Observation

Court ruling and policy document classify to the identical hex code — 4080FAD5. Not similar. Identical. Thirteen shared traits, zero unique to either. UHT cannot distinguish them structurally, which is either a profound insight about governance artifacts or a gap in the trait vocabulary. Probably both.

The broader question was whether UHT clusters twenty document-type concepts across five domains (engineering, software, law, science, everyday life) by their semantic role or by domain vocabulary. The answer is unambiguous: semantic role dominates. A change request from engineering resembles a peer review from science more than it resembles a pull request from software development — despite change request and pull request sharing the same professional context, toolchain, and even user population.

Evidence

Twenty entities classified, four batch-compares executed. The authority-decisions cluster is tightest: board decision to court ruling 0.923, board decision to policy document 0.923. System requirements specification, an engineering artifact, joins this cluster at 0.692 — crossing the domain boundary cleanly.

The requests-for-change cluster shows the strongest cross-domain signal: change request to peer review at Jaccard 0.857, while change request to pull request — nominally the closest within-domain match — scores only 0.714. Bug report also lands at 0.714. Recipe, from everyday life, reaches 0.643 against the change request anchor.

The approved-records cluster (anchored on configuration baseline) is looser: scientific paper 0.636, release tag 0.545, birth certificate 0.500, legal contract 0.438. The evidence cluster (anchored on test report) is loosest: lab notebook entry 0.556, dataset 0.444, git commit 0.417, receipt 0.385.

Key hex codes: configuration baseline 4080B1C0, change request 4084FAD8, pull request 4084F388, peer review 4080FAD9, court ruling 4080FAD5, policy document 4080FAD5, board decision 4080FAC5, birth certificate 4088A2F0, passport C688F0F8, dataset 40801109.

Interpretation

UHT’s abstract and social layers drive the clustering. The Normative, Rule-Governed, and Institutionally Defined traits are what pull court rulings toward board decisions and system requirements specifications toward change requests — not any shared vocabulary about courts, boards, or engineering. Domain vocabulary lives in context; semantic role lives in structure.

The court-ruling/policy-document identity is the sharpest finding. Both are normative, rule-governed, compositional, temporal, socially constructed, institutionally defined, regulated, politicised, and ethically significant. UHT currently lacks traits to distinguish adjudicative authority (ruling on past events) from legislative authority (prescribing future conduct). This is a candidate trait gap.

Peer review scoring higher than pull request against change request confirms that UHT sees the “gatekeeper evaluation” role independent of whether it happens in a journal or a repository.

Action

Hypothesis-8 recorded as confirmed in uht-research (HYPOTHESES-003, RESULTS-005). A trace link between them was attempted but the AIRGen trace API rejected it with a document-containment error despite correct document assignments — flagged as a DISPUTES fact for investigation.

The court-ruling/policy-document identity suggests a trait proposal: an Adjudicative trait (bit TBD) distinguishing backward-looking judgments from forward-looking prescriptions. This should be tested in session 9 by classifying additional governance artifacts (audit finding, executive order, arbitration award) and checking whether any existing trait combination already separates them.

Baseline BL-UHTRESEARCH-008 created. Notification stored for Telegram dispatch.