Two Thresholds in Humans and Artificial Intelligences

This article presents the Two-Threshold Model (TTM), a comparative experimental framework for investigating response emergence, rational reconstruction, and explanatory fidelity in human beings and artificial intelligence systems. The model introduces Preconscious Relational Latency (PRL) to examine whether a response can emerge before the complete explicit reconstruction of the relations that support it. The proposed protocol uses a Compositional Relational Grammar (CRG), controlled experimental conditions, and tasks involving identifiable, manipulable, and verifiable relations. It distinguishes the response-emergence threshold from the explanatory-reconstruction threshold and defines the Emergence–Explanation Interval as a central variable for comparing response production and explanatory completion. The study introduces operational measures including the Relational Latency Index (RLI), Reconstructive Fidelity, Reconstructive Precision, Metacognitive Gain, and Structural Transfer Index. It establishes nine discriminant experimental predictions addressing relational complexity, intuitive emergence, structural transfer, explanatory demands, confidence calibration, incubation, and the recovery of relations through cues. The protocol specifies experimental stages, sample-design principles, temporal recording procedures, explanation-coding criteria, statistical analysis methods, model-comparison strategies, and reproducibility requirements. Its comparative design enables human participants and AI systems to be evaluated through equivalent relational tasks while distinguishing observable performance from assumptions about internal processing. The central contribution is a reproducible research architecture for examining whether response correctness, explanatory quality, and fidelity to the relations underlying a solution constitute distinct evaluative dimensions. The framework provides a methodological basis for empirical investigation in cognitive science, metacognition, philosophy of science, scientific education, and artificial intelligence evaluation.