Attempting to deradicalise people using Facebook Messenger

Facebook Messenger has been used to try to deradicalise extremists in a pilot project funded entirely by the company. People posting extreme far-right and Islamist content in the UK were identified and contacted in an attempt to challenge their views. Of the 569 people contacted, 76 had a conversation of five or more messages and eight showed signs it had a positive impact, researchers claim.

That's a pretty dire "success" rate. And how many people showed signs of a negative impact ? Without that, this is meaningless.

This pilot was led by the counter-extremism organisation Institute for Strategic Dialogue (ISD), which says it was trying to mimic extremists' own recruitment methods. It told the BBC's Victoria Derbyshire programme and BBC World Service's World Hacks it used software to scan several far-right and Islamist pages on Facebook for targets. It then manually looked at their profiles looking for instances of violent, dehumanising and hateful language.

It employed 11 "intervention providers" - either former extremists, survivors of terrorism or trained counsellors, who were paid £25 per hour for eight hours' work a week.

One was Colin Bidwell, who was caught up in the Tunisia terror attack in 2015. Under a fake profile, he spoke to people who appeared to support Islamist extremism, including some who may support the Tunisia gunman, and was tasked with challenging their views with chatty conversation and questions. "I think I'm entitled to ask those questions after what I've been through," he explained. "If there's the smallest chance that I could make some form of difference or awareness, for me I'm in."

Many did not respond, but some entered into long conversations. Mr Bidwell would talk a little about religion, about the effect the attack has had on his wife and how he worries for the future of his children in "such a violent world". "One of the things I would say is, 'You can have your extreme beliefs, but when it gets to the extreme violence - that's the bit I don't understand'," he said.

The aim was to "walk them back from the edge, potentially, of violence", said Sasha Havlicek, the chief executive of the ISD. "There's quite a lot of work being done to counter general propaganda with counter-speech and the removal of content, but we know that extremists are very effective in direct messaging," she explained.

Privacy campaigners are concerned about the project, especially that Facebook funded something that broke its own rules by creating fake profiles. Millie Graham Wood, a solicitor at the Privacy International charity, said: "If there's stuff that they're identifying that shouldn't be there, Facebook should be taking it down. Even if the organisation [ISD] itself may have been involved in doing research over many years, that does not mean that they're qualified to carry out this sort of... surveillance role."

During conversations, the intervention providers did not volunteer the fact that they were working for the ISD, unless asked directly. This happened seven times during the project, and on those occasions the conversation ended, sometimes after a row.

Looks to me like the fraction showing a negative result could easily be as great as that showing a positive impact. This would be far more interesting with a much larger sample. Then one could start to compare if there was any correspondence between the methods different "intervention providers" use and their positive or negative results, whether this depends on the particular demographics of the extremists contacted, etc. Without that, I don't think there's much to be learned here.
http://www.bbc.com/news/technology-43170837