From an article in FT:
Nearly a quarter of web content shared on Twitter by users in the battleground state of Michigan during the final days of last year’s US election campaign was so-called fake news, according to a University of Oxford study.
Researchers at the Oxford Internet Institute (OII) also determined that these users shared approximately as many fake news items as “professional news” over the same period. The report, published on Monday, concludes that links to fake news stories accounted for 23 per cent of the links tweeted by a sample of 140,000 Michigan-based users during the ten days up to November 11 last year. The proliferation of fake news, particularly via social media, has been blamed for distorting public perceptions and political debate in several western countries, while some political leaders have seized upon the term to selectively dismiss credible but critical news coverage. OII researchers prefer the term ‘junk news’, defining it as “various forms of propaganda and ideologically extreme, hyper-partisan or conspiratorial political news and information”. Professional news, defined as political news and information by outlets that display the qualities of professional journalism, constituted another 23 per cent of the links tweeted by the Michigan group, suggesting that this content was no more likely to be shared than fake news.
The researchers also categorised links to Russian-origin news stories and unverified WikiLeaks content. Taken together with fake news, content in these three categories was shared more widely overall than professional news. “I think it’s safe to say that’s a bad thing for public life and the political conversation in [Michigan],” said Professor Philip Howard, principal investigator with the Project on Computational Propaganda at OII. Polls of potential voters in Michigan taken immediately before the November 9 election indicated a lead of more than 3 percentage points for Democratic candidate Hillary Clinton over now-US president Donald Trump. Mr Trump ultimately won the state by 0.2 points, becoming the first Republican candidate to carry Michigan since 1988. Despite being the subject of much debate, particularly in the aftermath of the US election, there has been little quantitative research into the fake news phenomenon. “We believe this is the first real research on fake news over social media”, Prof Howard said. A report by economists at Stanford University in January suggested that social media are a relatively unimportant source of political news for many Americans, with television by far the most important source. Unlike the Oxford study, however, this research was based largely on online survey data collected after the election. A second OII report, also published on Monday, stated that the share rate of fake news was only 10 per cent among users tweeting about the recent German presidential election between February 11 and 13. Moreover, professional news was shared at almost double the rate it had been in the Michigan study — 45 per cent. Prof Howard explained that, although the German president has a largely ceremonial role and is elected by parliament, it was important to conduct the February study in order to establish a baseline share rate for fake news in advance of the country’s federal election in September. In addition to fake news sharing, the researchers also set out to study the levels of “bot” activity generated by fully or partially automated Twitter accounts. The levels of activity were lower than expected, Prof Howard said. “We think bots are much less likely to put in geolocation information, so by focusing on a subset of [users located in Michigan], we think we discarded a lot of bot activity,” he explained. The German study, while not geographically specific, nonetheless identified an “overall low level of bot-driven automation”.