Autoři

Diana Rotar, Kirena Levytska, Rostyslav Levytskyi, Ishtiaque Ahmed Rafin

Anotace

Diplomacy Pulse AI is a research proposal and locally demonstrated prototype for analysingoverloaded online information environments affected by falsehoods, propaganda, bot activity,and misleading presentation. It is intended chiefly as an analytical tool for journalists,researchers, situation centres, government structures, and businesses requiring verifiedmarket or risk information. Rather than immediately answering a query, the proposed systemgathers multilingual web material, evaluates source credibility, cross-references claims, tracesa story’s origin, and returns structured evidence with confidence or credibility indicators.
A user begins by submitting an analytical question. The main agent interprets the query andperforms web searches rather than immediately composing a response. For the presentation’sUkraine–Poland relations example, the proposed search covered Ukrainian, Polish, andEnglish sources, broadening collection beyond English-language reporting. The system isdesigned for specialised analysis, although the team also described possible business uses,such as comparing competing products, prices, and capabilities before entering a market.Collected sources pass into a classification process. Known sources can be checked againsta trusted whitelist, while cross-source confirmation examines whether similar claims appear inmultiple independent news agencies, analytical centres, or international organisations.Sources outside the whitelist undergo additional reliability and reputation assessment through comparison with other collected material. This stage is intended to prevent heavily repeated,low-quality propaganda from being accepted merely because it appears many times.
An analytical agent then processes the classified material and prepares a report containing theprincipal ideas and conclusions together with their supporting evidence. The intended outputidentifies relevant news items, dates, source ratings, and confidence or credibility indicators.A further proposed capability traces a story’s timeline backward toward its earliest source.During the demonstration, the team showed a local prototype returning principal claims,credibility points, supporting sources, and ratings on an A–D scale. In the Q&A, they clarifiedthat the initial purpose is to flag information as suspicious or apparently reliable, rather thandirectly remove or correct content on external platforms.

licence

Creative Commons BY-SA 4.0