AquaForensicsIEEE OneAquaHealth
OneAquaHealth IEEE Global Hackathon 2026

Evidence-Guided
Environmental Reasoning

“AquaForensics doesn’t ask AI to decide what happened.
It asks what evidence we need to find out.”

Turn citizen science observations of urban water anomalies into structured investigations, competing latent hypotheses (Reach-wide vs Local vs Transient), uncertainty estimates, and mathematically computed Next-Best-Evidence (EIG) recommendations.

DEMO / SYNTHETIC DATA3–5 Minute Hackathon Walkthrough

Deterministic Investigation Scenario

Explore a synthetic AquaForensics investigation from first observation to updated evidence.

Traditional Citizen Science
OBSERVE→RECORD→SUBMIT

Static single-point capture without active investigative guidance.

AquaForensics Reasoning Loop
Observe→Hypotheses (W/L/N)→Compute EIG→Next Evidence

Active entropy reduction guiding citizen observers where to look next.

Scientifically Honest Environmental Reasoning

EXPERT_ELICITED_UNCALIBRATED

"AquaForensics does not ask AI to decide what happened. It asks what evidence we need to find out."

All numeric model scores represent assumption-sensitivity model support under expert-elicited engineering parameters (EXPERT_ELICITED_UNCALIBRATED), NOT calibrated real-world probabilities. Investigation states represent candidate hypotheses for triage, not confirmed causes or diagnoses.

Initialize New Investigation

Create evidence-guided investigation container

Bayesian v0.1
DEMO / SYNTHETIC
Primary OAH citizen science trigger
90%
Reference reliability baseline

Recent Investigations

Loading cases...