10.9 CLIC: creative causal learning

CLIC—Causal Learning with Infinitesimal Creativity—is the DIAL-SKFM realization developed in Chapter 11. Its target is an interventionally admitted causal structure rather than merely the highest-scoring graph. One repair direction estimates local intervention fields and their uncertainty; the other assembles or revises their global causal factorization. Lie-bracket nonclosure supplies a localization signal, but not by itself a causal interpretation.

CLIC makes the architectural creative mandate explicit. After a registered nonrealizability gate rules out the relevant current-sketch alternatives, it may propose a missing directed mechanism, latent common-cause bundle, altered latent dimension, or revised factorization. Unsupported alternatives participate in inference and may request new evidence or veto construction, but they are not silently converted into admissible repairs.

The controlled ladder in Chapter 11 keeps the repair metalanguage registered while isolating four burdens: exact closure repair, operator uncertainty, active counter-witness interventions, and explicit unsupported alternatives. It tests typed construction and guarded admission inside that metalanguage; it does not demonstrate invention of a new causal variable or mechanism type.