Digital Silence in the Age of Generative AI: How Algorithmic Hegemony Obscures Non-Western Cultural Heritage

This article focuses on how generative AI and search algorithms marginalize non-Western cultural heritage through language bias, Western-centric data prioritization, and shifts in knowledge paradigms, triggering a crisis of cultural visibility in the digital age. Technology is not neutral; it carries interwoven data, history, and power. We must confront algorithmic hegemony and promote decolonial technological practices.

The suppression of non-Western knowledge systems by colonialism did not end with political decolonization; instead, it continues in digital form. The training data for current large language models (LLMs) and search engines is heavily biased toward Western languages and perspectives, and the "objective" results of data availability mask deep structural inequalities.

Mainstream AI systems have significantly insufficient ability to process non-Latin scripts and non-English content. Users often get poor results when querying non-Western cultural heritage. The logic of algorithmic optimization (pursuing click-through rates) leads to a vicious cycle of low exposure → low citations → low weight for non-Western content, further marginalizing it.

Traditional living cultural memories (oral history, rituals, handicrafts, and other non-textual, contextual knowledge) are systematically excluded from digital archives because they do not conform to the structured, standardized formats preferred by algorithms. This is epistemic violence that negates the intrinsic value of non-Western knowledge systems.

Generative AI defaults to outputting Western cultural examples (such as Greek architecture, symphonies), reinforcing cultural hierarchy; its generated content is scraped to train the next generation of models, forming a self-reinforcing feedback loop that further dilutes non-Western cultural voices.

The digital absence of non-Western cultural heritage has multiple impacts: knowledge homogenization in education; difficulty for non-Western creators to earn traffic revenue in the economic sphere; suppression of marginalized groups' narratives in the political sphere; and the public's understanding of human civilization being narrowed to a synonym for the West in the cognitive sphere.

Solutions include: proactively including multilingual/cultural training data at the data level (in collaboration with community partners); introducing bias compensation mechanisms in algorithm design (such as weighting for low-resource languages); incorporating fairness indicators like cultural representation into evaluation standards; and empowering communities to build independent digital archives.

Technological development is not neutral; design decisions reflect cultural assumptions and power relations. Generative AI can be a cultural bridge or a tool of hierarchy—the key lies in confronting algorithmic hegemony. Reconstructing digital cultural justice requires joint efforts from technical experts, policymakers, cultural workers, and marginalized communities to appreciate the full spectrum of human civilization.