Topics

About this talk

SOAS World Philosophies Lecture 32 Recent debates on generative AI's philosophical implications have largely neglected how cultural particularity is invoked to justify domestically trained large language models. Using contemporary Japan as a case study, we show how current AI discourse rearticulates Nihonjinron, a tradition of cultural essentialism built on claims of Japanese civilizational uniqueness, through the language of LLM development, robot ethics, and AI safety. This rearticulation produces three epistemic distortions: it naturalizes cultural homogeneity as a design constraint, grants domestic researchers unwarranted authority to define moral standards, and forecloses moral universalism without sufficient justification. Culturally essentialist assumptions thus become embedded invisibly in AI development, shaping what counts as safe or intelligent output without philosophical scrutiny. Such epistemic essentialism produces an "illusion of understanding," whereby developers mistake training data's cultural assumptions for genuine knowledge. AI design is not culturally neutral, though acknowledging this need not entail cultural relativism.

→ Book tickets / register (opens in new tab)