The Luminous Prosperity Model Limitations Disclosure: Instrument 05 of the Luminous Governance Framework

Every other instrument in the Luminous Governance Framework specifies what an AI system will do. This one specifies what it cannot do, may not attempt, and must never say. It is a negative specification — a governance instrument defined entirely by its prohibitions — and it operationalises the framework's fourth pillar, epistemic humility. The governing argument is stated plainly: a system that cannot say what it does not know will be believed about things it does not know. Silence about limitation is not neutrality but a claim, and users fill that silence generously because machine output arrives fluent, immediate, and confident. This instrument therefore publishes the incapacity of the system in the same register, with the same specificity, and in the same places as its capability. Section II states the epistemic position honestly: a developmental inference is a probabilistic pattern-match over language and behaviour, produced by models trained on culturally situated data, evaluated against developmental frameworks that are themselves contested instruments. Section III enumerates ten classes of thing the system cannot know, including actual developmental stage as distinct from a correlated signal, interior experience and felt sense, intention, the truth of self-report, trauma history, clinical status, future trajectory, the valence of change, spiritual condition, and a person's worth or ceiling. Section IV enumerates eight categorically prohibited inference classes — clinical diagnosis, suicide and self-harm risk scoring, protected-characteristic inference, political affiliation, criminality propensity, employability and creditworthiness and insurability, parental and relational fitness — governed by a general test: any inference whose primary use would be to make a decision about a person rather than to offer something to them is forbidden. Section V prohibits six classes of speech act and supplies a binding required-language table of fifteen prohibited/required pairs written to be usable directly by engineers and copywriters. Section VI specifies the mandatory user-facing uncertainty affordance. Section VII is an honest register of seven known failure modes with mitigations, marked where mitigation is only partial. Section VIII states without softening that stage-model measurement validity is disputed in the peer-reviewed literature, that Spiral Dynamics has essentially no peer-reviewed validation record, and that the system's constraints are designed on the assumption that its own instruments may be wrong. Sections IX and X establish disclosure obligations and Governance Circle enforcement, including the operative rule that no inference class enters production before it is added to this register.