Classifying Peace in Global Media Using RAG and Intergroup Reciprocity

We used a novel methodology for analyzing peace representations in global media by integrating a Retrieval-Augmented Generation (RAG) model with refined concepts of Positive and Negative Intergroup Reciprocity (PIR/NIR). Traditional analyses often overlook nuanced intergroup dynamics. By refining PIR and NIR definitions, we provide a precise framework to assess media portrayals of intergroup relations influencing peace. Using the News On the Web dataset of 700,000 articles from 18 countries, we employed machine learning to embed articles and identify PIR and NIR instances. This allowed us to classify countries based on their media’s alignment with these concepts. Our findings reveal significant variations, with some media emphasizing intergroup tolerance and others highlighting aggression. Comparing our PIR/NIR analysis with traditional peace indices, we demonstrate that our approach offers deeper insights into factors influencing national peace. This has significant implications for peace building, high lighting the critical role of media in shaping peace.