Supplementary Dataset for "Discursive Behavior of Generative Language Models on Geopolitical and Humanitarian Topics"

This dataset accompanies the manuscript *“Discursive Behavior of Generative Language Models on Geopolitical and Humanitarian Topics: A Comparative Analysis”. It contains the full coded dataset (10 prompts × 5 models = 50 responses), coding matrices, distribution counts, high-resolution figures, and a sample of raw responses. Prompts are provided in both Italian (original wording) and English (translations). Raw model responses are reproduced in Italian, the original language of data collection, while coding of tone, framing, and alignment was performed directly on the Italian texts. The dataset is intended to support transparency and reproducibility in the study of discursive behaviors of large language models (LLMs) on sensitive geopolitical and humanitarian questions. It may serve as a resource for researchers interested in AI ethics, bias, framing analysis, and computational social science.