Lille, France: So called 'Twitter bots' that masquerade as real people and spread fake news and propaganda on social media have been detected in action in France in the lead-up to this weekend's presidential election.
However, French voters are sharing much better quality information online than what many US voters shared in last year's presidential election, according to an analysis released on Saturday by a group of experts from Oxford University.
The Computational Propaganda project at the Oxford Internet Institute said they have not yet seen 'bot' activity in France at the levels they recorded during the last days of last year's American presidential election, when just as much 'junk news' was being shared online as genuine, professional news.
However they found that bot activity was rising in France.
In their new analysis - based on a sample they took in mid-March - the group found that for every two links to professionally produced news about the election shared on Twitter, there was one link to 'other' content such as ideologically extreme propaganda and deliberate false reporting.
In this 'other' category there was a fraction of content from Russian sources, the analysis found.
"It is not as extreme as in the US election," said Samantha Bradshaw, one of the authors of the report.
Ms Bradshaw said their research put the lie to claims in France that fake news was "running rampant" in the campaign.
"The data we have just doesn't reflect that narrative," she said. "It may be that the US was a special case."
However, it also could be the case that - as in the US - the level of junk news being pushed by automated accounts would massively increase in the final week of the campaign.
"In the US junk news had its highest point just before the election, and professional news was at its lowest," Ms Bradshaw said.
Cecile Vaissie, a professor of Russian studies at the University of Rennes 2, told the New York Times last week that the Kremlin, building on methods and contacts developed in the Soviet Union, had assembled a "formidable machine of influence" in France that was working to promote its preferred candidates - Republican Francois Fillon and the Front National's Marine Le Pen.
The CrossCheck project, a collaboration between Google and a network of professional news organisations, has been tracking fake news being spread on social media during the French campaign.
The false reports included:
- a claim that centrist Emmanuel Macron was being funded by Saudi Arabia (and another that he was supported by al-Qaeda);
- a false Figaro poll showing Marine Le Pen had 'won' the TV debate;
- a doctored photo that had far-left candidate Jean-Luc Melenchon wearing a Rolex;
- a claim that Macron wanted to impose a new tax on property owners;
- a claim that protesters and an 'anti-Front National' march in Paris had chanted "Jews, thieves, murderers";
- a satirical article saying that Marine Le Pen's father Jean-Marie was caught growing marijuana on his estate near Paris; and
- a claim that the Kremlin had Tweeted that "Moscow will help Le Pen to win the elections".
During the US election a lot of the fake news was injected into Twitter by 'bots' - automated accounts that masquerade as real people and try to create trends by flooding the feed with their own hashtags.
The number of 'high frequency' bots in action in France in mid-March was "pretty low" compared with the US election or the Brexit referendum, Ms Bradshaw said.
They found that about 11.4 per cent of the traffic about Socialist candidate Benoit Hamon was driven by automated accounts, compared with 8.9 per cent and 5.7 per cent respectively for poll favourites Emmanuel Macron and Marine Le Pen.
They also saw the level of automation was increasing over time.
During their March snapshot, the team found about 47 per cent of the links being shared in political conversations on Twitter in France were to professional news content, from both big and small outlets, that respected journalistic values such as fact checking and separating news from commentary.
A further 16 per cent were to professional political content, such as candidates' websites.
But 20 per cent fell in the 'other' category, of which a fifth was 'junk news' - propaganda, conspiracy theories and deliberately false reporting.
"It seeks to persuade readers about the moral virtues or failings of organisations, causes or people??? and uses attention-grabbing techniques, lots of pictures, moving images, excessive capitalisation, ad hominem attacks, emotionally charged words and pictures, unsafe generalisations and other logical fallacies," the report said.
The report concluded that "political conversations about French politics over social media are not as poisoned as the same kinds of conversations about the US election in 2016".
Last week, it was reported that Facebook had targeted 30,000 fake news profiles in France, as apart of a global effort to combat misinformation on the social network.
Facebook said it had deleted some accounts with large followings, and had set up automated systems to spot the tell-tale signs of fake profiles such as repeated postings of the same content.
