Language and the framing of historical narrative in large language models: the case of Asia minor (1922)

This study examines linguistically dependent bias in the generation of historical narratives by Large Language Models (LLMs) and its ethical and societal implications. Through a qualitative comparative case study focusing on the events of 1922 in Asia Minor, the research analyzes how the language of the prompt is associated with terminology, attribution of responsibility, and historiographical framing in outputs produced by GPT-4o and Claude 3.5. The findings suggest that LLMs do not generate neutral historical accounts; rather, they exhibit algorithmic mirroring, aligning the structure and meaning of historical content according to linguistic input. This process results in divergent emotional receptions and interpretations of the past. Particular emphasis is placed on the ethical significance of restricted pluralism, as language-dependent narrative framing may constrain exposure to alternative historical perspectives through mechanisms of selective omission. The study underscores that language functions as a crucial framing mechanism in the construction of historical knowledge in the age of Artificial Intelligence, raising critical ethical questions regarding epistemic responsibility, historical mediation, and the role of LLMs in shaping collective memory. It concludes by highlighting the need for critical digital literacy and the responsible design and use of generative AI systems in knowledge-sensitive domains.