What authorises knowledge?
Qualify claims, sources, access, dependence, context, contestability and responsibility.
MOSES EPISTEMIC–COSMOGONIC FOUNDATION · VERSION 0.14
For millennia humanity has asked where the world comes from and how anything about the world can be known. Moses studies both histories as extreme laboratories of knowledge, inference and speculation about the inaccessible.
Epistemology qualifies the right to know. Cosmogony tests how intelligence crosses beyond what it can observe. Neither supplies Moses with inherited dogma.
Epistemology studies observation, testimony, memory, inference, practice, authority and proof. Cosmogony builds bridges from present traces to absent origins through analogy, genealogy, symbol, model, retrodiction, cyclicity, first causes and transformation. Their joint value lies in the operations they reveal — and in the errors they make visible.
Qualify claims, sources, access, dependence, context, contestability and responsibility.
Expose bridges from trace to cause, present to past, model to world and origin to authority.
They compare architectures without ranking civilisations and without granting equal evidential weight to every proposition. Presence and critique are universal rights; epistemic weight remains claim-specific.
The bridge is indispensable. It becomes dangerous when its joints disappear: analogy becomes identity, a model becomes the world, genealogy becomes nature, or a plausible history becomes the only possible history.
It happened and may have left traces. Several histories can produce the same present evidence.
It may exist without direct access. The name or model must not be confused with the thing.
It has not happened and has no traces of its own. Probability is not possession of the future.
A single confidence score hides why a system does not know. Each kind of ignorance requires a different response.
Naming a gap makes it available for inquiry; it does not explain it. Chaos, Nun, Te Kore, dark matter or “risk 0.82” must remain distinct from what the label actually authorises us to conclude.
M-PTL extracts a function, identifies its possible pathology, builds a control, and then submits that control to real cases, adversarial dialectic and the Constitutional Gate.
Use only the minimum depth the problem requires.
Map producers, custodians, verifiers and beneficiaries.
Expose models, providers, lineage, retrieval, tools, instructions and evaluators.
Show who chose each threshold, why, and who bears its errors.
Ask whether a prediction or classification changes what it describes.
Compare irreversible harms of action, omission and the status quo.
Separate the right to object from thresholds for reopening or suspending.
An origin never justifies present legitimacy by itself.
A conceivable catastrophe still requires a supported causal chain.
Separate genealogical continuity from present reasons for authority.
Use cosmogonic analysis only where origins, invisible pasts, analogy or genealogy matter.
MOSES EVIDENTIAL BURDEN PRINCIPLEThe burden of justification rises with inferential depth, the number of people affected, the severity and irreversibility of plausible harms, the difficulty of correction and dependence on invisible bridges — while also counting the harms of waiting and omission.
The synthesis separates comparative findings from cognitive patterns, pathologies, controls, universalist candidates and questions that must remain visible.
Plural origin-forms; no universal first-cause grammar; recurring retrodiction; repeated origin-to-authority moves.
Separation, ordering, analogy, genealogy, symbolisation, named unknowns, cycles, beginninglessness and prospection.
Invented first causes, genealogical naturalisation, model reification, dependent consensus, predictive determinism and catastrophic veto.
Unknown Vector, Alternative Histories, Evidence Ecology, Applicability Gate, reflexivity, bilateral irreversibility and parameter registries.
Do not hide ignorance, confuse repetition with independence, derive authority from origins or treat probable futures as possessed.
Gate governance, source independence, irreducible moral conflict, restricted knowledge, community standing, action thresholds and AI status.
A persuasive pattern may be accidental, copied across dependent sources, valid only in its sample, attached to the wrong referent or narrated beyond the evidence. “AI pareidolia” is used here only as a critical metaphor, not as a diagnosis or a claim that machines perceive like people.
State whether the pattern asserts similarity, identity, cause, prediction or norm.
Separate independent observations from copies, correlated sources and system-produced data.
Remove superficial attributes, vary irrelevant conditions and seek counterexamples.
Use independent, realistic cases or checks suited to the proposition and its referent.
Expose false positives, false negatives, affected people and the cost of delay.
Look for drift and unexpected consequences; yesterday’s success is not a permanent licence.
THREE ADMISSIBLE OUTCOMESCorroborated for a bounded purpose · rejected · insufficient evidence. A failed control cannot disappear inside an average score. Abstention and human escalation remain visible decisions with costs of their own.
M-CCM v0.1–v0.5, M-ATT v0.1 and M-HCS v0.1 remain research materials: historically situated, open to specialist correction, sensitive to community custodianship and incapable of transferring a tradition directly into constitutional authority.
Cosmogonic structures, taxonomies of the unknown, beginninglessness, origin-to-authority, retrodiction and convergences.
Cases where incomplete knowledge meets irreversible medicine, prediction, welfare, climate, finance and AI deployment.
Complexity inflation, pluralism laundering, synthetic independence, provenance capture, catastrophic veto and contestability regress.