The CHILD Model: A Five-Phase Cognitive Loop for Creative Reasoning and Its Division of Labor with Artificial Intelligence

Abstract

Creative reasoning is usually studied one capacity at a time. This paper introduces the CHILD Model, which treats five such capacities — Curiosity (question-generation), Heuristics (compression), Imagination (recombination), Laconic expression (cognitive distillation), and Deductive reasoning (verification and return) — as phases of a single recursive loop. Cyclical accounts of the creative process exist, from Wallas through the Geneplore model and the process models of Mumford and Basadur; what the CHILD Model adds is the loop's internal architecture. Its phases alternate by cognitive direction — each expansive operation is followed immediately by the compression that disciplines its output, before a resolving phase closes the cycle — and each handoff passes an output of a specified kind, from targeted question to problem representation to possibility set to testable proposition. Two elements of that architecture are, to the author's knowledge, new: the formalization of distillation as a distinct phase between generation and evaluation, and the residual — the unresolved remainder of verification — as the specified mechanism by which one cycle's close becomes the next cycle's opening. From the architecture the paper derives a diagnostic vocabulary of five phase-located failures (scatter, rigidity, drift, minimalism, paralysis), offered as hypotheses pending validation, and a three-tier division of cognitive labor with generative AI — human-anchored, AI-accelerated, and shared-evaluative — stated as a default under current capabilities, with a phase-located atrophy risk under heavy reliance. The model's name records its motivating intuition, the fluency with which children integrate these operations; the paper treats that observation as a hypothesis, not a premise. Limitations and empirical directions, including a falsifying test of the model's central phase distinction, are stated.