Distinguishing Between Truth and Fake: Using Explainable AI to Understand and Combat Online Disinformation

Disinformation campaigns have become a major threat to democracy and social cohesion. Phenomena like conspiracy theories promote political polarization; they can influence elections and lead people to (self-)damaging or even terrorist behavior. Since social media users and even larger platform operators are currently unready to clearly identify disinformation, new techniques for detecting online disinformation are urgently needed. In this paper, we present DeFaktS, an Information Systems research project, which takes a comprehensive approach to both researching and combating online disinformation. The project develops a data pipeline in which (i) messages are extracted in large quantities from suspicious social media groups and messenger groups with the help of annotators. Based on this corpus, a Machine Learning-based System (ii) is trained that can recognize factors and stylistic devices characteristic of disinformation, which will be used for (iii) an explainable artificial intelligence that informs users in a simple and comprehensible way about the occurrence of disinformation. Furthermore, in this paper an interdisciplinary multi-level research approach focusing on media literacy and trust in explainable artificial intelligence is suggested in order to operationalize research on combating disinformation.