COWL: Cognitive Warfare Analysis through LLMs

In this Phase I effort titled Cognitive Warfare Analysis through LLMs (COWL), we will focus on assessing key LLM-enabled technologies for presenting qualitative and quantitative emic information resulting from psychographic variables inferred from messages in a variety of media types (e.g. video, audio, images, text) and vectors (e.g. social media, blogs, newswire).  The resulting technology will provide an emic perspective of messages within a geographically bounded region possessing key language indicators (Task 1).  From communication relevant to a regional event, actor or message we shall automatically discover and articulate Frames of Communication (FoCs) which correspond to beliefs shared by all authors evoking the same FoCs. Affect information in the form of stance will be identified for each FoC, and in form of emotions for each communication. FoCs and their stance in turn will enable the discovery of the elements of our psychographic model, by first identifying the cultural and moral values implied by FoCs, and then uncovering sentiment, emotions, goals, attitudes, motivation and intentions, which finally lead to assigned hypothesized behaviors (Task 2).  The emic results will be displayed within a UI which will facilitate analysis as well as assessment of the emic information sought by an analyst, while providing indications of potential outcomes and possible course of action. (Task 3).  The emic information shall be evaluated against reference data originating from crowd sourcing based on cultural expertise, providing a proof-of-concept for potential transition partners.