Holistic Analysis of Organised Misinformation Activity in Social Networks – HAMISON

Project Summary

Given a scenario of organised intentional misinformation campaigns, often termed as disinformation, we as a society must be aware not only about fake news, but also about the agents that introduce false or misleading information, their supporting media, the nodes they use in the social networks, the propaganda techniques they use, their narratives and their intentions.

Therefore, we must address this challenge in a holistic way, considering the different dimensions involved in the spreading of disinformation and bring them together to really identify, characterise and describe the orchestrated disinformation campaigns. At the message level, we will explore: claim worthiness checking, stance detection and multilingual verified claim retrieval; at the social network level, we will model disinformation propagation and apply social network analysis techniques to identify sources and main players.

Then, the challenge is how to integrate both levels. To address this challenge we must be aware about the intentionality of disinformation: agents that create and introduce disinformation in the social media networks carefully select narratives aimed to have a concrete impact such as polarise, destabilise, generate distrust, destroy reputation, etc. This adversarial game has, at the end, benefited and injured agents. Therefore, we must also address the identification of these malicious intents and bring everything together to collect all the evidence and give it to final analysts and users in explainable ways.

Identifying misleading messages, knowing their narratives and hidden intentions, modelling the diffusion in social networks, and monitoring the sources of disinformation will also give us the chance to react faster to the spreading of disinformation. Thus, the project is articulated around theses goals: (i) Identify disinformation (claim worthiness checking, stance detection and verified claim retrieval); (ii) Analyse the sources of disinformation and their narratives; (iii) Model the propagation of disinformation; (iv) Develop demonstration applications on video shorts and clips, regular YouTube videos and Tweets; (v) Create evaluation datasets of Tweets and videos in English, Spanish, German, French and Estonian; and (vi) Organize shared tasks for competitive evaluation on stance detection in twitter and claim-checking worthiness on videos.


Relevance to the call

The project will apply Natural Language Processing, Social Network Analysis and Artificial Intelligence techniques (Machine Learning, Multi-Agent based simulation, epidemic modelling, audio transcription, image recognition, etc.) to aid analysts in the identification of disinformation in messages and video streams in several platforms. Among others, we will develop new methods for stance detection, claim worthiness checking, verified claim retrieval and disinformation propagation modelling. The project will release datasets in several languages (Spanish, German, French, including a very low resource language such as Estonian) and modalities (text and video). We will organise shared tasks for competitive evaluation in other languages than English.

By modelling the propagation of disinformation in social networks we aim at developing tools that will allow analysts to evaluate the effect of mitigation actions in the spreading of disinformation. By analysing the sources of disinformation, their narratives and modelling their propagation on social networks we expect to develop tools to raise in society a deeper understanding of misbehaviour in media and help to anticipate the spreading of disinformation. In particular, making explicit the intentions behind the messages, the propaganda techniques they use, the sources that support it, their subjectivity, stance and bias degree will increase awareness in society, helping people to assess the quality of the information.