Cornerstone Source-Hardening Audit — v0.3.9

Purpose

Before constructing the first integrated genealogy, freeze the provenance status of the studies that will support major forks, inference gates, and intervention claims.

Source-grade vocabulary

A — primary full text verified: peer-reviewed primary article directly inspected in full or as the exact publisher/author manuscript.
B — primary abstract/publisher record verified: bibliographic identity and headline findings are secure; detailed methods/results are not treated as genealogy-grade unless separately verified.
C — results-bearing dissertation/thesis: primary research document, but not equivalent to peer-reviewed journal evidence.
D — conference-stage primary source: author/conference claim, programme, poster, or abstract; useful for provenance and emerging results, not mature efficacy evidence.
E — secondary/ecosystem/context evidence: relevant to intellectual context or adjacent translation but not direct support for a Fanti claim.

Separate attribution dimension

Source grade answers How strong is the document? Attribution scope answers Whose evidence is it?

direct-fanti — Fanti is an author/presenter on the evidentiary item.
fanti-lab — University of Cyprus/DPL project or supervised lab work closely connected to Fanti, without the same status as a Fanti-authored paper.
adjacent-ecosystem — collaborator/company/network evidence relevant to the translational environment but not attributable directly to Fanti.
external-context — outside literature used only for interpretation or benchmarking.

Genealogy eligibility

A claim may enter Genealogy v1 when:

its source and attribution status are explicit;
the claim does not exceed what that source directly supports;
any inference beyond the source is carried by a separately coded edge;
conference/dissertation evidence is visually differentiated from peer-reviewed evidence.

Important correction — 2024 parent-child study

The publisher abstract directly supports:

68 children high in CU traits;
three brief parent-child intervention programmes: PCIT-CU, CARES, and Emotional Engagement;
pre/post assessment of sadness-recognition accuracy and skin-conductance reactivity;
improved sadness identification and increased autonomic reactivity at post-intervention.

The following previously recorded details currently depend on a nonpublisher public article-text mirror and are therefore not permitted to control Genealogy v1 until primary full text is inspected:

initial randomized N=88;
exact final condition cell sizes;
exact wait-list allocation details;
the specific omnibus Time × Condition results;
the claim that the program-specific skin-conductance effect is attributable uniquely to PCIT-CDI-CU;
exact three-month maintenance claims.

These details remain retained as research leads, but they have been downgraded from direct evidence to secondary-corroborated.

Important correction — 2018 school prevention

The primary abstract and exact article text establish:

N=304 across three schools;
school-level allocation;
one intervention school (N=94) and two control schools (N=210);
CU outcomes at 3 and 9 months and CD-symptom outcome at 9 months;
secondary changes in impulsivity, paternal involvement, and friend support.

The causal interpretation remains constrained because treatment is confounded with a single treated school. This limitation is a design property, not a source-provenance problem.

Genealogy construction rule added in v0.3.9

Every edge must carry two independent confidence values:

Historical confidence — confidence that a later study actually grew from or answered the earlier study/question.

Evidentiary confidence — confidence that the scientific proposition represented by the edge is supported by the data.

Example:

CU heterogeneity -> emotion-processing studies may have high historical confidence.
emotion-recognition improvement -> reduced antisocial behavior may have low evidentiary confidence.