Lightweight Accountability

Inspiration

The inspiration for this project came directly from the persistent frustration experienced in our residential college. We noticed that shared spaces frequently suffered from an accountability vacuum; residents would leave messes behind, causing a decline in overall living standards and sparking toxic complaints in floor group chats. We realized that traditional analog solutions, like chat reminders and cleaning rosters, consistently fail because they rely on voluntary compliance rather than actionable accountability. We wanted to solve this "tragedy of the commons" using technology, but without creating an invasive surveillance state.

What it does

Lightweight Accountability is a privacy-first system that introduces indirect accountability in shared spaces. It detects when issues are reported and provides anonymized, time-based visual summaries of who accessed the space—without identifying individuals.

How we built it

We designed a localized, low-cost system architecture centered heavily around privacy.

Cameras are placed strictly at the entrances of shared spaces, ensuring they never monitor the interior of toilets or pantries.
Cameras are motion-triggered to capture timestamped footage.
An AI extraction engine is deployed on an edge device (e.g., Raspberry Pi) to process video locally.
The computer vision model extracts only non-sensitive descriptors (e.g., clothing color, hair length, accessories), strictly avoiding facial recognition or identity databases.
A Telegram bot serves as the user interface, listening for trigger keywords like "dirty" or "not flushed" in student floor group chats.
When triggered, the bot retrieves anonymous visual logs from a temporary database and outputs them using emojis for easy visualization.

Challenges we ran into

Our biggest challenge was balancing the need for accountability with strict administrative compliance and user privacy. We had to ensure that the system could not be used for malicious tracking.

To address this:

We implemented edge processing so that raw video never leaves the local network.
We enforced a strict data retention policy where all temporary logs are automatically deleted after 24 hours.

Accomplishments that we're proud of

Privacy-First Design

We successfully built a monitoring system that provides actionable data without ever using facial recognition or identity tracking.

Edge Processing Implementation

By running our AI extraction engine locally on edge devices like a Raspberry Pi, we ensured that raw video feed never leaves the local network, significantly lowering the risk of data breaches.

Seamless Community Integration

We are proud of the Telegram bot interface, which transforms a complex backend into a simple, emoji-based visualization that residents can use without learning new software.

Social Engineering Impact

We shifted the focus from punitive "policing" to community-driven responsibility, proving that ambient observation is enough to elevate living standards.

What we learned

We learned that effective community management doesn't require absolute certainty. We don't need to know exactly who made the mess.

We discovered that simply letting the community know that shared spaces are passively observed is enough to create subtle, positive social pressure. By narrowing down responsibility indirectly, we can shift the culture from high-friction direct confrontation to low-friction, community-driven responsibility.

What's next for Lightweight Accountability

Community Feedback Gamification

We plan to introduce a community rating system and cleanliness feedback loops to further gamify and encourage positive behavior.

Multi-Facility Scaling

While currently designed for residential floors, we aim to scale the hardware footprint to be easily deployable across larger facilities like gyms or study lounges.

Dynamic Time-Window Refinement

We are working on refining the bot's logic to automatically adjust its query window based on the frequency and severity of reported issues.

Enhanced Consent Protocols

We will continue developing clear communication and consent frameworks to ensure residents are fully informed and comfortable with the system prior to any long-term deployment.