Arivu Mitra: Building a Motivation-Driven WhatsApp System for Rural Libraries as Community Knowledge Hubs

The problem

• Karnataka's Gram Panchayat Arivu Kendras are community knowledge centres run by rural librarians.

• Activating a large, distributed network of librarians is hard and the usual norm of top-down monitoring and reporting forms makes people feel surveilled and quietly kills motivation.

Our bet: motivation, not compliance

• We built Arivu Mitra as a companion, not a monitoring tool, which is designed around intrinsic motivation.

• It helps librarians get activity ideas, training, and support, share their work, and feel recognised so reporting becomes a by-product of meaningful work, not a chore.

Why WhatsApp + FOSS

• Librarians are Kannada and mobile-first and wary of new apps, so we built the intervention inside WhatsApp, that they already trust.

• We had a short policy window and couldn't procure a third-party vendor so a working prototype had to win approval. On the self-hosted, open-source Whatomate gateway, we went from idea to piloting with 100 librarians in a week: without any SaaS lock-ins, no per-message bill and full data ownership.

How We Build on Open Source and Give Back.

• Since Whatomate is open source, we not only used it but also extended its codebase. We developed an admin module for metrics, activity, progress and features that closed vendors typically do not provide.

• Being open also meant we could move at our own pace and add what we needed, instead of waiting for a vendor to prioritise it.

• As a result, we are responsible for ongoing maintenance, which motivates us to contribute improvements. For example, we resolved an issue where Kannada titles were truncated due to byte-based character counting. I will submit a pull request to address this, as well as add support for media in bulk campaigns and scheduled sends.

Two pilots, studied and iterated

• Across two pilots (≈100 librarians each), we studied real behaviour, ran small experiments, and turned findings into concrete design decisions.

 What users taught us:

• Librarians treated the bot like a WhatsApp group chatting freely, rarely tapping buttons, describing problems in their own words.

• This drove us to an LLM intent engine for free-form Kannada and English, voice-message input in their language, and since they don't read long texts- voice responses, plus rewards and a community newsletter for recognition.

Open by design

• We're open-sourcing our findings as a living knowledge base for anyone who wants to use or contribute.

• How a 3-person team of one engineer, two designers, shipped an inclusive, vernacular tool, fast, by using FOSS for everything.

• Designing for intrinsic motivation, specifically in rural contexts requires extensive iteration, since both interaction design and conversational UX evolve through continuous testing. Using Whatomate allowed rapid prototyping and quick implementation of learnings from user feedback.

• Inside Government institutions, the fastest way to win approval isn't a proposal, it's a working product in the decision maker’s hands which open source allows us to do. Self-hosting open source was also crucial to keep the data with the people it belongs to and not on an outside company's servers.

• Open source has enabled us to not just be users, but also to contribute. We are giving back our code fixes and our learnings, so others can build on both instead of starting over.