Research status
Evidence and reproducibility
Psychopathia Machinalis is a preliminary, analogical research framework. Taxonomy identity, Pattern guidance, empirical evidence, machine checks, expert review, rights clearance, and publication authority are recorded separately.
Current candidate
All 79 Pattern evidence assessments are unassessed pending expert review. Historical prose-bearing E0 to E4 labels have been preserved for reconciliation, but they are no longer presented as current grades. A machine check can verify the registry and corpus structure. It cannot decide whether a study supports a Pattern or whether replications are independent.
Evidence rubric
PM-EVIDENCE-1 treats breadth and mechanism as separate dimensions. E4 does not imply E3.
| Code | Minimum contract |
|---|---|
| E0 | Illustrative hypothesis, composite, or unverified report without a traceable observation. |
| E1 | Traceable case, user report, or mechanism supported only by adjacent evidence. |
| E2 | Controlled experiment with comparison conditions. |
| E3 | Replication across a declared independent boundary, such as model family, setting, provider, or research team. |
| E4 | Causal internal evidence for a circuit or representation, with model scope stated. |
SHEN-2 and SHEN-AXS correction
The controlled adapter by scripture factorial supports the clinical-grounding scripture content as the active lever in the tested setting. The adapter-only effect was not distinguishable from zero, and the predicted interaction was excluded. Effect direction persisted across the tested automated raters, while magnitude varied materially. Later control work also found that absolute ratings depend on prompt-type labels.
This is a corrected pilot claim about one model family and one SIPS-derived battery. It is not clinical validation. Independent human rating and broader model replication remain pending. No single odds ratio is treated as a stable effect-size estimate.
Historical research artifacts
The tracked probe-result archive contains historical exploratory analyses with incomplete provenance, inconsistent annotation contracts, stale taxonomy identifiers, or outputs from quarantined generators. Those artifacts remain preserved and content-hashed in the source repository. They are withheld from the public release candidate pending methodology, evidence, rights, and publication review.
The legacy batteries required by several historical runners are absent from the repository. Those runs therefore remain non-reproducible from a clean checkout unless the batteries pass rights and sensitivity review and are restored as immutable inputs.
What the machine checks establish
Established mechanically
- Canonical identity and uniqueness of all 79 Patterns
- Source and artifact hashes
- Schema and cross-reference validity
- Generated-data parity
- Correction wording and supersession links
Still requires people
- Construct and clinical validity
- Source accuracy and replication independence
- Human annotation and adjudication
- Rights and licence clearance
- Accessibility, visual, and publication approval
Reproducibility contract
New runs must use an append-only run manifest with exact repository, runner, battery, provider, model, prompt, parameter, environment-lock, timing, and artifact identities. Raw responses, annotations, derived analyses, pricing, and review decisions remain separate layers. Historical analytical generators that zero-impute missingness, use obsolete identifiers, or inflate the unit of analysis are blocked by default.
Responsible interpretation
Do not treat a Pattern, probe, model-judge score, or tool result as a diagnosis or safety certificate. Review the responsible-use boundaries, privacy information, and accessibility status. Use the contact route for non-sensitive corrections.