Leiden University: New way to rapidly detect fake news

0

With the emergence of social media platforms such as Twitter, Facebook and Instagram, it’s easier than ever to share information. Including disinformation. During his PhD computer scientist Xueqin Chen developed a new way to recognise fake news and predict how messages spread within online social networks using artificial intelligence (AI). Chen received his doctorate last week on October 25.

How does fake news spread?
It’s becoming increasingly difficult to distinguish news and fake news from each other. Chen managed to develop a technique that detects fake news in a way that is faster and better than previously possible. ‘Our new technique does not use images or texts to detect the difference between news and fake news. Using deep learning techniques, we look purely at the way messages spread within an online social network, for example Twitter.’

How does it work? ‘Fake news is often sensational. As a result, the diffusion pattern is different from real news. For example, fake news is shared more often and reaches a wider audience faster,’ explains professor of computer science Marcello Bonsangue of the Leiden Institute of Advanced Computer Science (LIACS). He supervised Chen during his PhD. ‘Chen has now shown, for the first time, that it’s possible to use features of distribution patterns to quickly and accurately identify fake news and predict whether a message will go viral, without looking at the content of the message. That is unique.’

A new algorithm
Quickly detecting false information is now more important than ever, according to Bonsangue. ‘It would be great if we could eventually reduce the spread of fake news by developing systems that automatically filter out fake news,’ Chen adds. This means that if a message is recognised as fake, it will no longer appear on your screen as a suggestion. Chen: ‘Unconsciously, we all deal with automatic suggestions every day. Think, for example, about the list of suggested songs on Spotify or Google advertisements. This makes this field of research so interesting to me.’

Endless applications
So far the deep learning technique that Chen developed has only been used to analyse information distribution within online social networks. Chen: ‘At the moment, my research is still in an early stage, but the possibilities are endless.’ As a postdoc, Chen has now moved to Delft to apply a similar framework based on AI to the management of water systems in the Netherlands. ‘We could, for example, possibly use it to predict flooding,’ Chen says proudly.

‘The spreading of fake news is different from real news’

Bonsangue also sees a lot of potential, especially within the medical field. But: ‘We’re not there yet. The next step is to find out what AI ​​programs exactly base their choices on. We have shown that it works, but the automatic explanation of the programs themselves are still lacking.’ According to the professor, this is one of the biggest challenges of today’s AI techniques. ‘As long as we cannot explain exactly what’s happening, it’s hard for people to trust decisions based on AI.’