A research programme that has to remain capable of being wrong.
Decision Epistemology is developed under explicit comparator pressure, bounded claims and conditions under which its constructs should narrow, be absorbed or be retired.
Scientific seriousness requires more than internally coherent terminology. It requires the possibility that stronger evidence or stronger existing explanations can win.
The language does not get to decide.
A compelling term is not evidence.
A repeated pattern is not automatically a mechanism.
A mechanism is not automatically a principle.
A construct is not validated because it is intuitively persuasive.
The programme therefore separates:
- observation
- interpretation
- hypothesis
- construct
- empirical result
- validation
The terminology is provisional.
The evidentiary burden is not.
The programme has to survive stronger explanations.
Decision Epistemology assumes that established disciplines already explain many of the functions relevant to long-lived decisions.
Its distinct contribution therefore has to survive comparison with stronger existing architectures.
- Maintained assurance / Assurance 2.0
- Systems safety / dynamic safety governance
- Situation Awareness / Human Factors
- Organisational sensemaking
- Voice / silence
- Organisational memory / decision provenance
- Management of Change
- Certification / configuration management
- Organisation design / decision rights
- Runtime authorisation / policy enforcement
- Strategic control
- Resilience
- Real options / irreversibility
- Decision analysis / decision quality
If an established architecture performs the same function at equal or lower burden, the new construct should be narrowed, absorbed or retired.
A programme that cannot lose is not functioning as a research programme.
If a construct adds no residual explanatory or analytical value
If another architecture performs the same function with lower burden
If the research question survives but the terminology fails
Preserve the question, not the terminology.
A result should not travel further than its evidence allows.
A successful test does not automatically establish universal validity.
A finding demonstrated under one implementation does not automatically establish cross-domain transfer.
A public discussion does not constitute independent replication.
A manuscript under review is not an accepted result.
A working construct is not a validated construct.
Claim strength should remain bounded by evidence strength.
Public does not mean validated.
Public research notes are used for:
- challenge
- hypothesis generation
- provenance
- public falsifiability
- intellectual-development record
They are not treated as:
- empirical validation
- consensus
- independent replication
Visibility can expose an argument to challenge.
It does not change the evidentiary threshold required to validate it.
Attribution is part of the scientific method.
Ideas, distinctions, results and research contributions retain provenance.
Public discussion does not erase intellectual origin.
Joint work does not become single-author intellectual property by being incorporated into a wider programme.
Independent convergence does not establish common origin, equivalence or validation.
Public does not mean ownerless.
Agreement is not validation.
Disagreement is not defect.
Where feasible, the programme seeks testing arrangements in which independent judgement is allowed to differ.
The purpose is not to maximise agreement.
It is to determine whether distinctions, classifications and propositions remain interpretable and testable when exposed to genuinely independent judgement.
Independent review is not a headcount.
Publication status and scientific status are not the same thing.
- UNDER REVIEW≠ACCEPTED
- PUBLISHED≠VALIDATED
- PUBLICLY DISCUSSED≠INDEPENDENTLY REPLICATED
- CURRENTLY USEFUL≠ESTABLISHED
Each output should be presented with the weakest accurate public label supported by its evidence.
Designed to be capable of losing.
If no meaningful residual remains after strong-comparator review, construct discrimination, empirical testing and transfer testing where relevant, the programme should narrow.
Individual constructs may be absorbed or retired.
If the research object survives while the terminology fails, preserve the question.
If even the distinct programme label loses justification, that result should remain scientifically admissible.
The evidence must be allowed to decide.