You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
[Principia M5] Add approximation and empirical-premise governance #23
Make the boundary between mathematical proof, physical modeling, approximation, and observation explicit in packet structure, policy, and release reporting.
Required work
Define packet conventions for physical/model assumptions that carry no proof authority.
Define approximation-role packets with explicit domains, limiting regimes, regularity conditions, asymptotic variables, and error terms.
Define empirical-premise packets with source/data/version provenance, measurement assumptions, uncertainty, and no proof authority.
Add relation-policy checks so empirical_premise targets cannot be proof-bearing packets.
Ensure a kernel-verified analytical implication may depend on a non-verified empirical premise without promoting the combined physical claim to kernel_verified.
Add coverage/report fields showing which source claims have:
mathematical analytical cores;
explicit physical assumptions;
approximation justifications;
empirical support;
unresolved model gaps.
Add adversarial fixtures for authority leakage, hidden approximation assumptions, missing error terms, and unverifiable empirical provenance.
Demonstration requirement
Using the model-boundary slice from #22, publish a complete dependency chain that separately displays:
reviewed historical claim
→ modern physical model assumptions
→ conditional mathematical theorem
→ approximation theorem or explicit wall
→ empirical premise/data record
→ combined interpretation with bounded authority
The final interpretation must state exactly which components are kernel verified and which are not.
Acceptance criteria
Mathematical proof authority cannot propagate into model assumptions, approximations, observations, or the source wording.
Approximation claims specify an actual domain and error/limit statement, or are classified as unresolved.
Empirical premises have immutable source/version provenance and no proof authority.
Public reports visibly distinguish verified implications from physically asserted premises.
The model-boundary demonstration passes all policy, export, and replay gates.
Activity