The Art of Making Problems Simple: A Theory of Intelligence

Human intelligence is flexible and general, yet its underlying structure remains poorly understood. We propose that intelligence arises from the ability to construct and revise representational normal forms, that is, coordinate systems in which prediction, comparison, and inference become simple. This capacity rests on four operations: identifying lawful generators of variation, quotienting task-irrelevant differences, constructing measurable axes for ordered reasoning, and selecting representations that locally flatten the space of possibilities.Across development, these operations support a progression from early object perception to abstract and metaphorical reasoning. The framework explains why classic measures of fluid intelligence, such as Raven’s matrices, analogical reasoning, and mental rotation, generalize so broadly: each probes the ability to reorganize a problem into an appropriate normal form. The theory also accounts for systematic individual differences, including cognitive profiles associated with autism, as principled trade-offs in representational precision and flexibility. Together, these results identify structural constraints on intelligence that are not jointly captured by existing cognitive or computational frameworks.