Adaptive interview
Questions are selected from the current brief, structured answers, research findings, contradictions, missing authority, affected populations, tradeoffs, funding assumptions, outcomes, and implementation responsibility.
The methodology is designed to reduce busywork without hiding uncertainty, opposition, source quality, legal limits, funding conditions, or human responsibility.
Questions are selected from the current brief, structured answers, research findings, contradictions, missing authority, affected populations, tradeoffs, funding assumptions, outcomes, and implementation responsibility.
Need, evidence, comparable policy, authority, funding, and constraint workflows create source records with publisher, date, retrieval, jurisdiction, limitations, snapshot, and verification state.
The system actively looks for conflicting evidence, failed examples, objections, unintended effects, legal and fiscal vulnerabilities, disparate impacts, privacy risks, and alternative designs.
Plain-language and official clauses map to each other and to claims, sources, funding, constraints, amendments, and human decisions.
AI workflows require thresholds for citation accuracy, source existence, claim support, jurisdiction and funding accuracy, uncertainty, fidelity, malicious-document resistance, and sensitive-data leakage.
Labels describe the state of the record. They do not confer legal authority, certification, or government endorsement.
A research artifact is not a verified fact until its sources and claims pass review.
Expertise is narrow and does not override affected participants or government authority.
Substantive policy choices preserve the actor, context, rationale, and version consequence.
The production domain publishes the reviewed public experience. Private AI, database, and worker services remain inactive until their deployment and provider gates pass.
Open the fictional evidence workspace to inspect supporting evidence, counterevidence, source quality, limitations, and human-review states.