From the Field to the Algorithm: Understanding Indian Ethnographers' Perspectives on Responsible AI

Paper Title

From the Field to the Algorithm: Understanding Indian Ethnographers' Perspectives on Responsible AI

Publication Info

• Topic area: Ethnographers' perspectives and contributions to Responsible AI in the Indian context.
• Keywords: Responsible AI, ethnography, India, AI governance, cultural grounding, power dynamics, AI policy, methodological innovation, interdisciplinary collaboration, AI ethics.

Background and Problem

• Problem / challenge: Existing Responsible AI (RAI) frameworks are often Western-centric and fail to address India's unique social realities, including caste, class, and linguistic diversity. Ethnographers, who can provide critical contextual insights, are largely excluded from AI development and governance.
• Significance: The exclusion of ethnographic expertise leads to AI systems that perpetuate harm and fail to align with the lived realities of diverse Indian communities, undermining ethical and effective AI deployment.
• Motivation and related work: While ethnographic methods have been used to study AI impacts, little is known about how ethnographers themselves perceive RAI or how they can contribute to its development. Prior work critiques Western-centric RAI frameworks but lacks empirical accounts of ethnographers' roles and barriers to their inclusion.

Solution

• Proposed approach: The study explores Indian ethnographers' perspectives on RAI through workshops and interviews, identifying critiques, methodological innovations, and pathways for their integration into AI governance.
• Novelty:
  1. First empirical documentation of Indian ethnographers' understanding of and engagement with RAI.
  2. Critical assessment of India's RAI policy and its alignment with local social realities.
  3. Concrete methodological proposals for embedding ethnographic expertise in AI development and governance.
  4. Identification of structural barriers and strategies for integrating ethnographers into RAI ecosystems.
• Procedure and key techniques:
  • Conducted a one-day workshop with 20 Indian ethnographers, followed by semi-structured interviews.
  • Reflexive thematic analysis of workshop and interview data to identify knowledge pathways, critiques of RAI frameworks, methodological innovations, and structural barriers.

Results

• Concrete findings:
  • Knowledge heterogeneity: Only 3 participants had direct RAI experience; most engaged indirectly through seminars or fieldwork.
  • Critiques of India's RAI policy: The policy is seen as industry-tilted, inattentive to caste and class dynamics, and reliant on abstract ethical frameworks (helpfulness, harmlessness, honesty) that fail without contextual grounding.
  • Methodological innovations: Proposals include ethnographic metadata in model cards, field-conditioned evaluation, and AI-assisted ethnography under human supervision.
  • Structural barriers: Ethnographers face exclusion from technical decision-making, stereotypes about their methods, and lack of interdisciplinary training pathways.
• Advantage over baselines:
  • Highlights the inadequacy of Western-centric RAI frameworks in India.
  • Provides actionable insights for integrating ethnographic expertise into AI governance, moving beyond critique to concrete proposals.
• Experiments / evaluation:
  • Workshop with 20 participants from diverse regions and disciplines.
  • Follow-up interviews exploring participants' experiences, critiques, and envisioned roles in RAI.
  • Reflexive thematic analysis to identify key themes and insights.
• Limitations and future work:
  • Limited representation from rural and non-academic ethnographers.
  • Geographic concentration in urban areas; no participants from Southern India.
  • Future research should include longitudinal studies, broader geographic representation, and comparative studies across Global South contexts.

Summary

This study investigates Indian ethnographers' perspectives on Responsible AI, revealing significant gaps in existing frameworks' ability to address India's social complexities. Participants critiqued India's RAI policy for its industry bias and reliance on abstract ethical principles, proposing methodological innovations such as ethnographic metadata and field-conditioned evaluation. Structural barriers, including disciplinary gatekeeping and limited technical training, hinder ethnographers' integration into AI governance. The study outlines multi-level strategies for inclusion, emphasizing the need for interdisciplinary training, institutional reforms, and recognition of ethnographic expertise as essential to ethical AI development in India. These findings provide actionable pathways for creating contextually grounded, community-centered RAI frameworks.