Bots aren’t to blame for fake news
False claims posted on Twitter spread faster and further than the truth, study finds.
Humans spread false claims faster and further than the truth, a new scientific analysis finds.
A new scientific analysis offers proof of something that social media acolytes have known for years: Twitter is an excellent platform for spreading actual news.
Unfortunately, the analysis shows, it’s even better at spreading fake news.
Compared with tweets about claims that were verifiably true, tweets about claims that were undeniably false were 70% more likely to be retweeted in the Twitterverse. And false claims about politics spread further than any other category of news in the analysis.
A team of data scientists and social media experts from MIT came to these conclusions after examining the spread of thousands of tweets shared by millions of people over 12 years. The findings were reported last week in the journal Science.
“Falsehood diffused significantly farther, faster, deeper, and more broadly than the truth in all categories of information,” wrote Soroush Vosoughi and Deb Roy of the MIT Media Lab and Sinan Aral of MIT’s Sloan School of Management.
“It took the truth about six times as long as falsehood to reach 1,500 people,” the trio added.
The researchers considered “news” to be “any asserted claim made on Twitter” expressed in words, a photo or a link to a full article.
Thanks to politicians, the term “fake news” now means information that does not support one’s point of view. The researchers made a point of avoiding this phrase.
Instead, they categorized news as either “true” or “false.” A tweet labeled “false” doesn’t imply that the writer is trying to pull a fast one. It only means that the claim is inaccurate.
When any type of news claim spreads on Twitter, it becomes a “rumor.”
The pattern by which a tweet spreads is a “rumor cascade.” If a tweet is retweeted 10 times in an unbroken chain, it is a cascade with a size of 10. If two people tweet the same news and each of those tweets is retweeted five times in an unbroken chain, we have two rumor cascades, each of size five.
Vosoughi, Roy and Aral used this framework to map the spread of information on Twitter since its creation in 2006 through last year.
They mapped out every rumor cascade rooted by a claim that had been factchecked by snopes.com, politifact.com, factcheck.org, truthorfiction.com, hoax-slayer.com or urbanlegends.about.com. They wound up with roughly 126,000 rumor cascades.
For each cascade, the researchers determined the size (the number of people involved in the cascade from start to finish), the depth (the number of retweets in a single unbroken chain), the maximum breadth (the largest number of people who were part of the cascade at any depth), and the structural virality (a measure of the number of people who helped a tweet spread). The more a rumor spreads, the more that all four factors increase.
Here’s a sampling of what the researchers found:
Rumor cascades based on true news rarely spread to more than 1,000 people. But at least 1% of rumor cascades based on false news did this routinely.
In the top 0.01% of both true and false rumor cascades, the false ones “diffused eight hops deeper into the Twittersphere than the truth.”
False news was more likely to be “viral.” So not only were the retweet chains longer, but they were more likely to branch off into new chains.
Rumor cascades about politics outnumbered those of all other topics, including urban legends, business, terrorism, science, entertainment and natural disasters. The news that ultimately spread the most concerned politics, urban legends and science.
False news about politics spread to 20,000 people almost three times more quickly than any other kind of false news reached 10,000 people.
Compared with people who spread true news, users who spread false news were newer to Twitter, had fewer followers, followed fewer people and were less active with the platform.
The researchers believe false news has more novelty, making it more surprising and more valuable — and more likely to be retweeted.
They figured this out by studying a random selection of about 25,000 tweets seen by 5,000 people and comparing their content with the other tweets those people would have seen in the previous 60 days. They also examined the emotional content of replies to these tweets and found that false tweets prompted greater feelings of surprise and disgust. (True tweets generated replies expressing sadness and trust.)
The researchers made a separate map that excluded fake Twitter accounts identified by a bot-detection algorithm. Removing rumor cascades that started with bots did not change the patterns that propelled false news.
“False news spreads farther, faster, deeper, and more broadly than the truth because humans, not robots, are more likely to spread it,” the trio wrote.
If all this has you feeling hopeless about the future, the researchers advise you to hang tight.
“Understanding how false news spreads is the first step toward containing it,” they wrote. “We hope our work inspires more largescale research into the causes and consequences of the spread of false news as well as its potential cures.”