Open Research Questions
Research agenda derived from the public stack · not claims already resolved by the papers
Interpretation rule: the questions below identify areas for further research. Their inclusion does not mean the current papers have already solved or empirically validated them.
1. Measuring effective independence
- How should authority domains be counted when they share cloud infrastructure, model providers, credentials, root administrators, or legal control?
- What evidence is sufficient to distinguish organizational separation from causal independence?
2. Dynamic refusal margins
- How should refusal latency be updated when model speed, automation depth, or deployment topology changes?
- How much safety margin is required when detection and verification latency are uncertain rather than fixed?
3. Objective-persistence evidence
- What qualifies as an external anchor when the evaluator shares data, models, infrastructure, or assumptions with the evaluated system?
- Can persistence claims remain meaningful after repeated self-modification without collapsing evaluation into internal self-attestation?
4. Human-revisable development
- Which observable indicators can represent fallback capacity, retained human competence, institutional alternatives, and substitution reversibility?
- How should revisability be evaluated when no single deployment is irreversible but cumulative dependency is?
5. Cross-layer failure
- How should a safety case represent a system that passes PCR-C-style infrastructure checks but fails L30-RET because the human decision process is too slow?
- How should a system be evaluated when refusal remains effective locally while PISF-style authority concentration or societal dependency erodes wider revisability?
6. Empirical program
- Which real-world governance or infrastructure systems are suitable for non-certifying case studies?
- How reproducible are RET gate assessments across independent reviewers?
- Which false-positive and false-negative patterns appear when applying these concepts to operational systems?
Research discipline
The public stack should distinguish conceptual definitions, toy/reference cases, observational studies, and empirical validation. It should also preserve the separation between research evidence and the LUMINA-30 canonical boundary framework.