PJC Evidence and Cases
Disclosure date: 25 August 2026
Current conclusion: PJC has an executable method structure, but insufficient evidence that it improves calibration, accuracy, efficiency, or team performance.
1. Status overview
| Evaluation target | Status | Public basis |
|---|---|---|
| Method definition | Specified | v0.6 documents and field schema |
| Minimum operability | Under evaluation | No complete pilot dataset published |
| Record completeness | Not measured | Internal pilot pending |
| Settlement rate | Not measured | Internal pilot pending |
| Confidence calibration | Not validated | Comparable records required |
| Error detection | Not validated | No baseline or control |
| Work efficiency | No claim | Recording cost unmeasured |
| Cross-context validity | Not validated | No external cases |
| External peer review | Not conducted | Criticism invited |
2. What current evidence supports
PJC has a consistent field schema and lifecycle; append-only commitment can preserve pre-outcome text; the claim register separates mechanism claims from empirical hypotheses; and the method declares non-use conditions and invalid states.
These are facts about method design and implementation, not behavioural effects.
3. What current evidence does not support
PJC must not currently be described as making people more rational, improving decision accuracy, making agents reliable, lowering costs, fitting every team, or being scientifically validated.
Any future effect statement must link to evidence or remain explicitly untested.
4. Internal pilot plan
The first 30–60 day pilot will cover recurring, low-risk product, engineering, agent-workflow, user-behaviour, and release-risk judgments.
Priority measures:
| Measure | Definition |
|---|---|
| Completion rate | Records with every required field |
| Timely commitment | Records committed before outcome availability |
| Settlement rate | Records settleable under the original rule |
| Invalidation rate | Records excluded for leakage, rule switching, or data failure |
| Continued use | Participants recording across multiple cycles |
| Recording cost | Time spent creating, reviewing, and settling |
Calibration and decision quality come only after basic operability.
5. Case publication standard
Every published case includes context, roles, agent involvement, duration, record count, original propositions, settlement criteria, actual outcomes, evidence sources, observed and unobserved changes, failed records, data limits, conflicts of interest, and reviewability.
Internal cases carry this notice:
An internal case is not independent validation.
6. Case register
No case currently meets the publication standard.
This is an evidence boundary, not a placeholder. Demonstrations will not be presented as empirical validation.
7. External evaluation
Send criticism, failed cases, or replication proposals to contact@sedesmentis.com. We especially value examples of unsettleable tasks, field-induced bias, cost exceeding benefit, agent contamination, or no improvement over a lighter baseline.
Substantive changes are recorded in the changelog.