Historical Revisionism Supercharged: Large Language Models (LLMS) in Education in Japan

Volume 10, Number 1

Historical Revisionism Supercharged: Large Language Models (LLMS) in Education in Japan

Authors

Shanyi Yang , The University of Hong Kong, China

Abstract

This paper argues that the current structure of large language models (LLMs) implemented in Japanese schools, especially primary and secondary classrooms, constitutes a supercharged mode of historical revisionism. LLMs are structurally disallowed to maintain internal world models and thus rely on existing corpora for outputs, which, in Japan’s context, risks perpetuating existing historical revisionist texts saturated in Japanese-exclusive training data. The state’s caution against “hallucinations,” coupled with the contemporary rise of revisionist groups such as Tsukurukai, creates a structural symmetry between narratives where majoritarian revisionist historiography is more likely to thrive. Moreover, Japan’s cultural and political (robot) anthropomorphic animism, also perpetuated by the state and especially susceptible in children, means that LLMs may exist on the same plane as humans in the making and recording of history, which extrapolates the flawed human essence to Japanese imperial atrocities, relativizing them as inculpable narratives. Ultimately, LLMs’ structural oblivion creates a form of historical revisionism two orders above simple redaction: to manufacture an epistemology of symmetrical narratives over historical “truth” and to imbue the transcendental human spirit, kami, in the nation-state and its atrocities in the form of “narrative humanism,” depoliticization being revisionism’s final goal.

Keywords

Large language models, historical revisionism, education in Japan, collective memory