False news is more novel than true news, and that may be why we share the false much faster and more widely. Prominent responses to false news include surprise, fear and disgust. True news tends to be met with sadness, joy, anticipation and trust. Humans are more likely than automated processes to be responsible for the spread of fake news.
These insights emerge from a large and impressive study published on 9 March in the journal Science. Researchers at Massachusetts Institute of Technology, interested in how and why true and false news stories spread differently, used 126,000 stories that had been tweeted by 3 million people a total of 4.5m times. The data spanned 2006, when Twitter began, to 2017.
The study is unsettling reading, especially in light of what has so far emerged from US intelligence agencies, congressional inquiries and the special prosecutor Robert Mueller about use of social media to distort the 2016 presidential election. I hope the research helps to persuade more people that fake news powered by social media is a serious threat to all democracies’ health. A growing bundle of studies shows that this is a qualitatively and quantitatively new problem, not just a digital manifestation of the yellow press of old. Apart from effects on elections and referendums, fake news in social media can assist hate speech to turn into communal violence more quickly. And some government responses are troubling on free-speech grounds, such as Sri Lanka’s week-long ban on social media, or “digital curfew”.
The MIT researchers studied what they called “rumour cascades”. A cascade starts with a Twitter user making an assertion about a topic – with words, images or links – and continues in an unbroken chain of retweets. The researchers analysed cascades about news stories that six fact-checking organisations agreed were true or agreed were false. The study found that “falsehood diffused significantly farther, faster, deeper and more broadly than truth in all categories of information”. False political news reached more people faster and went deeper into their networks than any other category of false information.
The study compared the emotional content of replies to true and false rumours by using about 32,000 Twitter hashtags and a lexicon of about 140,000 English words that associate with eight basic emotions: anger, fear, anticipation, trust, surprise, sadness, joy and disgust. Were automated processes, or “bots”, the main culprits in spreading falsity? No – the researchers found, it’s humans.
Calling for more effort to identify the factors in human judgment that spread true and false news, including interviews with users, surveys, lab experiments and neuroimaging, the paper points to some obvious reasons to look deeper. “False news can drive misallocation of resources during terror attacks and natural disasters, the misalignment of business investments, and misinformed elections.”
Two features of this study, besides its published results, are heartening. Artificial intelligence was successfully deployed to good effect, for example, a bot-detection algorithm. And Twitter provided access to its data, some funding, and shared its expertise. The researchers have conditionally offered to share their dataset.
More openness by the social media giants and greater collaboration by them with suitably qualified partners in tackling the problem of fake news is essential. Traditional journalism organisations are potential partners too. They find, check and disseminate news, are well placed to assess veracity, attract masses of comment online and discussion on social media platforms, and have a clear incentive to maintain trust in their own contributions to democratic life.
• A Guardian podcast released on 16 March includes more detail about the MIT study, including an interview with one of its authors.
• Paul Chadwick is the Guardian’s readers’ editor
