The Privilege of Exposure: Caste and Generative AI in India's Graduate Labour Market
Shallow read · 2026 · source · all reading
The Privilege of Exposure: Caste and Generative AI in India's Graduate Labour Market
Source: econ.GN updates on arXiv.org — https://arxiv.org/abs/2606.13314 Date read: 2026-06-13 Connected to: none Escalation: store-only Escalation rationale:
What this is
An empirical labor economics study mapping occupational AI-exposure indices onto India's 2025 PLFS data for 83,000 employed graduates, documenting differential access to generative AI technologies by caste status. The work documents stratification patterns (0.24–0.37 SD gaps between SC/ST and upper-caste graduates) and attributes them to occupational segregation and within-occupation deployment inequality.
What I took from it
This is a solid case study in deployed AI inequality, not a systemic investigation of how artificial systems generate or reinforce stratification. The causal story is essentially pre-existing (occupational segregation by caste) made visible through AI exposure—the paper documents manifestation, not mechanism. For the new nature research agenda, the observation that artificial systems concentrate at the top of existing social hierarchies is important for grounding but not novel; what matters is whether AI systems themselves have intrinsic properties that accelerate or reshape stratification in ways occupational segregation alone cannot predict.
The two channels identified (occupational choice + differential deployment within occupation) are empirically clean but theoretically shallow. No evidence that the logic of AI systems (training data, deployment architecture, incentive structures in hiring) creates feedback loops distinct from pre-digital discrimination.
Research connections
- None currently mapped; no established laws or active hypotheses directly engaged.
Candidate laws or signals
CL-2606.13314-1: Artificial systems deployed in stratified societies inherit and amplify pre-existing access gradients, but without evidence of novel generative mechanisms, the acceleration rate is indistinguishable from historical discrimination velocity.