The fake news topic is increasing in importance nowadays, with more and more newspapers writing on it. Social media has taken on the battle of creating “fake news websites” lists, as if you could shoot a moving target with a paper plane! It is clear to me that this Internet-age issue is not fully understood and requires up-to-date approaches in order to be solved.
The way fake news propagate is similar to an epidemic. We’ve found this while working on our TrustServista algorithms. Once the original article gains momentum – in the form of mentions, likes, shares in media and social media – it spreads into the digital space beyond control. The origin of the information is lost after less than 1o iterations and soon it ends up being “transcreated” (not translated, but re-written in a different language) into many-many languages. All direct references, links and mentions of the original article are lost, making it very very hard to track down the source or “patient zero”, as we call it.
“Patient zero” is the most important concept when dealing with the propagation of fake news, because identifying it can then lead to detecting all other digital items that were built using that information from patient zero. If you want to imagine the complexity, once Auto BILD incorrectly claimed that BMW rigged their emission test in Sep 2015, more than 500,000 articles were created around the world stating the same thing. Even after BILD changed their post and apologized for the false information (not after BMW’s stocks fell by 10%), those hundreds of thousands of articles, written in some dozens of languages, still exist on the web.
The core of TrustServista‘s “news trustworthiness” algorithms is based on detecting “patient zero”, using advanced text analytics techniques, in order to create a full mapping of how information propagates across the web. The trustworthiness of “patient zero” is the one the ultimately influences all other similar articles’s trust score. In a couple of week’s we’ll be able to demonstrate visually how this concept works and show some real live results on how TrustServista finds “patient zero” of trending news stories. If you want to take part in the prelaunch of TrusServista please register on www.trustservista.com.