It’s a real epistemological mess out there. The US president is tweeting wild lies about the election, and because Twitter refuses to ban him from the platform, it’s some poor employee’s job to write little alerts under his tweets saying that his claims are “disputed,” as though the truth of the matter is somehow unknowable. There has been a pervading sense within political discourse for years that this kind of rejection of settled fact is bad, but remarkably little rigorous research into the ways or extent it might be harmful. On the next two editions of Deep Dive, we’ll look at some of the work that has been done on that subject, and try to better understand the dreaded “disinformation threat.”
Though he might like for you to believe otherwise, Donald Trump did not invent lying. Nor, despite concerns about election disinformation originating in Russia, did Vladimir Putin. Disinformation has been a part of regular political communication for as long as there have been politics. With that in mind, one promising way to study modern disinformation is to treat it like other forms of speech by political leaders — rhetoric that may be taken more or less seriously by voters as the result of a number of factors.
That’s the approach taken by political scientists Katherine Clayton, Nicholas Davis, Brendan Nyhan, Ethan Porter, Timothy Ryan, and Thomas Wood in their new working paper. Their work examines the effect of Trump’s tweets about the 2020 presidential election on the attitudes of a cross-section of American voters. Clayton et al. aren’t particularly interested in the veracity of the president’s tweets. Instead, they treat the tweets as speech that matters because Trump, in his capacity as president and leader of the Republican party, sent them in an attempt to influence the political process.
Some of the tweets were fairly anodyne — recognizing National Doctors Day, for example — while others were, to use a word favored by many a mealy-mouthed newspaper editor, charged.
Their study relies on a panel survey experiment, which ran for 17 days in October, just before the 2020 election. Participants were asked for their baseline demographic data, political leanings, and views on the legitimacy of the election, and then shown a series of the president’s tweets. Some of the tweets were fairly anodyne — recognizing National Doctors Day, for example — while others were, to use a word favored by many a mealy-mouthed newspaper editor, charged. Of the more aggressive tweets, which the researchers categorized as “norm violations,” there were two types: general norm violations (“THE RIGGED AND CORRUPT MEDIA IS THE ENEMY OF THE PEOPLE!”) and election norm violations (“Rigged Election, and EVERYONE knows it!”). Some participants were shown just the boring tweets, while others were shown a mixture of boring and norm-violating tweets. After having seen the tweets, the respondents were again asked their views about the legitimacy of the election.
Perhaps unsurprisingly, seeing the tweets had no measurable effect on population-wide attitudes about the election. Trust in the election, willingness to accept the election results, support for political violence, support for democracy — all those measures stayed put, on average, no matter what kinds of tweets people were shown. Once Clayton et al. started subdividing the data, though, things got interesting. When respondents were grouped by their approval of the job Trump has done as president, measurable effects started to appear. When people who opposed Trump saw his tweets saying that the election was fraudulent, their trust in the legitimacy of the election actually went up. When people who supported Trump saw Trump’s norm-violating tweets — either about the election or just about Adam Schiff being corrupt and “probably sick” — their trust in the election went down.
The election tweets caused Trump supporters not just to doubt the legitimacy of the election but to believe the substance of Trump’s accusations. Nearly across the board, Trump supporters who saw his election tweets were more likely to say that the election was rigged in favor of his opponents.
The increased suspicion among Trump supporters did not lead to an increased taste for electoral violence. Indeed, the researchers were careful to note “we find that no evidence that support for political violence or belief in democracy changes after repeated exposure to” Trump’s id. Nevertheless, the divergent effects of the tweets suggest that elite rhetoric is capable of causing significant reductions in confidence in American democratic institutions. The researchers offered no speculation on whether Twitter’s little blue alerts under Trump’s tweets served to restore that confidence.