Testing a Coherence Architecture Across Domains of Human Cognition: Prospective Identification, Falsification, and Boundary Conditions
Type: article
Abstract: This paper asks whether the context-gated human-side model defined in How Morals Become Ethics (Paper I v20) can generate prospectively discriminating tests across domains without treating a common graph vocabulary as an explanation. Paper II remains in exact zero closure: A1-A17 and the canonical CML, BOWB, EBG, CSLP, LPC, and BSIC mechanisms belong solely to Paper III v20 and do not operate here. The inherited F/G/L/R/N family retains typed representation, bounded nonnegative transmission, nonredistributing gates, exogenous cue/context inputs, fan-centered log weights, compilation, readouts, and residual traces. CAP/CSC permit controlled external composition; Paper II adds only prospective tests of frozen protocol behavior and applies Paper IV v20's IAP/FHFS freeze, human-reference, held-out, intervention, rival, and stopping requirements without importing an equation. Passing establishes at most finite profile-relative evidence for the named composition, never universal human-likeness. The manuscript establishes neither identification nor empirical success, and all mappings remain prospective, exploratory, or Residual until their declared recovery, leakage, calibration, rollback, settlement, consequence, selective-ablation, rival, and held-out criteria are met.
Author: Spargur, Brian
Language: English
Subfield: Philosophy
Field: Arts and Humanities
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W7220218636
