AI systems encode Western moral biases, marginalizing global ethical frameworks in algorithmic decision-making

Indigenous moral frameworks often center on relationality, reciprocity, and ecological stewardship, which are incompatible with the atomistic, extractive logic embedded in LLMs. For instance, Indigenous legal traditions like the Māori concept of 'kaitiakitanga' (guardianship) or the Andean 'ayni' (reciprocal labor) are entirely absent in AI training data, despite their relevance to global ethical challenges. The erasure of these traditions reflects a deeper colonial pattern where Indigenous knowledge is either appropriated as 'data' or dismissed as irrelevant. Reclaiming these epistemologies requires dismantling the Western-centric AI development pipeline.

The dominance of Western moral frameworks in AI reflects centuries of epistemic extraction, from colonial-era anthropological classifications to Cold War liberal individualism. The 20th-century rise of behavioral economics and cognitive psychology further entrenched Western moral assumptions as 'universal,' a legacy now embedded in AI training data. Historical precedents like the Universal Declaration of Human Rights (1948) reveal how Western liberal ethics were globalized without adequate representation of non-Western traditions. This historical amnesia in AI ethics perpetuates the same power imbalances that shaped earlier waves of globalization.

Cross-cultural comparisons reveal that moral reasoning is not a monolith but a spectrum shaped by cultural context. For example, while Western ethics often frame morality as a set of individual choices (e.g., 'rights'), many non-Western traditions view it as a collective practice (e.g., 'duties' in Confucianism or 'harmony' in Daoism). The conflation of Western liberal individualism with 'universal morality' in LLMs is a form of cultural imperialism. Integrating diverse moral frameworks could improve AI's adaptability and ethical robustness across global contexts.

The study likely uses Western psychological frameworks (e.g., Kohlberg's stages of moral development) to evaluate LLM outputs, which may not capture non-Western moral reasoning. Scientific evidence shows that moral judgment is culturally situated, yet AI training data often aggregates diverse sources without accounting for these differences. The lack of interdisciplinary collaboration with anthropologists, ethicists from the Global South, and Indigenous scholars further undermines the scientific validity of these models. Future research must adopt participatory, culturally grounded methodologies.

Artistic and spiritual traditions often encode moral frameworks in ways that resist Western rationalism, such as through storytelling, ritual, or symbolic representation. For example, Indigenous oral traditions convey ethical lessons through metaphor and collective memory, while Sufi poetry explores moral dilemmas through spiritual metaphor. These forms of knowledge are incompatible with the binary, decontextualized outputs of LLMs. Recognizing these dimensions could humanize AI by integrating moral reasoning with emotional and spiritual intelligence.

If unaddressed, the current trajectory will entrench Western moral biases in global AI systems, exacerbating cultural conflicts and epistemic injustice. Scenario planning suggests that by 2035, LLMs could become dominant arbiters of 'global ethics,' sidelining non-Western traditions in policy and education. However, alternative futures are possible: AI could be redesigned to function as a 'moral translator,' facilitating dialogue between diverse ethical systems. The key is to embed pluralism into AI governance frameworks before these systems become irreversible.

Marginalized communities—particularly Indigenous peoples, Global South scholars, and non-Western ethicists—are systematically excluded from AI ethics discourse despite bearing the brunt of algorithmic harm. Their knowledge systems are either commodified as 'training data' or dismissed as 'unscientific.' The research itself is led by a graduate student in a Western institution, highlighting the lack of indigenous leadership in AI ethics. Centering these voices requires structural changes, including funding mechanisms, publication platforms, and decision-making power for marginalized communities.