During an emergency, the information flowing through social media is both your greatest asset and your most dangerous liability. TruthStream turns the noise into signal — in real time, at scale, with a human commander always in the loop.
Within minutes of any emergency, thousands of posts flood social platforms. Some are accurate. Many are not. All of them spread. Inaccurate posts misdirect responders, trigger unnecessary evacuations, and overwhelm 911 with phantom reports — at exactly the moment your resources are most constrained.
The system runs continuously from the moment an incident is declared. Every step is logged, auditable, and under Incident Commander authority.
TruthStream scans social media platforms continuously — Twitter/X, Facebook, Nextdoor, local news comment sections — filtered by geographic radius around the incident and emergency-specific keywords. Every relevant post is captured in real time and queued for verification.
Each post is evaluated against a continuously updated RAG — a real-time knowledge base drawing from authoritative federal, state, and field sources. The AI determines whether the claim is accurate, inaccurate, or unverifiable, and assigns a confidence score. When data is stale or ambiguous, the system flags the Incident Commander immediately rather than responding with false confidence.
If a post is inaccurate, TruthStream drafts a correction — not a generic broadcast, but a response calibrated to the demographic profile of the poster. Age, apparent technical literacy, emotional register, and platform all inform the tone and language. A frightened parent on Nextdoor receives a different response than a journalist on Twitter. Each draft is presented to the Incident Commander with its confidence score and source citations before any distribution occurs.
The Incident Commander reviews the draft and confidence score. One button distributes the verified response across all relevant platforms simultaneously. High-confidence corrections can be pre-authorized for auto-distribution. Every response is logged, timestamped, and linked to the RAG data that supported it — a complete audit trail for after-action review. The RAG is updated immediately with any new information the IC provides — keeping ground truth current as the incident evolves.
Three posts. Three very different situations. TruthStream handles each in seconds — and puts the IC in command of every response.
The verification engine is only as good as its data. TruthStream draws from the broadest possible set of authoritative sources — federal, state, and field — and updates the RAG continuously as the incident evolves.
Live weather observations, storm tracks, flood stage predictions, wind speed and direction — updated every 5 minutes.
Shelter registry, logistics assets, disaster declarations, resource availability — authoritative federal emergency logistics.
Earthquake data, landslide risk, stream gauge readings, flood inundation maps — critical for ground-truthing terrain claims.
Road closures, bridge status, traffic conditions, evacuation route availability — real-time across NC.
Incident commander updates, field officer reports, perimeter maps — the most current ground truth in the RAG.
Live perimeter mapping from aerial assets, updated on each flyover cycle. The RAG flags when perimeter data is stale.
Air quality, stream levels, wind sensors — automated feeds that update the RAG without human intervention.
Every RAG entry carries a timestamp. When data ages past a defined threshold, the system flags the IC rather than responding with potentially outdated information.
Every verification produces a confidence score. That score determines the response pathway — not the response itself. The IC always decides. The system always shows its work.
Multiple authoritative sources agree. Data is current. Draft response is presented to IC with recommendation to distribute. Auto-distribution available if pre-authorized.
Partial verification. Some sources agree, some are silent or stale. Draft response presented to IC with source gaps clearly marked. "Working on it" response recommended.
Cannot verify from available data. IC is alerted immediately. Field unit requested. No auto-response generated. "We're checking this now" is surfaced as safe default.
Below is a conceptual TruthStream operational map during a wildfire in West Asheville. Every asset has a live GPS position. Every social media post is pinned to its origin. Predictive fire modeling shows projected perimeter growth based on current wind, terrain, and fuel load — updated as conditions change. The IC sees ground truth and public narrative simultaneously, from a single screen.
* Positions and perimeters are simulated for conceptual demonstration. In deployment: asset positions feed from agency CAD/GPS systems in real time; predictive spread integrates FARSITE/Phoenix fire behavior modeling with NOAA wind forecasts and USGS terrain/fuel data; weather overlay feeds from NWS API. Social media locations approximated from device geolocation or profile data where available.
TruthStream is not an autonomous system. It is an intelligence amplifier. The Incident Commander remains the authority for every public communication — the system simply makes it possible to exercise that authority at the speed and scale that modern emergencies demand.
"The measure of this system is not how often it responds automatically. It is how clearly it presents the situation so the Incident Commander can make the right call — every time."
Every draft is presented with its confidence score, its source citations, and a clear indication of what is known, what is uncertain, and what requires field verification. The IC is never asked to distribute something they haven't reviewed. The audit trail is complete — every response linked to the data that supported it, every IC decision timestamped and logged.
TruthStream is a concept developed by the Full Measures Foundation in collaboration with NC Emergency Management. The system is designed to be deployed rapidly during declared emergencies and stood down cleanly when the incident closes.