Who is to blame for spreading false rumours online? A new study suggests it's not just the bots. It's us.
False news spread "farther, faster, deeper, and more broadly" than true news on Twitter between 2006 and 2017, a team of US scientists has found.
Their study, published today in the journal Science, is one of the largest long-term investigation of fake news on social media ever conducted.
The role of social media in spreading misinformation, propelled by bots or malicious actors, has been heavily scrutinised since the election of US President Donald Trump in 2016.
In February, the US Justice Department charged 13 Russians with allegedly trying to "promote discord in the United States" by posing as Americans on social media.
"It just so happens our paper finishes around the same time as 'fake news' became the talk of the town," said Dr Soroush Vosoughi of the Massachusetts Institute of Technology.
The study's findings suggest veracity may be a rare commodity: the truth typically took about six times as long as a falsehood to reach 1,500 people.
Axel Bruns, a professor at Queensland University of Technology's Digital Media Research Centre, who was not part of the study, said the result helps proves the proverb:
"A lie can get half way around the world before the truth gets its boots on."
While the study was funded by Twitter (which also gave the team access to its full historical tweet archives), Dr Vosoughi said it was conducted independently.
What is true and what is false?
While recent concern about "fake news" has focused on political stories, Dr Vosoughi and colleagues also looked at urban legends, business, terrorism, science, entertainment and natural disasters.
They analysed 126,000 "rumour cascades" — tweets that contained links, comments or images about a story — spread by 3 million people more than 4.5 million times.
Political news was the largest rumour category with about 45,000 "cascades".
For example, rumour spikes occurred during geopolitical flashpoints such as the annexation of Crimea in 2014 and the 2012 and 2016 US presidential election.
The researchers selected stories that had been investigated by six fact checking sites including snopes.com and politifact.com.
But this method may have missed other, more subtle ways of spreading false news, said Professor Bruns.
"There is also a tendency for the people who spread this kind of misinformation or disinformation to … hide the fake news payload in a story that otherwise seems quite logical and fact-based," he said.
"There might a larger grey area of half-truths and half-false stories that are disseminated."
Bots are bad, but how bad?
Recently, social media companies have been made to face US and UK lawmakers about the role of bots on their platforms.
The researchers found bots accelerated the spread of news, but there was little difference in how false or true news spread when bots were removed from the analysis.
"We're not saying that bots did not have an effect, but bots cannot explain everything," Dr Vosoughi said.
He said he was unsure whether or not bots would be more prominent if the study had focused solely on political news.
"If we limited our study to only political rumours around the 2016 election, then my guess is — and this is just a guess — you would see bots playing a much greater role," Dr Vosoughi said.
Not all bots are malicious, for example, publishers use bots to automatically tweet news headlines.
Professor Bruns suggested strategically set up bots may also aim to make stories visible to Twitter's trending stories tab by mass retweeting, for instance, rather than distributing the false stories in the first place.
"The bots may initiate the visibility, but as it becomes visible, it's humans that play a big role in passing it on," he said.
Why do we spread false news?
The study provides only one part of the picture of how falsity spreads online.
For one thing, it only looked at English-language rumours, Dr Vosoughi said.
While the researchers suggested false news was shared more often because of its novelty, Dr Vosoughi said more work is needed about the motivations of people who share this content and its impact.
Is it actually changing people's minds?
So far, Twitter has mostly resisted being "the arbiters of truth," said Nick Pickles, Twitter's head of public policy for the United Kingdom.
"We are not going to remove content based on the fact this is untrue," he told British MPs in February.
"I don't think technology companies should be deciding during an election what is true and what is not true, which is what you're asking us to do."
Dr Vosoughi is now looking at interventions to try and stop the virality of false news.
For example, people who are likely to share information, but who are receptive to messages about whether an item is true or false, could act as "nodes of influence" to keep questionable news from spreading.
Bot panic or not, Dr Vosoughi said social media companies may need to intervene.
"In a way, the more engagement Twitter gets, the better it is for their business model.
"I wouldn't say that the platforms are complacent, but they have a big role to play in addressing this issue," he said.
Twitter declined to comment.