This is the single most important boundary in the product. Internally it is tracked as guardrail #563, but you only need to remember one sentence: causation is the licensed clinician's opinion, full stop.
CareHub Intelligence must never form, suggest, estimate, or apportion a causation opinion. It never determines industrial versus non-industrial. It never analyzes the good-faith-personnel-action question, the Rolda analysis, or the §3208.3 threshold. It never opines on compensability.
Take Sofia Reyes. CareHub Intelligence may gather the evidence — her account of the collision, her screener scores, the diagnosis on the record. It may never tell you the injury is industrial, never propose "work 70%, pre-existing 10%," never weigh the personnel-action question. That apportionment is your opinion to form. Crossing this line is the one failure the product is built to make impossible.
The guardrail is enforced in three layers, so it does not depend on anyone remembering it.
- Prompt preamble: every intelligence request carries an injected preamble that enumerates the prohibited outputs, so causation is off the table before any work begins.
- Surface removal: the route that could once suggest causation was removed, so causation is authored manually only — there is no path for the platform to propose it.
- Output scanner: after every response, a scanner checks the output and blocks known causation-leakage patterns before they ever reach you.
Belt, suspenders, and a second belt. Even if one layer were bypassed, the next would hold.
And on every causation surface — like the one pictured — a standing disclaimer makes the boundary visible to anyone looking at the screen:
Causation is determined by the clinician. CareHub Intelligence organizes evidence but does not suggest causation.
That sentence is not a legal footnote to scroll past. It is the contract. The platform hands you an organized, cited pile of evidence; you read it, weigh it, and form the determination that only your license permits.
For Sofia Reyes, which of these is CareHub Intelligence allowed to do?