Political overconfidence worsens polarization in online debates

A new study published in Communication Research suggests that people who overestimate their political knowledge are more likely to react negatively during online conversations with those who hold opposing views—and as a result, become more emotionally polarized over time. The researchers found that this “overconfidence effect” helps explain why cross-cutting discussions, or exchanges between people with different political views, often fail to reduce hostility between political groups and may even backfire.

The study explores two psychological patterns that may be undermining efforts to reduce political polarization through dialogue. One is affective polarization—the tendency to dislike or distrust those from the opposing political side. The other is the Dunning-Kruger effect, a well-documented bias in which people with low ability or knowledge in a domain tend to overestimate their competence. The researchers proposed that individuals who overestimate their political knowledge are more likely to respond to disagreements with hostility, especially in online spaces where social norms are weaker. This oppositional behavior, in turn, deepens affective polarization.

To test this idea, the researchers conducted a two-wave online survey in South Korea during early 2022. They recruited a nationally representative sample through a professional survey firm, yielding 1,175 responses in the first wave and 948 in the second. Participants answered questions about their political beliefs, behaviors on social media, and attitudes toward people with opposing views. Crucially, the study measured both perceived and actual political knowledge, allowing the researchers to assess overconfidence—how much individuals overestimated their knowledge relative to their performance on a factual quiz.

Cross-cutting discussion was measured by asking participants how often they talked about public affairs with people who held different political opinions on their most-used social media platform. Oppositional responses to disagreement were assessed one month later by asking how often participants engaged in negative behaviors on social media, such as clicking “dislike,” posting critical content, or making hostile comments in response to opposing views. Affective polarization was also measured in the second wave using a composite score of how warmly or coldly participants felt toward candidates and supporters from opposing political parties.

The results showed several key patterns. First, the Dunning-Kruger effect was clearly present in the political domain. Participants who scored the lowest on objective political knowledge tended to believe they were more knowledgeable than average. In contrast, those who performed well were more likely to underestimate their knowledge. This misalignment between actual and perceived knowledge helped identify who was politically overconfident.

Free daily newsletter

Second, the study found that cross-cutting discussions were not directly associated with either an increase or decrease in affective polarization. However, the relationship changed when looking at behavioral responses. People who more frequently engaged in conversations with political opponents were also more likely to respond with oppositional behaviors, such as publicly criticizing the other side or reacting negatively to their posts. These behaviors, in turn, predicted higher levels of affective polarization over time. In other words, cross-cutting exposure alone did not reduce hostility—but when it led to antagonistic reactions, polarization worsened.

Most importantly, political overconfidence influenced this chain of effects. Among participants who overestimated their political knowledge, cross-cutting discussions were significantly more likely to produce oppositional reactions, which then predicted increased polarization. For those who underestimated their knowledge, this pathway was not statistically significant. In other words, overconfident individuals were more prone to interpreting disagreement as a threat or challenge, responding with hostility, and becoming more emotionally polarized as a result.

These findings help clarify why conversations across political divides on social media often do not lead to greater understanding. Rather than reflecting an honest effort to engage, such exchanges may become battlegrounds for self-affirmation, particularly among those who falsely believe they have superior political knowledge. When people feel certain they are right, they may become less willing to listen and more inclined to attack, reinforcing existing divisions.

The study has several implications for how we think about political dialogue and polarization in the digital age. First, it highlights the importance of metacognition—people’s ability to accurately assess their own knowledge—in shaping the tone and outcome of political conversations. Second, it suggests that interventions aimed at reducing polarization should not only focus on improving political knowledge, but also on addressing overconfidence. Helping individuals recognize the limits of their knowledge may reduce the likelihood of hostile responses to disagreement.

But there are limitations to consider. The study relied on self-reported data, which may be influenced by biases or inaccuracies in how participants recall their behavior. The researchers also focused exclusively on online interactions, particularly on social media platforms where anonymity and low accountability may encourage more extreme behavior. It remains unclear whether the same patterns would hold in face-to-face conversations.

Additionally, the study used only two waves of data, which limits the ability to draw conclusions about long-term effects or the direction of causality. For instance, it is possible that those who are already polarized may be more likely to behave negatively and misjudge their knowledge.

Future research could build on these findings by incorporating more waves of data, examining how overconfidence develops over time, and testing interventions designed to reduce it. Experimental studies could also help clarify how specific features of online platforms—such as anonymity, comment visibility, or feedback mechanisms—shape the relationship between overconfidence and hostility. Another promising direction would be to explore how individual traits like intellectual humility influence responses to disagreement and whether cultivating such traits can help counteract polarization.

The study, “How Political Overconfidence Fuels Affective Polarization in Cross-cutting Discussions,” was authored by Han Lin and Yonghwan Kim.