A collaboration between two research teams with opposing perspectives found that, despite claims to the contrary, simply reminding people about the concept of accuracy improves the quality of information-sharing on both sides of the political aisle.

Both groups – one of which included Gordon Pennycook, associate professor of psychology in the College of Arts and Sciences – conducted exhaustive analyses of 21 surveys of more than 27,000 people on liberal versus conservative ideology, including whether they voted for Donald Trump in previous presidential elections.

They found that “nudges” regarding the importance of accuracy reduced the sharing of misinformation, regardless of political affiliation. But half of the 70 models they employed suggested that the prompts were slightly less effective on those identifying as Republican and having voted for Trump (relative to Democrats and Clinton/Biden voters).

“Still, the one thing we found across all 70 analyses was that the effect is present among Republicans,” said Pennycook, a co-author of “On the Efficacy of Accuracy Prompts Across Partisan Lines: An Adversarial Collaboration,” which published March 20 in Psychological Science.

The research was spurred by a 2022 letter written to the editors of Psychological Science, in response to earlier work by Pennycook, with the title “Meta-Analysis Reveals that Accuracy Nudges Have Little to No Effect for U.S. Conservatives.” The letter was sent by a group led by Sander van der Linden, professor of social psychology in society at the University of Cambridge and a co-author of the current work.

“Our primary concern was that their letter said ‘little to no effect’ among Republicans,” Pennycook said. “People may take that to mean that it doesn’t work for Republicans. But we found, in every single analysis, that it does.”

Co-senior authors are van der Linden and David G. Rand ’04, the Erwin H. Schell Professor and Professor of Management Science and Brain and Cognitive Sciences at Massachusetts Institute of Technology.

Pennycook said the idea for the “adversarial collaboration” – typically research involving two groups with a theoretical disagreement, where that theory is tested by both – was important given the other group’s claim that accuracy prompts have little impact on Republicans.

“Most of the research looking at misinformation-sharing in the U.S. is about Republicans – a lot of it is right-wing stuff,” Pennycook said. “And so if accuracy-prompting doesn’t work among Republicans, that really calls into question whether this intervention would actually be feasible to roll out at a large scale.”

Other prior work has shown that reminding people of the importance of accuracy increases the quality of their news-sharing decisions, so the two groups agreed to analyze the same datasets in a multitude of ways.

The research teams were: Rand, Pennycook and lead author Cameron Martel, a doctoral student at MIT; and van der Linden, psychology professor Jay Van Bavel of New York University and Steve Rathje, postdoctoral researcher in psychology at NYU.

Study participants were presented with a set of actual true and false current news headlines, taken from social media and presented in the format of a Facebook post, and asked whether they would share the information online. All false headlines were from fact-checking sites such as snopes.com and factcheck.org; all true headlines came from mainstream news sources.

In 17 of the 21 experiments, participants were asked how socially and economically conservative or liberal they were. In two experiments, participants were asked how ideologically liberal or conservative they were; in the remaining two, participants were asked for their views on income equality and government’s role in individuals’ everyday lives.

Across 70 different modeling types and five accuracy-prompt treatments, the groups found that prompts – even just subtle, unrelated reminders of the importance of the truth – did have an effect on information-sharing for both left- and right-leaning individuals. More specifically, they found that nudges decreased the sharing of falsehoods among Republicans by 3.3% to 14.6%. This was compared to a decrease in falsehood-sharing of 7.6% to 19.1% among Democrats.

Pennycook said the opposing group, based on the pre-analysis specifications they stated as being most central, mostly confirmed their original hypothesis: Accuracy prompts were less effective for Republicans than for Democrats. But overall, Pennycook said, the study showed that prompts improved the quality of information-sharing for both.

“We took the implication of their previous paper to be, essentially, that nudges don’t work for Republicans; on this, at least, we agree that there is, in fact, a clear effect,” Pennycook said. “However, it is pretty ambiguous whether the effect is genuinely smaller for Republicans than Democrats. On this, we did not really come to an agreement. In any case, if it is smaller, it’s not to such a degree that it would be easy to detect.”

This research was funded in part by a grant from the Alfred P. Sloan Foundation.