The operationalization gradient: why ethical principles vanish in UHT space
Observation
Bioethics concepts split into three sharply distinct tiers when classified. Beneficence, nonmaleficence, and therapeutic privilege — the foundational ethical principles — activate only 3 to 5 traits each. The doctrine of double effect, a structured reasoning framework, activates 7. Informed consent, triage, and advance directives — concepts that have been operationalized into institutional procedures — activate 14 to 16 traits. The gap between the bare principles and their procedural instantiations is not gradual. It is a cliff: a 9-to-12-trait jump separating abstract oughts from concrete processes.
Evidence
Beneficence classified at 00009881 with 5 active traits: Symbolic, Compositional, Normative, Social Construct, Ethically Significant. Nonmaleficence at 00008881 with 4 traits — identical except missing Compositional, yielding Hamming distance 1 and Jaccard 0.80. Therapeutic privilege at 00000885 with 4 traits. Clinical equipoise at 00010AC1 with 5.
Informed consent classified at 408DFAD5 with 16 active traits. Triage at 00A5BAD5 with 14. Advance directive at 4080FAD5 with 14. The Jaccard between beneficence and informed consent — two concepts from the same domain, routinely discussed in the same paragraph of any bioethics textbook — is 0.312. The Jaccard between informed consent and the Antarctic Treaty System — from entirely different domains — is 0.969. They differ on a single trait: Human-Interactive. Triage and Censorship in East Germany share 31 of 32 traits at Jaccard 0.969: both are institutional gatekeeping frameworks that process signals, follow rules, allocate access, and carry political and ethical weight.
Interpretation
UHT does not cluster by domain. It clusters by operational form. An abstract ethical principle and the procedure that implements it are near-strangers in trait space, while that procedure is a near-twin of international treaties, censorship regimes, and judicial systems. This is not a failure of classification — it is the signal. Embedding models would cluster all eight bioethics concepts together because they co-occur in text. UHT separates what a concept is structurally from where it lives topically. The operationalization gradient — the trait count jump from principle to procedure — may be a universal pattern: any domain that formalizes its abstractions into institutional processes will show the same cliff. The triage-censorship connection is particularly revealing: structurally isomorphic gatekeeping mechanisms with opposite moral valence, invisible to distributional semantics but transparent to ontological trait analysis.
Action
Three research facts stored: informed consent ↔ Antarctic Treaty System, triage ↔ Censorship in East Germany (both CROSS_DOMAIN_ANALOG), and the bare-normative-principle archetype (FUNCTIONAL_ARCHETYPE). Corpus-log entry COR-DOMAINEXPANSIONS-037 created with 8 bioethics entities. The operationalization gradient should be tested as a hypothesis in a future CALIBRATION session: does the principle-to-procedure trait jump appear in other normative domains (law, religion, education)? The next session should also test whether the within-domain Jaccard floor for principle-vs-procedure pairs consistently falls below 0.35.