Homogeneous Self-serving Information Bubbles

3. Factors That Make Conflict Intractable

This introductory article was written by ChatGPT at the direction of Heidi Burgess, who reviewed, edited, and approved the final content.
For more information

Homogeneous self-serving information bubbles form when people spend most of their time with others who are much like themselves. Since it is more comfortable, most people tend to sort themselves geographically, socially, professionally, and culturally — they live in neighborhoods with people similar to themselves, work with people similar to themselves, socialize with similar friends, and follow similar news sources. As a result, they tend to share similar assumptions about what is true, fair, and good, versus what is false, unfair, and bad, even evil. Progressives tend to hear and believe progressive explanations of events; conservatives tend to hear and believe conservative explanations. Religious, racial, class, educational, and professional groups often develop their own information worlds as well.

Social media and personalized news feeds can make information bubbles worse, although Reuters Institute’s review of echo-chamber research cautions that the simple story that everyone is trapped by algorithms is too crude. Many people encounter some cross-cutting information on social media, but highly partisan citizens often actively choose like-minded communities and sources. In practice, offline sorting and online sorting reinforce one another.

Information bubbles become dangerous when they make one’s own side seem normal, reasonable, and well-informed, while making the other side seem ignorant, irrational, selfish, hateful, or corrupt. A 2025 Pew Research Center study found large partisan gaps in which news sources Americans use and trust. Republicans were much more likely to use and trust sources such as Fox News, Newsmax, The Daily Wire, and Breitbart, while Democrats were more likely to trust a broader set of mainstream and public-media sources such as CNN, NBC, ABC, MSNBC, NPR, PBS, AP, BBC, and The New York Times. When different groups rely on different sources, they often do not merely disagree about what should be done; they disagree about what is happening.

The result is a growing inability to understand the other side. More in Common’s “Perception Gap” project found that Americans often have distorted views of what people on the other side believe. Pew has similarly reported that eight-in-ten U.S. adults say Republican and Democratic voters cannot agree on basic facts, not just policies and plans. When people know little about the experiences, fears, and reasoning of those outside their own bubble, disagreement is easily interpreted as bad faith. People begin to assume that the other side must be stupid, immoral, brainwashed, or dangerous, rather than asking why reasonable people might see the situation differently.

These bubbles drive polarization because they reward certainty and punish curiosity. Within a homogeneous group, people gain approval by repeating the group’s favorite arguments and attacking its favorite enemies. They risk criticism if they ask hard questions, acknowledge the other side’s legitimate concerns, question someone on their own side, or point out inconvenient facts.

Research on social media suggests that simply forcing people to see opposing views is not enough; one PNAS study found that exposure to opposing views on Twitter did not reduce polarization and, in some cases, increased it. Other studies, however, showed changing social media algorithms and approaches can be helpful. For instance, the Civic Health Project's Prosocial Ranking Challenge is testing "alternative [social media] ranking algorithms to determine whether different content prioritization strategies can lead to healthier civic and individual outcomes." Although the project is ongoing, initial indications are that changing algorithms does help, although it certainly is not a complete solution. Breaking out of information bubbles requires more than putting opposing arguments into people’s social media feeds. It requires relationships, trust, humility, and credible messengers who can help people encounter unfamiliar views without feeling attacked.

The larger danger is that self-serving information bubbles make democratic problem solving almost impossible. Shared problems such as immigration, crime, inequality, climate change, education, public health, foreign policy, and democratic reform require citizens to debate from at least some shared understanding of reality. When each group lives in its own information world, every proposal looks like a trap, every institution looks captured, and every compromise looks like surrender. Constructive conflict does not require everyone to agree. It does require enough contact across social and informational boundaries that people can see one another as fellow citizens with understandable concerns, not as enemies living in an alien world.

This page was created by ChatGPT in response to this prompt. It was then reviewed, edited, supplemented and approved by Heidi Burgess. More information about how and why we are using AI in this way, and about the growing number of ways in which Beyond Intractability is using ChatGPT, Claude and other AI systems to generate content and build out the BI system, is available on our BI/AI Overview Page.