The most common explanation for political polarization is also the most comforting: that the other side has simply become more extreme. It suggests a clear problem with a clear location—out there, among them.
A large body of research, however, points to a stranger and more uncomfortable diagnosis. We are not primarily divided by what we believe. We are divided by what we believe about each other. And on that score, the evidence is unambiguous: we are consistently, systematically, and consequentially wrong.
The foundational finding is known as false polarization. In a nationally representative probability sample, political scientists Matthew Levendusky and Neil Malhotra found that Americans significantly misperceive the public to be more divided along partisan lines than it is in reality. This error is not symmetrical. People’s misperceptions of opposing partisans are larger than their misperceptions of their own party. We see the other side as more extreme, more unified, and more hostile than it actually is.
We see the other side as more extreme, more unified, and more hostile than it actually is.
This distortion has a self-fulfilling quality. In a separate study, Douglas Ahler provided experimental participants with the actual average policy-related predispositions of liberal and conservative Americans. Those who learned the truth later reported their own opinions as 8–13% more moderate, on average. The reverse also holds: when we assume a bitter divide, we adjust our own positions toward the poles to match the fiction we have mistaken for fact.
But the story is not simply that we exaggerate how extreme the average opponent is. That is where the evidence becomes genuinely counterintuitive. A 2024 study by Nicholas Dias and Yphtach Lelkes, using surveys and pre-registered experiments with nearly 9,500 participants, found little evidence that partisans overestimate how extreme the “average” Republican or Democrat is.
The real error is different. American partisans vastly underestimate the diversity of each party’s attitudes. We collapse a wide distribution of beliefs, intensities, and nuances into a single, caricatured point. The problem is not that we think the other side is uniformly extreme. It is that we think they are uniformly anything at all.
This misperception of homogeneity is not a neutral cognitive error. It has direct consequences for democratic function. If you believe that out-partisans are unified in their opposition to a policy proposal, compromise about that proposal may seem impossible. A unified enemy faction is also perceived as more effective, and therefore more threatening, than a fractured one. The perception of a monolith is what makes the other side feel like an existential danger rather than a negotiating partner.
The neural evidence in the evidence pack supports this framework. Zhang Hejing, Chen Yiming, and Ma Yina describe how the brain computes “subjective value” for ingroup and outgroup members. Crucially, this computation is not binary. It is a continuous weighting process, a “group reference point” that reflects an individual’s relative emphasis on ingroup versus outgroup interests. The brain does not deal in caricatures. It deals in gradients.
This is why interventions that correct perceptions can work. Jeffrey Lees and Mina Cikara argue that a specific psychological domain—meta-perception, or one’s judgment of how they are perceived by others—is ideal for corrective interventions. The reputational nature of meta-perception makes people sensitive to the truth, even when that truth pertains to the other side. We care what others think of us, and that concern can be leveraged to correct our distorted views of them.
The problem is not that we are locked in a war of incompatible values. It is that we have mistaken a diverse and largely moderate population for a nation of extremists. As john a. powell puts it, we create caricatures and tell stories of people from other groups that make it hard for us to see their full humanity.
This process is not inevitable. Research from a Stanford course on polarization found that Gen Z is more optimistic than any other age group about the possibility of bridging the political divide. That optimism may be warranted. The same research found that while greater social media engagement is linked to more polarized views, the effect is similar across all generations, and Gen Z did not appear to hold wildly distorted views about the other side on immigration. Social media hardens feelings more than it distorts facts.
The choice, then, is not between agreement and conflict. It is between seeing the other side as a monolith or as a distribution. The students at Dickinson College demonstrate what the latter looks like in practice: a deep friendship between the presidents of the College Republicans and Democrats created the conditions for regular, collaborative listening across contentious issues. They chose to lean in rather than make an assumption and walk away.
This is not a call for false consensus or for papering over real disagreements. There are legitimate divisions in politics over competing policies and values. But polarization often drives us apart unnecessarily. The first step toward reducing it is not to change what we believe, but to correct what we assume about what others believe. The evidence suggests we are far less divided than we think—and that the belief in a deep divide is itself a primary cause of the conflict we fear.