condition Lesson 3.2 described from the other end. Contempt at scale is the antechamber to moral exclusion, and it does not need anybody to hate anybody. It only needs enough people to stop finding the others worth understanding.

The Honest Limit

Do not take from this that every deep conflict is a misunderstanding. It is not.

Some disagreements are real, large, and about who gets what. Correcting a misperception does not resolve those, and a person who treats every serious dispute as a failure of communication will be useless in exactly the situations that matter most. Lesson 4.2 made this same point about security dilemmas and the caution transfers whole.

The claim here is narrower. The hostility is much larger than the disagreement that justifies it, and the excess is doing the damage.