Dimitris Papaevangelou and Ilias Dimos, talk about their innovative platform that tracks misinformation.
-this project was realized during the 1st cycle of iMEdD’s #incubator.
Over the last decade, there developed a trend for convergence towards a new, toxic reality in the news industry, that threatens to jeopardize – if it has not already done so – the fundamental values of democracy, undermining the critical thinking and reliability of journalism, through fabricated information, tampered content, reproduction of erroneous perceptions or campaigns for the manipulation of public opinion, affecting not only the very mission of journalists but citizens first and foremost.
Seeking solutions to this problem, we created MediaWatch.
An ecosystem for monitoring, studying and analyzing journalistic content aimed at the detection of “Propaganda Networks” and “Misinformation Chains”, a procedure in which existing information or narratives are transformed into many similar ones, to draw further attention to the specific agenda or context.
This is a system that monitors online media – of any form – in real time and collects information about any important article shared by the media. For each article collected, we “run” a series of algorithms, extracting significant information, such as the title, subtitle, text, media, author, tags, time, etc., as well as details such as the entities (to whom it refers, at what locations, companies, organizations, institutions, etc.). We then search for a specific time margin, in case we have identified a similar article, using previous information. In case such an article exists, we compare the extent of similarity of these articles, using specialized artificial intelligence procedures and natural language processing. All similar articles are grouped, thus creating a network (graph) where the vertices are the articles, media, authors, and entities, while the connection between them (edges) is similarity. We, therefore, map in this way the sharing or development procedure of specific information. Up this point, the procedure is fully automated and at the same time algorithms have no power over the processed content, given that we only search for similarity and do not characterize content.
Find a demo of the application here: https://mediawatch.io/
The idea came about in early 2017 in the form of an open community and the subsequent difficulties we faced were countless! How were we to fund this, how to communicate it without going beyond our role, how to convince journalists to address the issue that they themselves had allowed to spread or how to prevent contribution in its further development.
The iMEdD was also a part of the solution. For a year now, we came to know a huge network of journalists, photographers and industry professionals, learning from them and helping us evolve.
In this context, we managed to turn the community into an organization, the Civic Information Office, a non-profit organization aimed at the development of open source technology for the public interest. We managed to map the Greek – political – Twitter and how Greek politicians use social media to manipulate public opinion (see https://elections.mediawatch.io/). Moreover, we developed the platform in a manner that allows multilingual content analysis.
We currently work to unify the two systems, namely MediaWatch (https://app.mediawatch.io/) and TwitterBots (https://elections.mediawatch.io/) in order to monitor all content in online media and the methods of transmission in social media, aiming at the upcoming American elections.