Title: A Human-Centered AI Framework for Multilingual Rural Field Reporting, Qualitative Insight, and Community Support (Sruthi · Sakhi · Viveka AI · AI-Facilitator)
Core Research Question: How can human-centered AI systems be designed to support multilingual field reporting, qualitative interpretation, and facilitator action in low-resource community settings — without weakening trust, accessibility, or local voice?
Description
Most AI is built for scale, speed, and engagement. This project asks a different question: what happens when AI is built for listening — to women’s collectives, to field practitioners working in multilingual communities, to voices that standard data systems routinely miss? The portfolio — Sruthi, Sakhi, Viveka AI, and an AI-Facilitator — brings together conversational AI and multilingual speech tools that help community support field teams, women’s collectives, and researchers capture, interpret, and act on community knowledge across languages and literacy levels.
Each tool in the portfolio is designed around what real practitioners and women’s collectives actually need: Sruthi, a peer-mode reflective companion for field practitioners; Sakhi, a planned support tool for Gender Point Persons navigating gender-based violence; Viveka AI, a multilingual pipeline that turns field audio into usable insight; and an AI-Facilitator that supports inclusive, real-time Participatory Rural Appraisal.Grounded in the AWESOME framework for women’s empowerment ( https://doi.org/10.1016/j.wdp.2020.100245 ), the work takes seriously what mainstream AI often sidesteps: that in low-resource, high-trust community settings, an AI that maximises engagement, flatters users, or harvests data without consent does not help — it harms. Every design decision is therefore grounded in the principles of transparency, human oversight, and community accountability, aligned with UNESCO’s Recommendation on the Ethics of AI (2021) and NITI Aayog’s Responsible AI for All framework. Governance, consent, and contestability are built in from the first sprint, not bolted on at the end.
Tagline: AI that listens first — built with women’s collectives, for the realities of multilingual, low-literacy community life.settings.
Project Type: Combination: Applied Research + Product Development + Pilot Implementation + Evaluation. Includes participatory co-design, naturalistic field deployment, and longitudinal implementation-science evaluation.
Launch: Initiated 2023 (Sruthi field pilot 2024; Viveka AI pipeline and ICGT 2025 publication; ongoing portfolio integration through 2026).
Domain: Context-Aware AI for Social Good, Women’s Empowerment, Implementation Science in Community Support, Participatory AI, Human-Centered Computing
Audience: Frontline community workers; field practitioners and emerging social workers; program coordinators, supervisors, and researchers; tech enthusiasts motivated to make real social impact, institutional decision-makers in rural development, women in Self-Help Groups (SHGs) or cooperatives and Gender Point Persons (GPPs).
Technologies
Principal Investigator: Dr. Srividya Sheshadri
Co-Principal Investigator: Dr. Bhavani Rao
Funding Agency: Internally Funded Project
Duration: 2024 – Ongoing
Official Project Logo
Human-Centered AI Framework for Multilingual Community Development & Support designed for ethical, multilingual, and community-centered AI engagement.
What We Mean by Context-Aware AI
Building AI systems that prioritise trust, inclusion, dignity, and meaningful human participation over surveillance and engagement-driven design.
Designing AI systems that prioritize trust, inclusion, human dignity, and meaningful participation over engagement metrics and surveillance.
Most commercial AI is designed to keep users engaged — not to help them think clearly or act well. Sycophancy, surveillance, and the exclusion of people without literacy or dominant-language access are increasingly common outcomes.
In community settings, where trust is fragile and the people most likely to be harmed often have the least recourse, these are serious concerns. This project is built as a deliberate response to those risks.
“Context-aware AI” in this framework means designing systems that refuse to treat people as data sources or engagement metrics. Instead, the AI responds to four interconnected layers of context that shape real-world community interaction and support.
1. Linguistic Context
Supports multilingual, code-switched, and voice-first communication with minimal dependence on formal typing or literacy.
2. Social Context
Recognizes role relationships and trust dynamics across peer groups, supervisors, institutions, and community spaces.
3. Intervention Context
Integrates with real-world workflows including SHG cycles, GBV response systems, PRA activities, and facilitator practices.
4. Ethical Context
Treats consent, ownership, contestability, and community oversight as foundational design requirements from the beginning.
Alignment with Sustainable Development Goals
The project contributes to global sustainability priorities focused on inclusion, education, ethical innovation, well-being, and community-centered development.
Community-Centered Innovation for Global Impact
The framework aligns with key United Nations Sustainable Development Goals by strengthening ethical AI, women’s empowerment, multilingual accessibility, inclusive participation, and collaborative social innovation ecosystems.
Understanding the real-world challenges that inspired the development of ethical, multilingual, and community-centered AI systems.
Preserving Community Knowledge Before It Disappears
Picture a community field worker at the end of a long day — she has just witnessed something deeply important: a moment of trust inside a Self-Help Group meeting, a woman cautiously speaking about safety, or an aspiration that took months to surface.
She opens WhatsApp and types a few lines. No one responds. Her supervisor is managing dozens of similar reports. The formal reporting system asks for structured data but leaves no space for the emotional and human realities she observed.
By the next morning, the moment is gone.
This project responds to the challenge of preserving and interpreting rich community knowledge that is often lost across fragmented reporting systems, informal conversations, and disconnected voice-based interactions.
Building Ethical AI Systems for Community-Centered Action
Interdisciplinary leadership bringing together AI innovation, gender research, implementation science, community engagement, and rural development expertise.
Lead Institution
Center for Women’s Empowerment & Gender Equality (CWEGE)
Amrita Vishwa Vidyapeetham,
Amritapuri Campus, Kerala, India
Collaborating Partners
Team & Project Research Lead
Project Coordinator
Senior Advisor
Rural Technologies Lead
AI & Data Systems
Applied AI & XR
Evaluating how multilingual and community-centered AI systems support participation, reflective learning, qualitative insight, and ethical field engagement.
Community Participation Across Rural Contexts
Reflective Learning & AI Evaluation Insights
Beyond AI Assistance Toward Long-Term Governance
The findings suggest that reflective peer-mode AI systems can strengthen collaborative communication and qualitative understanding within community environments. The next phase focuses on governance-aware integration, long-term reflective ecosystems, and ethical field-to-insight AI deployment models.
Voices from the Field
International Conference on Gender & Technology (ICGT) 2025
Moving from pilot systems toward long-term multilingual, governance-aware deployment ecosystems embedded within real community practice.
Building Governance-Aware AI Ecosystems for Community Practice
The next phase aims to operate Sruthi, Sakhi, Viveka AI, and the AI-Facilitator as a unified field-to-insight ecosystem embedded within live SHG program cycles and governed collaboratively with the communities it serves.
Rather than treating ethical concerns as secondary risks, the framework approaches transparency, accountability, consent, human oversight, and anti-surveillance safeguards as core architectural principles shaping every stage of deployment.
01. Longitudinal Deployment
Extending the pilot into 6–12 month naturalistic deployment cycles integrated within live SHG and PRA ecosystems.
02. Offline AI-Facilitator
Developing Android-first, offline-capable facilitator tools designed for low-resource multilingual environments.
03. Ethical Governance
Co-developing governance protocols covering consent, ownership, anonymization, contestability, and withdrawal rights.
04. Scalable Replication
Adapting reflective AI systems across community health, education, GBV support, PRA activities, and rural facilitation ecosystems.
Peer Framing Matters
Non-evaluative peer-mode AI changes what becomes “sayable” in sensitive group environments.
Adaptive Pacing is Essential
Reflection deepens through persistence, but ethical systems must also prevent fatigue and interaction overload.
Multilingual Access
Code-switching and multilingual interaction are essential for genuine inclusion in rural AI ecosystems.
Voice-First Design
Low-literacy environments require voice-first interaction models from the earliest design stages.
Human Oversight
AI interpretation of PRA and community artifacts remains fragile without continuous human supervision.
Embedded Adoption
Tools succeed only when integrated into channels, rhythms, and practices communities already trust.
A university-led interdisciplinary ecosystem integrating multilingual AI systems, participatory methodologies, implementation science, and ethical governance frameworks.
Building Community-Centered Reflective AI Systems
The implementation model combines conversational AI, PRA interpretation systems, multilingual field intelligence, and governance-aware deployment frameworks into an evolving research ecosystem.
Sruthi Reflective AI System
Designed and evaluated a peer-mode conversational AI system through participatory design methods, practitioner reflection, and implementation-science evaluation.
– Workstream 4.1
Viveka AI Multilingual Pipeline
Built multilingual audio-to-insight systems capable of handling code-switched field recordings and qualitative community reflections across low-resource contexts.
– Workstream 4.2
Generative AI PRA Evaluation
Benchmarked leading generative AI systems using multilingual hand-drawn PRA village artifacts to study interpretation reliability and oversight needs.
– Workstream 4.3
International Conference Dissemination
Research findings presented at ICGT 2025 and disseminated through international AI-for-social-good publication channels.
– Workstream 4.4
Longitudinal SHG Deployment
Expanding toward long-term SHG ecosystems integrating reflective conversational systems, PRA interpretation layers, and facilitator-facing dashboards.
– Workstream 4.5
Ethical Community Oversight
Co-developing governance systems covering consent, anonymization, transparency, ownership, and responsible deployment safeguards.
– Workstream 4.6
Documentation, evaluation reports, and scholarly publications emerging from the Sruthi, Sakhi, Viveka AI, and AI-Facilitator research portfolio.
Pilot Completion Report
Internal pilot completion documentation and deployment summary report.
Impact Evaluation Report
Mixed-methods evaluation of Sruthi: Collaborative and Transformative Learning trajectories, post-use perception survey (41 items), and three FGDs.
Contact Information
For academic collaboration, interdisciplinary research partnerships, implementation-science discussions, multilingual AI systems, and community-centered innovation initiatives.
Principal Research Contact
Dr. Srividya Sheshadri,
Associate Director, CWEGE
Academic Email: srividya.sheshadri@ammachilabs.org
Research Center: Center for Women’s Empowerment & Gender Equality (CWEGE)
Research Platform: www.ammachilabs.org
