From Observable Behaviour to Developing Intelligence

This preprint examines what it means to explain intelligence scientifically by reconstructing three influential traditions in twentieth-century psychology: behaviourism, the Cognitive Revolution, and Piagetian constructivism. Rather than presenting these traditions as successive stages in which one approach replaced another, the paper interprets them as foregrounding three complementary explanatory dimensions. Behaviourism established the importance of observable performance and reproducible behavioural evidence. The Cognitive Revolution expanded explanation towards internal representations, processes, and cognitive architectures. Piagetian constructivism introduced a diachronic concern with how cognitive organisation emerges and changes through development. On this basis, the paper proposes a Performance–Mechanism–Development framework: Performance: What can a system reliably do? Mechanism: What organised processes produce its behaviour? Development: Through what history of learning, interaction, and reorganisation did its capacities emerge? The framework is applied to contemporary artificial intelligence, including benchmark evaluation, explainable AI, mechanistic interpretability, continual and lifelong learning, curriculum learning, and developmental robotics. The paper argues that performance is indispensable evidence of capacity but does not by itself explain intelligence. A more complete scientific account must distinguish what a system does, how it does it, and how it became capable of doing it. This is the fourth preprint in the Rethinking Mind research project: A Historical and Conceptual Journey from Ancient Philosophy to Artificial Intelligence.