cubed rumour check Guardian  ©

The Guardian Limited

Image above: screen shot from The Guardian's interactive visualisationof rumours spreading during the London riots in 2011

A team of scientists are developing a system to check potentially dangerous rumours circulating in social media

There’s a rumour going round that gossip isn’t all that bad. But in the age of smartphones and social media, unchecked rumours that circulate instantly can have a seriously negative impact. It’s why an EU-funded project comprising five universities – Sheffield, Warwick, King’s College London, Saarland in Germany and MODUL University Vienna in Austria – worked on developing a ‘lie detector’ for checking social media. The three-year project had been named after the mythological Greek Goddess of rumour – ‘Pheme’. Lead researcher Dr. Kalina Bontcheva, from the Department of Computer Science in the University of Sheffield’s Faculty of Engineering explains that the work began back in 2011 and 2012, looking at rumours circulating in the London Riots.

A multi-disciplinary team lead at the time by Rob Procter who is now Professor of Social Informatics at Warwick University manually analysed the tweets circulating, including the evocative rumour that animals had been released from London Zoo. ‘They identified different kinds of misinformation, disinformation and speculation,’ explains Dr Bontcheva, ‘mostly focussing on seven rumours which were pure disinformation.’ Disinformation is rumour intended to deceive.  Professor Rob Procter, a team from the London School of Economics and The Guardian newspaper later produced an interactive visualisation of the spreading rumours. 

Automated Rumour Check

Dr Bontcheva has been working on ‘text mining’ since 1996, originally analyzing news and longer media before moving into the emerging world of social media. The current project aims to automate rumour-checking in real time – the previous study had been all manually done. They are looking at four types of rumours – speculation, controversy, misinformation and disinformation – each with specific characteristics.  

For example says Dr Bontcheva, the area of ‘speculation’ concerns activities such as conjecture on  ‘whether the Bank of England is going to raise interest rates. You don’t know until proven otherwise. The real challenge for us is how to recognise these things automatically and the different properties that they have over time.’  The system will assess where the information is coming from, for example from an established journalist or an instantly created twitter account.

Healthcare Rumour

It’s not just news events that rumour builds and circulates around. Healthcare professionals already factor in the internet search by patients of their symptoms, through official information sites or through unofficial patient forums. One of the research areas Dr Bontcheva is most excited by is the use for medical professionals addressing controversial issues in the public space – Alzheimer’s disease for example, and the debate in press, social media, and in medical publications around the impact of aluminium.  The system the team are developing would help unpack medical rumours. Is a certain kind of information, ‘more in line with what the medical establishment and mainstream opinion is? Or is it not? It doesn’t make it wrong or the patient forum wrong,’ says Dr Bontcheva. But it helps doctors advise on the trustworthiness or otherwise of the data. 

While the internet breeds ‘viral’ information and misinformation – to use a biological metaphor – technology also develops new technological ‘antibodies’ to help cure the effects of this misinformation. ‘That idea of ‘antibodies’ is a nice way of looking at it,’ says Dr Bontcheva reflecting on how this app may address what happens when we worry ourselves unnecessarily. ‘It is pretty much straightforward to convince yourself that you have got a particular disease, whatever it seems to be.’ 

External links