<aside> ℹ️ This presentation features at PAIRS 2026 in New Delhi on 18th February 2026 at 14:00 IST

🎙️ Thread on PAIRS Discussion Server (Discord) (register first)

</aside>

Abstract

The expanding penetration of artificial intelligence (AI) into governance, communication, and livelihoods underscores an urgent need to reconfigure development paradigms around inclusion and equity, particularly at the intersections of technology, social justice, and development. Prevailing models of AI innovation are predominantly techno-centric and institutionally concentrated, often marginalising those most affected by algorithmic systems. This paper, drawing on empirical insights from the Digital Empowerment Foundation (DEF), advances a praxis-oriented framework for Participatory AI Development grounded in long-term community engagement within rural India.

Since 2002, DEF has facilitated digital access, literacy and entrepreneurship in India's last mile with more than 2,400 rural women and persons with disabilities entrepreneurs across 26 states and 250+ districts, reaching over 10,000 villages and serving more than 35 million citizens. Built on the 'SoochnaPreneur' ("information entrepreneurs") model, the recent initiatives such as 'Just AI: Data and Alogorithms for Communities and 'Smartpur' (shifting the focus to building self-sustainable smart villages) extend this mandate into the AI domain, foregrounding participatory methodologies that integrate community epistemologies into the design, deployment, and evaluation of AI-enabled systems. The research adopts a mixed qualitative approach comprising participatory action research and community ethnography to examine how marginalised actors articulate their technological aspirations and negotiate algorithmic interventions.

Findings indicate that participatory AI development, when situated in socio-economically diverse contexts, yields three interlinked outcomes. First, it reconfigures knowledge hierarchies by positioning local actors, particularly women, youth, and informal workers as epistemic contributors rather than data subjects. Second, it enhances procedural trust in AI systems through transparency, vernacular communication, and collective data stewardship. Third, it generates situated innovation, enabling the co-creation of AI tools responsive to local linguistic, cultural, and ethical conditions. Illustrative examples include community-curated voice datasets for underrepresented languages, participatory content moderation frameworks, and AI-assisted livelihood platforms co-designed with women entrepreneurs.

However, the analysis also surfaces structural constraints: sustaining participation within resource-limited ecosystems, mediating tensions between expert and experiential knowledges, and the persistent institutionalisation gap between grassroots participation and national AI governance frameworks. Addressing these challenges requires a shift from episodic consultation toward an infrastructural model of participation, where inclusivity becomes integral to AI research, funding, and policy design.

The paper situates DEF’s praxis within broader debates on localising AI and epistemic justice, contending that participatory AI in the Global South is not merely a methodological preference but an ethical and developmental necessity. Through field-level engagement into a replicable participatory framework, DEF’s experience contributes to the global discourse on equitable AI ecosystems, one that recognises communities as active co-authors of technological futures rather than passive beneficiaries of innovation.