Fake news is a threat to American democratic institutions, whether through online election interference or, in extreme cases, inciting violence. It also can destroy a person’s reputation and career. Social media companies are spending huge amounts of money to combat the problem but with little success.

New research published April 30 in Proceedings of the National Academy of Sciences offers a roadmap for dealing with false information. In “Believability of Evidence Matters for Correcting Social Impressions,” a team led by Cornell psychology professor Melissa Ferguson provides new evidence that people’s beliefs about the source of information affects how they take in that information, even at the level of their automatic responses.

They also found that new information can modify or even undo existing impressions created by fake news.

“We wanted to know whether offering information about the source of news matters for people’s gut-level, automatic reactions. Does knowing that something is fake have lingering pernicious effects that can later shape and influence our thoughts and behavior toward the person? Our studies suggest that establishing credibility for news sources is the right policy to combat fake news,” said Ferguson, one of the paper’s authors and senior associate dean for social sciences in the College of Arts and Sciences.

Ferguson and fellow researchers Jeremy Cone, Ph.D. ’12, of Williams College, and Kathryn Flaharty of Georgetown University conducted seven experiments with more than 3,100 participants to assess how the truth value of new information about others affected both their reported feelings and their gut-level, automatic reactions. The experiments ranged from those using video games and narratives of intergroup conflicts to studies featuring an individual named Kevin.

The researchers used the character of Kevin to assess how attitudes toward him changed depending on what information was provided. In one series of experiments, Kevin was first depicted entirely positively. Then participants were told something disturbing – in one version of the experiment, that he had been arrested for domestic abuse of his ex-wife. The researchers found that when the news about the arrest record was attributed to police records, participants’ rapid, gut-level attitudes toward Kevin immediately became more negative.

But when the same information was attributed to a friend of Kevin’s ex-girlfriend – more questionably reliable than the police – participants retained their positive attitude toward Kevin. They saw the ex-girlfriend as perhaps having a reason to spread malicious, untrue gossip and, by discounting the source, discounted the information.

“In other words, whether participants thought this new information was true determined even their automatic feelings,” the researchers wrote. “And, in a separate experiment, this occurred even if participants initially thought the information was true and only later discovered that it was from a questionable source.”

In another set of experiments using celebrities as the subject, the researchers found that even a single exposure to false news shifted participants’ automatic reactions to the negative from their previous positive attitude. However, the researchers found if they immediately informed participants that the news was fake, participants’ reactions did not switch to negative. The negative effects of the fake news could be undone simply by the researchers – who were seen as reliable sources of news – telling the participants that the news was false.

“The very same new information about an individual or event may or may not lead to substantive correction depending on the perceived credibility of the source – perceptions that are likely to differ markedly among people,” the researchers wrote. “Perhaps, then, the philosopher [Gloria] Origgi is correct when she suggests that we live not in an ‘age of information’ but rather in an ‘age of reputation’ in which the most impactful information is that which is shared by the most reputable among us – not just for deliberate evaluations, but also for relatively more unintentional ones.”