There have been many studies of single fake news incidents, but this report takes a different approach.
The team crunched data on 126,000 stories (or rumour cascades) that appeared between 2006 and 2017 and were tweeted by 3 million people over 4.5 million times.
Definitions
The team specifically rejects the term fake news which has not only been rendered "meaningless" by politicians but which also implies wilful distortion of the truth. Instead, the researchers use the more objectively verifable terms true or false news.
For the purposes of their study, the researchers define news as "any story or claim with an assertion in it" and rumour as "the social phenomena of a news story or claim spreading or diffusing through the Twitter network".
In other words, rumours are inherently social and involve claims being shared between people. Rumour cascades are defined as the propagation of a single assertion via an unbroken chain of retweeting. Rumours can have many separate cascades. Each time someone tweets (rather than retweets) an assertion, that is counted as a cascade.
True or false?
Next the researchers set out how they classify true or false news. They used six independent fact checking organisations (factcheck.org; hoax-slayer.com; politifact.com; snopes.com; truthorfiction.com; urbanlegends.about.com). They then identified rumour cascades which the fact-checkers demonstrated 95-98% (whether false, true or ‘mixed’) agreement on.
The cascades were then measured for depth (the number of retweet jumps from the original post); size (the number of users involved in the cascade over time); maximum breadth (the maximum number of users involved in the cascade at any depth) and structural virality (is the cascade sourced through a single large broadcast or by many people, each responsible for just a small part of the cascade?).
Key findings
- News the fact-checkers identified as false was 70% more likely to be retweeted than true news
- False political news spreads three times faster than other kinds of fake news – it travels deeper, more broadly, and is more viral
- It took the truth six times as long as falsehood to reach 1500 people
- It took the truth 20 times as long as falsehood to reach a cascade depth of 10
- Users who spread false news had significantly fewer followers, followed fewer people, were less likely to be verified, were less active and were more recent to Twitter. "Falsehood diffused farther and faster than the truth despite these differences than because of them."
Blame the humans not the bots
The team found that bots don’t differentiate between false and true. The problems are more human-behavioural. Humans respond emotionally to content and this may affect their decision to retweet. Novelty is also a factor.
Takeaways for information professionals
- Stop using the term fake news. Let’s start using the term false.
- Fact-checking sites are becoming even more important. Get to know the individual strengths of each, and work to raise their profile with your users.
- Don’t blame it on bots alone – it’s humans who help create cascades of false news.
- Time to revisit the CRAAP test, and help your users to build their CRAAP detectors!
The study appears in Science; spotted on Futurism.