A survey conducted during the 2019 presidential elections in Argentina and published in late February showed that fact-checking is effective in reducing the spread of false news, even if it does not change the opinion of most people.
The study was done by Ernesto Calvo, professor of government and politics at the University of Maryland, and Natalia Aruguete, at the National University of Quilmes (Argentina). The researchers had already been investigating the phenomenon of fake news when they were invited by the Argentine fact-checking organization Chequeado to independently conduct the study.
For the experimental part of the research, 2,040 people were interviewed in Argentina in April 2020, in a representative sample of the population. In the observational segment, the researchers assessed the impact, consumption and viralization of Chequeado’s checks on social networks during the electoral period, between June and December 2019.
One of the most important conclusions, which combines results from the interviews with the analysis of social networks, is that fact-checking has little capacity to impact people's opinions, but increases the cost of disseminating, on the internet, something that has already been categorized as false.
One of the cases analyzed with the dissemination of false information – that then re-election candidate Mauricio Macri was wearing a headset during a televised debate. According to the study, 720 retweets were published on the topic by people aligned with Frente de Todos, a group opposed to Macri, in the three hours prior to Chequeado’s verification.
In the three hours following the publication of the correction, the number of retweets linked to Frente de Todos fell to 74. At the same time, activity in the group hampered by the false news, aligned with Macri and political coalition Cambiemos, increased, from 37 retweets to 146 in the same time interval.
The experimental part, on the other hand, confirmed what other research had been pointing out, that people tend to believe in fact-checking corrections only when they reinforce their previous beliefs.
For this, the researchers showed the interviewees five publications and then divided the group in two. Both parties received an alleged check from Chequeado, but one said the information was true, while the other said it was false. In the end, the researchers asked whether people continued to believe in the original publication or not.
Using the example of Macri’s alleged headset, 66 percent of respondents who were aligned with the candidate believed in Chequeado’s "false" tag and responded that the original tweet was certainly false (46%) or probably false (20%). Among the interviewees who saw themselves as opposed to Macri, only 46 percent believed Chequeado’s verification, with 21 percent saying that the original content was "certainly false" and 25 percent that it was "probably false.”
When they received verification with the label "true," the ideological bias was maintained. Most of the interviewees who were against the candidate believed in the check, and the majority of those in favor of Macri are not convinced by the Chequeado check.
“What the study proves is that, even when they continue to think what they thought, people choose not to expose themselves to confront something that was published as false. Instead of publishing with the same speed that you would share something, because you like it, because it coincides with your thinking, if Chequeado posted that it is false, you tend to do it less. That is, the sharing of false content is lowered after a Cheqeuado intervention,” Laura Zommer, executive and journalistic director of Chequeado, told LatAm Journalism Review (LJR).
Zommer explained that the phenomenon of disinformation feeds on "bad actors” who create and disseminate false news in order to gain money or power, but also thousands of people who, ignorantly or inadvertently, share the content.
“What the study allegedly shows is that there is a way to proceed with these people, that is, there is a tool, which is fact-checking, which lowers the incentives for these people to share something false,” she said.
According to Calvo, one of the authors of the study, the result is positive, because, after Chequeado's intervention, “the circulation of the false message decreases dramatically in the community that was circulating it, although they do not believe in the correction.” And, for him, it is more important to prevent the spread of false news than to convince people.
“The key to managing fake news, since you cannot eliminate that people believe it, is to decrease its amplification. It is unrealistic to believe that people will read 200 tweets critically,” he told LJR.
Calvo reinforced that people do not necessarily communicate everything they believe. To share content, it is necessary that the informational and reputational cost is low. This is where fact-checking works, according to the survey.
“One is never going to be able to correct Bolsonarista or Trumpist actors who are creating false information and distributing it, instrumentally. So what one wonders is whether it can be prevented that people, who are not, in that sense, politically cynical, and do not share things that they do not believe, if these people can be less exposed to false news or less naive when it comes to sharing those things. For that, we do not need them to believe in the correction, we need that false news does not circulate."
The false tag causes "cognitive damage"
Another conclusion of the study is that the classification "false," attributed by a fact-checking organization, causes "cognitive damage" in the person who is corrected. According to Zommer, the way people interact on the Internet is political and emotional, linked to feelings of affinity and belonging, and not just governed by the factual. And this should be considered by the checkers, because it is in this context that their verifications will be read, she said.
"So every time a checker says to someone that something that is in his affinity is false, it is like a soccer referee that annuls a goal, you like him less, because he is saying something that you would like to be different. They say that every time a checker publishes a check with the false tag, he loses some of his symbolic capital, that is, there are people who like him less,” Zommer said.
Calvo said that this cognitive effect is measured at the end of the study. Each time Chequeado contradicts a person's conviction by means of a "false" tag, it places the organization on an ideological scale, increasingly distant from oneself.
“When a fact-checker is always saying false, false, false, his reputation goes down with each community that he is saying believes in something false. Because people think: it's full of false news, but they’re correcting mine all the time. They do not need to think that the correction is a lie, it is enough to think that the checkers are choosing the correction because they have a political bias,” Calvo said.
On the contrary, when the organization publishes a check stating that something the person agrees with is true, it is understood as a prize, a cognitive reward. For this reason, the researchers said that fact-checking organizations need to manage the number of "false" and "true" tags they use, considering the impacts on long-term institutional credibility. Especially since the "fake" tag tends to be the most frequent.
“This study does not say 'do not tag things as false,' what it does is highlight that checkers should not despise ‘true’ checks, because it helps us to recover brand value,” Zommer said.
Another advantage of the "true" tag is that it goes viral more often than the "false" tag. According to the survey, people tend to share a "true-true" check more, that is, they believed in the content and this was confirmed by the check. This case generates greater mobilization than a "false-false" check, that is, people did not believe in the content and this is confirmed by the verification.
“What we found regarding the double affirmation is that 'you are right, it was true' is validating original content that the person was already excited about. It is different from 'you are right, it was false.’ The interesting thing is that in both cases the person is right. The difference is that 'you are right, it was true' reaffirms a belief and excites you. While with 'you are right, it was false' the person is angry with the original Tweet and then does not get excited, that turns off the activation of the correction,” Calvo said.
Another benefit of the "true" tag is that it does not contribute to spreading an original false message, as is the case with the "false" tag, Calvo said. For this reason, he evaluates the result as very positive, because the "false" checks do not go viral so much, but they have the important effect of reducing the diffusion of the deceptive original content. At the same time, the 'true' ones tend to go viral more.
"The fact-checkers believe that it is a problem that the verification does not go viral like the false news, but for us that is a virtue. If the verification went viral it would amplify the message of the false news. The double true, which is the validation of the original news, produces virilization, but the correction of false news does not, that is good. Because if the original news is true, it is not a fake news issue. When one has false-false, which is the most common correction of fake news, all communities stop circulating the correction,” Calvo said.
For him, the concern that the verifications with "false" tags will go viral is that this contributes to keeping the public debate revolving around the themes raised by the fake news.
“When you see false news, many people think that what you have to do is make the truth go viral. But the question is whether one can put the [fake] news back in the middle of the forest, so that it is not heard.”
The professor argued that false news can cause huge damage, even if many people do not believe it. This is because they are able to guide, manipulate and control public discourse, what he calls agenda setting.
“If one believes agenda setting is important, and that people's opinion is not changed, but the subject they are talking about changes, then false news controls the agenda even if people do not believe it and although we correct it,” he said.
According to Zommer, the research signifies a change of mind within Chequeado. She said that, among journalists, there is an outdated idea that positive news does not matter to people and, in fact-checking, this is reflected in a tendency to devalue "true" checks.
“Before, if we had several true checks in the same week, it worried us a little and now it frees our heads, to say: ‘okay, we're doing our job well, we don't have to find false ones,” she said.
This does not mean that it is necessary to stop doing “false” checks, but that the "true" may be of greater use than previously thought, researchers said. For example: in the case of checking the speech of politicians, there is a clear benefit in the use of the "false" classification.
In addition to lowering the sharing rate of the original fake news, the "false" tag for authorities permits the “raising of the cost of lying and promotes accountability,” Zommer said.
On the other hand, in situations where the false news has no clear author, it may be interesting to use a "true" tag. That is, it may be more efficient to say "vaccines protect against COVID - true" than "vaccines do not protect against COVID - false.”
“In these cases, instead of doing a false, it will be more effective to do the double negation. There is a better chance that the verification will circulate,” the journalist said.
Although they were hired by the Chequeado, the researchers' work was done completely independently, Calvo said. The agreement was possible because the topic was already of interest to the researchers and, for them, it was important to have access to Chequeado’s internal data and its fact-checking methodology.
At the same time, it was agreed that the report with the results would be published, even if it was detrimental to Chequeado. Calvo said the organization "behaved very well" and there was no attempt at interference.
Zommer said the researchers were chosen, on purpose, for being "outside the movement" of checkers and skeptics of fact-checking. “They are two researchers who wrote a lot about disinformation and were not fans of fact-checking,” she said.
The journalist said, in the research prologue, that Chequeado took a risk when contracting the study. And that some members of the team were wondering why the organization was investing in it and what would happen if the result was negative. Zommer replied that if that happened, it was better to know and change work strategies.
Zommer stated in the text that the criticisms of the checkers, in all countries of the world, are always the same: “Is what they do any use (if the phenomenon of disinformation seems worse every day)? And are they really fair?”
She argued that it was, but acknowledged that there was a lack of research to show this. Zommer said that, as “it doesn’t seem appropriate to us to demand of others something different from what we demand of ourselves,” she decided to go after evidence about the impact of their work. And it worked. “Today we not only say that fact-checking works, we proved it.”