Algorithmic Adharma? Bias in Large Language Models Interpreting Dharma

Testing three major LLMs on questions of social duty revealed a distinct bias toward caste-neutral but hierarchical interpretations of varna, often citing 19th-century colonial interpretations over classical textual nuances. This analysis details how training data skews AI understanding of 'Dharma' toward a liberal-agnostic framework, potentially erasing the complex, contextual nature of duty described in the Manusmriti and Dharmasutras.

Full article content coming soon.