Resilience Ukraine

CS-DC is very supportive of our scientific colleagues in Ukraine.

The UK Open University supports Ukraine through one of its Open Societal Challenges:

Press the back button to return: Link to Open Societal Challenge - Resilience Ukraine

AI4TRUST.: Artificial Intelligence for Thwarting Russian Disinformation.

A consortium of Greek, Ukrainian, Swedish and UK partners were recently awarded funding from NATO for the project:

The project addresses a critical and escalating security challenge: the rapid spread of fake news in the context of Russia’s war against Ukraine and its hybrid war against NATO countries, including electoral disinformation, which poses a serious threat to democracy, political stability, and the security of NATO member states. These threats manifest as deliberate falsehoods designed to mislead policymakers and voters, deepfake videos used to incite hatred or defame public figures, and a broader erosion of trust in online and traditional media. Campaigns today employ advanced techniques such as deepfakes, coordinated bot-driven amplification, and psychological operations (PSYOPs) designed to inflame divisions and exploit cognitive biases. This is exacerbated by the growing sophistication of AI-generated content. Since such content becomes more lifelike and harder to detect, the public's ability to reliably assess the credibility of information is eroded. Social media, instant messaging, and digital news platforms enable the mass dissemination of false content at unprecedented speed and scale, making it increasingly difficult for institutions and citizens to separate fact from fiction. This leads to widespread skepticism, not only toward misleading content but toward all forms of media, undermining the foundation of informed democratic participation. The project recognizes that disinformation is not merely a communication issue, but a national security threat, particularly in the context of hybrid warfare. By focusing on the detection, analysis, and thwarting disinformation, especially in its most advanced, AI-enhanced forms, the project will contribute directly to the Alliance’s strategic objectives of preserving democratic governance, protecting public discourse, and reinforcing social cohesion.

Disinformation is the deliberate construction and propagation of falsehoods. A fundamental challenge for detecting disinformation is the absence of a universally accepted methodology for determining the truth or falsehood of information entities. Addressing this requires the development of new formal and algorithmic approaches to ‘truth’ and reasoning, particularly in complex and ambiguous information environments in different content formats and languages. Countering disinformation requires falsehoods to be distinguished from non-falsehoods.

Formal logics define non-falsehoods by truth sets (binary, multivalued, fuzzy, non-monotonic, etc.), with truth values assigned to well-formed strings of symbols (sentences, arguments) by logical and grammatical rules. Our approach goes beyond this to consider messy ill-formed sentences or slogans that have no absolute meaning but have meanings for some people. These meanings and beliefs in their non-falsehood by individuals and groups change through time. We will develop a new relational mathematical theory for machine-representation of the highly dynamic multilevel structures of sentences and arguments in the creation, propagation, and evolution of false sentences and narratives. Using a unique database of international newspaper articles collected by University Institute of Lisbon over the last ten years, we will create new kinds of multilevel large language models (LLMs).

This will enable comparative analysis of articles in Russian, English, Ukrainian, Greek and Swedish for their multilevel falsehood or non-falsehood. We will create a new AI-based methodology to detect and counter the propagation of Russian disinformation. This will involve developing a methodological and analytical toolkit grounded in the latest advances in artificial intelligence (AI), machine learning (ML), and data science. It will be designed to identify and interpret the features of disinformation, manipulation, PSYOPs, and fake news in a multilingual environment. This goes far beyond simple translation, requiring a deep understanding of the linguistic, cultural, and contextual nuances that shape how disinformation is framed and perceived within different language communities. Addressing this challenge demands both linguistic adaptability and semantic sensitivity in model development. Modern disinformation campaigns often combine text, imagery, audio, and video to produce persuasive, manipulative narratives. Analysing such content requires models that can interpret each modality independently, while also detecting inconsistencies and manipulations across them. The project will therefore build systems capable of multimodal fusion and disaggregation, enabling robust cross-verification and holistic content understanding.