The Liberty Model

Rules-as-code triage for undertrial entitlement — not a risk predictor.

Roughly three in four prisoners in India are undertrials. Many already qualify for release under existing law but are lost in a system where prison and court data are never joined up. The Liberty Model does not predict who deserves liberty. It checks, deterministically, who the law has already granted it to — and cites exactly where.

🔗 **Live demo:https://poonam160498.github.io/Liberty-Model-Algorithmic-Triage-for-Bail-Applications/

Why rules, not machine learning

Predictive bail models estimate reoffending or flight risk from historical data — bias in, bias out, and opaque at the point where it matters most. This project takes a different, narrower approach: every check answers a factual, verifiable question ("has this person served the custody threshold the law itself set?"), not a predictive one. See methodology.html for the full rulebook and citations (BNSS s.479, Satender Kumar Antil v CBI, Hussainara Khatoon v State of Bihar, and others).

What's real and what's synthetic (read this first)

• Dashboard case data: 120 synthetic cases, generated to resemble realistic patterns. No real individual's data is shown.
• Homepage statistics: real aggregate figures from NCRB Prison Statistics / NJDG, used for context only — not linked to any case in the dashboard.
• Data extraction pipeline (/pipeline scripts, not connected to the live site): a proof of concept for pulling anonymised case metadata from public eCourts/NJDG pages, built to test feasibility for a future, properly-reviewed real-data version.

Full caveats and what would need to happen before any real-world use are on about.html.

Tech

Static HTML/CSS/JS. js/engine.js is a deterministic rules engine — no ML, no learned weights. cases.json holds the synthetic case set consumed by the dashboard.

Author

Poonam Nauroji — LLM (Criminal Law) · MSc Business Analytics, University of Exeter. Part of a research thread that also includes an analysis of UK asylum appeal outcomes and an LLM dissertation on unequal application of criminal law in India.

Disclaimer

This is a research and portfolio project, not a legal service. Nothing here constitutes legal advice or an automated legal decision.