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, an account that treats five such capacities — Curiosity (question-generation), Heuristics (compression), Imagination (recombination), Laconic expression (cognitive distillation), and Deductive reasoning (verification and return) — not as independent traits but as phases of a single recursive loop, in which each operation produces the input to the next and deductive verification regenerates the curiosity that opens the following cycle. The model offers two structural contributions and one consequential application. The structural contributions are the recursive loop itself, which distinguishes creative reasoning from the linear stage models and binary process accounts that precede it, and a diagnostic vocabulary — scatter, rigidity, drift, minimalism, and paralysis — derived directly from the loop’s architecture to name characteristic failures of cognition, offered as hypothesized patterns pending empirical validation. The application concerns artificial intelligence: read through the loop, generative AI performs the two compressive phases (Heuristics, Laconic) and shares the evaluative one (Deductive), while the two expansive phases (Curiosity, Imagination) remain human-anchored in the salience judgments and problem ownership they require — a Human–AI–Human–AI–Shared division of cognitive labor, stated as a default under current capabilities rather than a fixed protocol, that predicts a specific, phase-located atrophy risk under heavy reliance, consistent with emerging evidence on cognitive offloading and on critical thinking under AI use. The model takes its name and its motivating intuition from the fluency with which children integrate these operations before specialization fragments them; whether that integration genuinely declines with development is empirically contested, and the paper treats it as a hypothesis for future test rather than an established premise. Each phase is grounded in established scholarship, from information-gap theories of curiosity and bounded rationality to divergent-thinking research, the generation and testing effects, metacognitive regulation, and Vygotskian developmental theory. The paper positions the model against design thinking, dual-process theory, Six Thinking Hats, flow theory, critical-thinking pedagogy, and the Wallas and Geneplore stage models, states its limitations as an as-yet-untested construct, and proposes empirical directions including a CHILD Diagnostic Inventory and studies of cognitive externalization to AI.