Abstract

The use of artificial intelligence in peacebuilding and the construction of collective memory represents an emerging opportunity to promote recognition and reconciliation in post-conflict societies without revictimization. This article presents the conceptual and methodological development of Cattleya Llanera, a GPT-based language model designed to facilitate empathetic and context-sensitive dialogue about peace in Colombia. Grounded in the GPT Personalization Methodology based on Life Stories and Dialogue, the model prioritizes the generation of meaningful narratives over information delivery, with the aim of fostering reflection and empathy. The article provides a theoretical framework and a structured methodology for building this AI model, distinguishing it from traditional applications by integrating life testimonies as the core of its identity. While a pilot implementation was carried out after the manuscript was completed, the results are being analyzed and will be presented in a separate study. The technical configuration and development process are documented in a complementary manual that details the model’s internal architecture. This work highlights the relevance of combining technology and human experience to address the challenges of transitional justice and peace education. It sets a precedent for the responsible use of generative AI in conflict mediation, historical memory, and social transformation, while acknowledging the ethical, technical, and methodological limitations that future research must address.