header.html

An official website of the European Union How do you know?      
European Commission logo

handle.jsp

cover
Policy analysts working on conflict early warning, prevention, and crisis response increasingly rely on heterogeneous data sources. The European Commission’s Joint Research Centre (JRC) already produces conflict forecasting models, trend analyses and data briefs that support early warning, but building on these outputs in applied work remains challenging. Analysts work under time constraints and often lack the technical expertise required to analyse structured and unstructured data and draw insights from them. This report presents the Conflict Risk Analysis (CoRA) AI Assistant prototype, a multi-agent system based on large language models (LLMs) that complements the JRC’s existing conflict early warning tools. By enabling natural-language interaction with integrated datasets offering global coverage and daily to monthly updates, the CoRA-AI Assistant aims to lower the technical barrier to working with conflict data. The system orchestrates specialised components for query classification, code generation, document retrieval, and synthesis, allowing users to produce outputs that combine quantitative results with qualitative context. The project team evaluated the system through user testing, interviews, surveys, and written feedback. Overall, users perceived the CoRA-AI Assistant as a useful first-pass analytical and briefing tool, while also identifying limitations in handling complex queries, time-related queries, and multi-step interactions. Some of these limitations may be related to the underlying LLM architecture and system configuration used during testing, although the evaluation does not establish a direct causal relationship. The report outlines priorities for further development, including usability, analytical robustness, and data coverage.
2026-08-27
Publications Office of the European Union
JRC147669
978-92-68-42598-5 (online),   
1831-9424 (online),   
EUR 40821,    OP KJ-01-26-350-EN-N (online),   
https://publications.jrc.ec.europa.eu/repository/handle/JRC147669,   
10.2760/5009002 (online),   
NameCountryCityType
Datasets
IDTitlePublic URL
Dataset collections
IDAcronymTitlePublic URL
Scripts / source codes
DescriptionPublic URL
Additional supporting files
File nameDescriptionFile type 

footer.html