AI could help reduce the spread of false information

A team of academics from the US and the UK has been working on developing a machine-learning tool that could predict when newly registered internet domains will be used to spread false information. Hence, these sites can be shut down before the misinformation has the chance to escalate.

Bu using domain registration data with Mozilla web browsing data, the academics have been able to create a machine-learning classifier that can determine which websites will spread the misleading content.

False information is used to describe misinformation and made-up news that look authentic and that serve an agenda rather than the public interest. It has become a major concern following the 2016 US elections, where a Russian campaign gained influence. Since then, there are more concerns about what is true and what is not.

The research has been focused on the role websites play in facilitating the spread of false information. Indeed, websites are easy to install and cost nothing so, once the false information is in the system, the people behind it only have to rely on the viral nature of communications to carry out the message across other platforms.

The project thus aims to identify the websites that doing so as early as possible to prevent the spread before it begins. The early-identification system will be able to help policy-makers in deploying their resources quickly and efficiently and prioritizing sanctions or monitoring for certain domains.

In order to develop it, academics rely on various data points available from public domain registrations. By using machine learning and automation, they can detect and stop more quickly corrupted websites.

For now, it would seem that the classifier works quite well, with 96.2% of false information domains identified, and they hope that it will help policy-makers to reduce the possibility of taking actions built on possible false positive classifications.

More
articles