This morning’s Observer column:
In the 1930s, a maverick young journalist named Claud Cockburn resigned from the Times and, with £40 borrowed from an Oxford friend, bought a mimeograph machine (a low-cost duplicating machine that worked by forcing ink though a stencil on to paper). With it he set up the Week, a weekly newsletter available by subscription in which Cockburn printed news and gossip that came to him from his diverse group of contacts in both the British and German establishments.
From the beginning the Week printed stuff that the mainstream newspapers wouldn’t touch because of fears of running foul of the Official Secrets Act, the libel laws or the political establishment. Cockburn, having few assets and a rackety lifestyle, proceeded as if none of this applied to him. But people in the know – the third secretaries of foreign embassies, for example, or City bankers – quickly recognised the value of the Week (for the same reasons as they now read Private Eye). Nevertheless the circulation of Cockburn’s scandal sheet remained confined to this small elite circle – and its finances were correspondingly dodgy.
And then one day everything changed…
So called ‘sock puppets’ are multiple social media accounts which are actually controlled by a single person. They’re a pain and the pain is getting worse — as we discovered in 2016 — because they can have a distorting impact on online discourse.
But now comes some good news. New Scientist reports that researchers have developed some tools that can detect these pests with reasonable accuracy.
[Srijan] Kumar and his colleagues at the University of Maryland and Stanford University in California analysed commenter accounts on news websites including CNN, NPR, Breitbart and Fox News. They identified the sock puppets by finding accounts that posted from the same IP address in the same discussion at similar times. This approach isn’t always possible, so they wanted to develop a tool that automatically detects sock puppets based only on publicly accessible posting data.
They found that sock puppets contribute poorer quality content, writing shorter posts that are often downvoted or reported by other users. They post on more controversial topics, spend more time replying to other users and are more abusive. Worryingly, their posts are also more likely to be read and they are often central to their communities, generating a lot of activity.
Based on their findings, the researchers created a machine learning tool that can detect if two accounts belong to the same person 91 per cent of the time. Another tool can distinguish between a regular account and a sock puppet with 68 per cent accuracy. The research will be presented this week at the World Wide Web Conference in Perth, Australia.