Artificial intelligence and the “Coloniality of Knowledge”: Examining WesternCentric Narratives in Historical Outputs

Abstract

This study examines how contemporary artificial intelligence (AI) systems reproduce Western-centric historical narratives through the concept of “the coloniality of knowledge.” Utilizing Aníbal Quijano’s theoretical framework, the research analyzes how AI-generated interpretations reflect enduring epistemic hierarchies rooted in imperial knowledge production. Specifically, the article investigates whether AI systems privilege Western historical experiences and interpretations over non-Western perspectives. The novelty of this study lies in extending the concept of the coloniality of knowledge—originally developed in the context of Latin America—to the analysis of AI-generated historical narratives, with a specific focus on Siberia and Central Asia. This research broadens the geographical and analytical scope of the concept and contributes to emerging discussions on AI bias in historical knowledge production. A qualitative comparative methodology was employed, focusing on the outputs of three AI systems—ChatGPT, Gemini, and DeepSeek—using identical prompts related to four historical figures: Yermak Timofeevich, Mikhail Skobelev, Mikhail Chernyaev, and Kenesary Kasymuly. The findings reveal consistent patterns across all examined systems, where AI-generated narratives tend to prioritize imperial actors, particularly those associated with the Russian Empire, while portraying Indigenous societies in Siberia and Central Asian political entities as secondary or
reactive. Terms such as “expansion,” “conquest,” and “capture” normalize imperial violence and frame historical processes as inevitable. Moreover, AI systems frequently rely on analytical frameworks derived from Western military historiography, oversimplifying complex local socio-political systems and largely
omitting Indigenous epistemologies and oral traditions. This study demonstrates that AI systems not only reproduce dominant historical narratives but also reinforce the coloniality of knowledge through discursive normalization, the privileging of imperial sources, and the marginalization of local perspectives. These patterns contribute to the persistence of epistemic hierarchies and the reproduction of asymmetrical power relations in digital knowledge production. Overall, the findings highlight the need for more inclusive, contextsensitive, and epistemologically diverse approaches to AI training in order to mitigate representational bias and promote more balanced historical narratives in the age of artificial intelligence.