Formalization
Define the candidate property, terminology, claims, reference artifacts, and reproducible object of study.
A stable question that can be tested.Research hypothesis under active investigation
What assurance survives software change?
Assurance Continuity studies a problem underneath modern software delivery: whether a claim that was justified before a transformation is still justified after the system, evidence, dependencies, providers, or operating conditions change.
Why this research exists
Software changes faster than assurance relationships are reconsidered.
Tests, approvals, reviews, provenance, and operational evidence can all remain present after the relationship between those artifacts and the current system has changed.
Assurance Continuity asks when that relationship still holds, when it does not, and what evidence is necessary to tell the difference.
Four detachment modes
Where assurance can stop following the software.
Each condition expands independently. No item is preselected, and each describes a distinct failure in attribution.
Plan and execution diverge
The approved plan may describe one transformation while the system, provider, operator, or automation performs another.
Research question: Can the observed transformation be reconstructed precisely enough to determine whether prior assurance still applies?
Providers partially succeed
An external provider can report success while organizational state remains incomplete, delayed, contradictory, or split across systems.
Research question: Which state is authoritative, and what evidence is required before assurance can follow the reported outcome?
Evidence becomes stale
A correct test, review, approval, or provenance record can stop supporting the current decision after code, dependencies, conditions, or scope change.
Research question: What exact change invalidates the relationship between the evidence and the claim?
Rollback restores only part of the state
Code can return to an earlier version while authority, evidence, service state, data, provider effects, and continuity remain unresolved.
Research question: What must be restored or re-established before previous assurance can be attributed again?
Research progression
Formalize. Test. Measure. Replicate.
The program advances only when the preceding stage produces enough evidence to justify the next. A later stage is not presumed from the existence of an earlier one.
Protocol validation
Use independent fixtures, separated roles, preregistration, controlled allocation, and adjudication.
Evidence that the protocol is coherent and usable.Empirical evaluation
Measure agreement, reconstruction, false acceptance, false rejection, latency, burden, and domain sensitivity.
Effect and cost under controlled conditions.Independent replication
Support implementation and interpretation by organizations not involved in model authorship.
Portability findings, contradictions, and replication evidence.Measurement
Success is not the only outcome worth preserving.
Research integrity
The hypothesis must be allowed to lose.
Precommit the test
Outcomes, stopping rules, allocation, and analysis are fixed before exposure.
Separate material roles
Fixture authors, operators, evaluators, custodians, and adjudicators remain distinct where independence matters.
Preserve the complete record
Negative results, contradictions, deviations, corrections, withdrawals, and burdens remain explicit.
Keep research separate from commercial claims
Research participation does not certify customers, validate SAM, prove Causance outcomes, or convert commercial work into evidence without separate consent and authority.