Section 111 of 440
APPENDIX AA: CANDIDATE LAB AND RESEARCH OBSERVATORY
Lab discovery, admission and promotion pipeline
The Titan Research & Governance Lab gives HCE and Titan one controlled route for preserving observations, admitting research, defining candidates, authorising experiments and considering operational promotion without converting novelty into proof. Research stage, workflow state, evidence currentness, execution state, risk route and deployment role remain separate. A topic may be scientifically interesting while operationally unscored; an operational module may later be suspended while its historical evidence remains intact.
State | Meaning | What is allowed |
|---|---|---|
Observation | A phenomenon, mechanism or report worth tracking. | Research notes only; no Titan level, protocol claim or implication of general trainability. |
Candidate | A bounded proposed capability with a plausible task, outcome and safety route. | Prototype definitions and measures; clearly labelled unscored. |
Experimental | A governed pilot with predeclared criteria and adverse-event monitoring. | Controlled testing in an appropriate setting; no routine public promotion. |
Operational | An explicitly promoted module or exemplar with versioned claim, assessment and safety contracts. | Declared assessment under the Capability Standard. |
Maintained | Operational evidence and governance remain current under the exact comparison rule. | Current use and retesting inside the declared review window. |
Suspended | New evidence, safety concerns, data defects or governance failure block current use. | Historical records remain visible; no new operational claims. |
Retired / rejected | The protocol or claim no longer earns its burden, lacks validity or is superseded. | Preserve the decision and rationale; retirement is a valid positive outcome. |
Promotion is a hash-bound human decision followed by a separately verified registry mutation, not a formatting change, archive import or AI recommendation. Before a Candidate becomes Operational, Titan requires a bounded capability definition, explicit system of interest, objective direct outcome, predeclared criterion, delayed retention plan, meaningful novel transfer test, burden measure, stop rule, accessibility route, professional and jurisdictional gates, evidence review, data-field validation, migration plan, rollback route and proof that the measure cannot be passed by lowering quality or increasing risk.
The evidence must support the population and configuration actually claimed. A result in experts does not automatically generalise to novices. A finding in blind participants does not automatically establish the same practical value in sighted participants. A laboratory device does not become a consumer capability until equipment, privacy, cognitive load, transfer, device-off retention and long-term safety have been addressed.
Rejection, suspension and retirement are governed outcomes rather than failures of the discovery process. They show that the system can distinguish an interesting idea from a useful, safe and supportable capability. The Lab master register preserves Candidate and Observatory identity, versions, contrary evidence, decisions and rollback history so later evidence can reopen a question without rewriting the past.