When can you call a lie a lie? And, more to the point, when can you call a liar a liar? These are not abstract questions. From the day Donald Trump announced his candidacy in 2015 with a semi-extemporized, 46-minute speech that included numerous documentable falsehoods, journalists and news organizations throughout America and around the world have struggled to find the right language to describe his language.
Are words like “falsehood” and “untruth” adequate terms for a responsible journalist to use? Or should the word “lie”—which suggests intention, rather than accident or ignorance—be used instead?
It should. In new research published in the International Journal of Communication, we offer empirical proof that Trump’s intent is to deceive. He is, in other words, a verifiable liar.
There is no reasonable debate about Trump’s loose relationship to the truth. Claims like “we have no soldiers in Syria” and “Alabama will most likely be hit” by Hurricane Dorian were demonstrably false. And, instead of backtracking when getting called out, Trump generally doubles down—for instance, using a Sharpie to doctor a NOAA map of Dorian’s projected path in a hamfisted attempt to give his lie the veneer of truth.
There is also no question that Trump has made far more false statements while in office than any previous president. According to the Washington Post’s meticulous tally, he passed the 20,000 “false or misleading claims” mark in July, after 827 days in office. And the volume of untruths has starkly accelerated over the past four years; University of Oregon professor David Markowitz has shown that Trump’s “bursts of deception” have grown in size and frequency since he first took office. It’s also abundantly clear at this point that the toll of Trump’s lying can be measured in human lives; as journalist Bob Woodward recently revealed, Trump openly bragged to him about lying to the American public regarding the severity of the COVID-19 pandemic.
We shouldn’t be surprised that the administration which stands behind Orwellian claims like “truth isn’t truth” and “alternative facts” so frequently lies. Yet many leading news organizations still resist ascribing intent to his actions. As Glenn Kessler of The Washington Post explained last year, the paper is cautious about using the word “lie” because “you can’t get into someone’s head. Trump especially is very situational, so he may actually believe what he is saying, despite all evidence to the contrary.” Similarly, earlier this year, New York Times executive editor Dean Baquet argued that it’s not a journalist’s job to ascribe intent. In his words, “Let somebody else call it a lie.”
We disagree. According to our new research, there is empirical evidence of Trump’s intent to deceive, which can be demonstrated through computational linguistic analysis of the words he uses when making false claims on Twitter. In effect, we used his lies to build a lie detector test.
Essentially, it functions on the principle that people talk differently when they’re lying than they do when they’re telling the truth. Using this finding as our starting point, we ran a year and a half of Trump’s tweets (about four thousand of them) through a text analysis program built by psycholinguistics professor James W. Pennebaker, and found consistent differences in the language Trump used for independently verifiable true and false claims. We then created an algorithm to predict whether his tweets would be true or false based solely on their composition. So, for example, it predicted this tweet about an upcoming television interview would be true because of the prepositions and @ signs, which were common among Trump’s verifiable statements. And it predicted this tweet about a Washington Post article would be false because of the negative emotion words, commas, and parentheses, all hallmarks of Trump’s falsehoods on Twitter. Overall, it was able to distinguish between Trump’s true and false claims roughly 92 percent of the time.
Obviously, algorithms like this can’t read people’s minds or prove that somebody’s lying, any more than a polygraph test can. And there are several other limitations that we describe in our study. What algorithms can do, though, is recognize predictable changes in somebody’s speech patterns. And if those changes in language typically correspond to false information, you have compelling evidence that the speaker is intentionally telling you falsehoods.
And that’s what we found in Trump’s case. We know he makes false claims at a faster clip than any other president; we know he makes the same false claims repeatedly, some of them hundreds of times; and we know he often uses predictable linguistic “tells” when he’s making false statements. In other words, we don’t have to read his mind to prove that he’s a liar.
When claims are false and there’s empirical evidence of intent to deceive, it’s fair to call them “lies.” And calling lies “lies” is important because we have to recognize that somebody’s trying to mislead us before we can talk about why they’re doing it and what we’re going to do about it. The American public deserves to know that Trump is lying to us, at least as much as we need to hear the substance of the lies themselves.