A Blog by Jonathan Low

 

Oct 11, 2018

Visualizing the Spread of Misinformation On Twitter

The key finding is that it takes a very few false stories to generate a cascade of them, providing that the originators understand the influencer networks into which they need to connect.  JL


Daniel Funke and Alexis Mantzarlis report in Poynter:

Only 4,888 tweets of 60 million linked to one of the stories debunked at the time less than 0.01 percent of the total. Fake news is mostly circulated by politically affiliated accounts whereas debunks mostly came from unaffiliated accounts. Fact checks of misinformation get four times fewer shares than the original falsehood. The result is heat maps showing the interactions between different political groups on Twitter. The findings do shed light on how misinformation spreads.

Visualizing the spread of political falsehoods

How much misinformation made the rounds on Twitter during the French presidential election last year? Possibly not a lot.
That’s according to a study conducted by the Politoscope project at the Institute of Complex Systems of Paris Île-de-France, which analyzed the interaction between Twitter accounts. The report looked at 60 million exchanges from more than 2.4 million users by collecting data related to French politicians and political keywords in real time using an automated visualization platform.
According to the methodology, which Le Monde’s Adrien Sénécat covered for Décodeurs, the more accounts retweeted each other, the more they were considered to be “close.” The result is a variety of heat maps showing the interactions between different political groups on Twitter.
So how does that relate to misinformation?
Using data compiled by Décodeurs’ Décodex project, which keeps track of misinformation the fact-checker identifies online, Politiscope researchers were able to map how fake news links circulated during the French election. What they found is that only 4,888 tweets of 60 million they analyzed linked to one of the stories that Décodeurs had debunked at the time — less than 0.01 percent of the total.
That finding echoes what other researchers have found about the proportion of misinformation to all posts on Twitter. That includes the observation that, according to Politiscope, fact checks of misinformation get about four times fewer shares than the original falsehood.
Décodex’s data is not exhaustive, as the project did not fact-check every piece of misinformation on Twitter. And, as Sénécat noted in his article, the same domains they identified have racked up far more shares on Facebook.
Emeric Henry, an associate professor of economics at Sciences Po, told Daniel that much of what the study covers has already been documented elsewhere. But the findings do shed light on how misinformation spreads.
“One of the main interesting findings is that fake news is mostly circulated by politically affiliated accounts (in particular those of the embattled candidate Fillon) whereas debunks mostly came from unaffiliated accounts,” he said in an email. “This seems like quite a new observation.”
In an email to Daniel, Sénécat agreed, and said that fact-checkers should think about adapting their methodologies to tackle a range of political misinformation.
“We fact-checkers should not only work on 100 percent false information but also delve into other phenomena such as propaganda,” he said. “And considering that false information is a portion of a more massive phenomenon, which we could define as misinformation as a whole.”

0 comments:

Post a Comment