Skip to main content

People follow a crowd, no matter its politics

Amid the clamor of political polarization and mistrust, new Cornell Tech research has found cause for optimism: When it comes to evaluating news, people tend to trust the opinions of a large group whether it’s composed of liberals or conservatives.

The study of 1,000 participants found that Democrats were reliably influenced by Republican-majority crowds and vice versa, though the researchers also found that people are inclined to disregard news that contradict their own political views.

“In a practical way, we’re showing that people’s minds can be changed through social influence independent of politics,” said Maurice Jakesch, doctoral student in the field of information science at Cornell Tech and first author of “How Partisan Crowds Affect News Evaluation,” which was presented at the Conference for Truth and Trust Online, held online in October. “This opens doors to use social influence in a way that may de-polarize online spaces and bring people together.”

Political polarization has skyrocketed in recent years, exacerbated by the internet and social media, where people tend to be exposed to information conforming to their existing beliefs. With this study, the researchers sought to explore whether exposure to differing opinions could impact their preexisting views.

“When algorithms optimize for viewer engagement, they will often show content that people either like or are angered by,” Jakesch said. “That’s the reason we’re seeing a lot of extreme ratings online. But people’s evaluations would be less extreme if a broader, more representative audience had responded to the content evaluated.”

The researchers asked participants to rate 16 news claims, presented as headlines, as either true or false. Four of the headlines were consistent with Democrats’ views, four were Republican-consistent, and eight were collected from a list of headlines that were true but considered difficult to evaluate.

Participants were assigned to three groups: one in which participants could see how a group comprising mostly Democrats had rated the claims; one where the group of prior raters was mostly Republican; and a control group where participants did not see how others had rated the news.

For example, one headline read, “Trump’s First Mar-a-Lago Trip Cost Taxpayers $13.6 million.” Participants in one group were told, “75 Democrats and 21 Republicans answered so far,” and “24 say the claim is false and 72 say it is true,” and then asked to rate it as true or false.

Participants across political lines were 21% less likely to evaluate claims as true if they didn’t align with their views, the study found. But when it came to social influence, compared with the control group, both liberals and conservatives were highly influenced by a crowd’s opinion, regardless of its political makeup.

In almost all cases, the politics of the crowd didn’t have a significant effect on evaluations, except when a majority-Democrat crowd affirmed a Republican-consistent claim.

The findings offer opportunities for social media platforms to make design changes that decrease political polarization – or at least don’t exacerbate it.

“While platforms cannot show the same content to everyone, they could use the data they already collect about people  to estimate what feedback they would get from a more representative audience,” Jakesch said. “Statistically correcting sample selection bias doesn’t cost a lot, and based on our results, could move more people towards the political center. Even if I think a video is great and right, if I see that not everyone thinks so, that may influence my opinion.”

The paper was co-authored with Moran Koren of Stanford University and Cornell doctoral student Anna Evtushenko. The paper’s senior author is Mor Naaman, professor of information science at the Jacobs Technion-Cornell Institute at Cornell Tech.

 

Media Contact

Gillian Smith