AI Knowledge Graph Reveals 1,200-Year Philosophy Attribution Errors

A new research paper by Joy Bose tackles attribution bias in philosophical knowledge graphs, using a custom-built darshana-graph of 28,322 relationships across Hindu, Buddhist, and Jain traditions. The study shows that traditional corpus-frequency methods misattribute concepts to the school that mentions them most, conflating textual power, historical priority, and philosophical significance. For example, the concept of moksha is commonly attributed to Advaita Vedanta, but it first appears in Jain sources over 1,200 years earlier. Seven of the top 25 concepts by betweenness centrality predate their attributed school by 288 to 2,288 years. This bias is exacerbated between 300 CE and 800 CE, when the network grew from 18 to 1,028 nodes—97.4% carrying Advaita proxy dates—revealing that apparent dominance reflects textual survival, not philosophical history.

Bose proposes temporal sourcing as a corrective, taking a snapshot at 300 BCE using only explicitly dated sources. That snapshot reveals a genuinely pluralistic structure: 59% Vedic, 24% Jain, 18% Buddhist. Beyond correcting attribution, the temporally grounded graph enables structural homology analysis. Ego-network feature vectors applied to 48 temporally labelled concepts across eight traditions identify similar concept pairs across traditions. Known correspondences like purusha-jiva (Samkhya/Jain, similarity 0.990) and prakriti-maya (Samkhya/Vedic, 0.972) are recovered. Novel pairings include nibbana and samsara (0.954)—doctrinal opposites that both function as the ultimate reference concept in their soteriology—and cetana (Buddhist intention) with ajiva (Jain non-living matter, 0.923), a pairing absent from existing literature. These are structural homologies, not claims of doctrinal equivalence, showing how different philosophical vocabularies navigate a shared conceptual space.