A Social Media Lie Detector Hopes to Quash Viral Rumours

The art of lying is as old as conversation, but with the development of social media we’ve seen quite how devastating the consequences of a vicious rumour or or accidental untruth can be. Publishing falsehoods on the web leaves a permanent record that spreads a lot quicker than a whisper at the back of the classroom, and the result can be a tweet-fuelled mess of misleading information, complete fabrications, and viral hoaxes. In turn, they can lead to unfounded panic, confused emergency responses, false allegations, and libel cases.

To combat the problem, an EU-funded project is developing a “lie detector” for social networks; a system that could analyse in real-time whether a piece of information is true or false. They highlighted accusations of vote-rigging in Kenya, allegations that Barack Obama was Muslim, and claims that animals from London Zoo were set free during the 2011 riots, as examples of social media rumours that could have substantial real-world effects.

“In all of these cases—and many more—an ability to quickly verify information and track its provenance would enable journalists, governments, emergency services, health agencies and the private sector to respond more effectively,” a release from the University of Sheffield, which is leading the project, explained.

RT: London eye on fire!!!!! Is this true??? #Londonriot http://twitpic.com/636s8r

— Zadio (@zadi0) August 9, 2011

The idea behind the system is to stop false information spreading before things get out of hand. “People do believe things they hear on the internet,” lead researcher Kalina Bontcheva told the Times. “In critical situations, you can instead show reliable information or alert the authorities before things get out of hand.”

The lie detector system is called Pheme—a term coined to suggest “memes which are enhanced with truthfulness information,” and a reference to the Greek goddess of fame whose wrath consisted of scandalous rumours. On the EU project page, the researchers explain that their system will divide suspicious social media comments into four types of pheme: speculation, controversy, misinformation, and disinformation. The difference between those last two is that misinformation is spread unwittingly—people don’t realise it’s not true—whereas disinformation is disseminated knowingly, with malicious intent.

But automatically categorising comments into those groups is difficult without an awareness of the context surrounding the words; if something’s presented as fact, it reads like fact. Pheme will look at the source of the rumour, on the basis that news outlets and experts are more reliable  (though they’re not immune to social media fails). Automated bots and newly created Twitter accounts, on the other hand, are less trustworthy.

Corroborating sources will help a rumour pass the detector, and it will also analyse the content of the message and cross-reference it with trusted resources. In the case of medically themed assertions, this could be compared with information in the online medical publications database PubMed.

Bontcheva told the Times she hoped to have a final version of the system within 18 months, and would have working prototypes sooner. There are already a couple of real-world applications set up: the Swiss Broadcasting Corporation will test the system in relation to online journalism (which has been known to pick up on social media “stories” that turn out to be false) and King’s College London has an interesting proposal to test it in the health domain. They want to track what recreational drugs are trending online, and see how long it takes for them to appear in real-world patient records.

Given the insidious and relentless nature of lying on the internet, it seems unlikely any program could totally quash false rumours. But just flagging up potentially untruthful information acts as a useful reminder that everything we read online isn’t necessarily true, and that it’s our job to sift out the lies with an scepticism too often absent from cyberspace. And if it stops another of those phony “eavesdropped” arguments or made-up napkin note exchanges from garnering a few minutes of desperate internet fame, that’s got to be a good thing.