Causal Reasoning
Post Hoc Fallacy,
Illusory Causation
Planning
Planning Fallacy,
Typicality Bias
Counterfactual Thinking
Hindsight Bias,
Outcome Bias
Social Inference
Fundamental Attribution Error,
Stereotyping
Intelligence is hard to observe directly—but its failure modes are not. Across domains, we see systematic biases emerge not from randomness, but from structured pattern matching gone wrong. By tracing these consistent breakdowns, we can infer the underlying architecture of intelligence itself.
Recent advances in large language models have challenged long-held assumptions about the nature of intelligence, showing that behaviours once thought to require symbolic reasoning or abstract planning can emerge from large-scale pattern matching.
We argue that this is not an artifact of artificial systems, but a reflection of how human intelligence itself operates. We support this view with two converging lines of evidence. First, cognitive biases—such as stereotyping and base-rate neglect—reflect systematic misalignments in pattern recognition, rather than logical errors. Second, intuition, creativity, and sudden insight often stem from rapid alignment with stored patterns, enabling flexible responses without explicit reasoning.
We revisit core cognitive functions—language, causal reasoning, planning, and social inference—and show that they can be unified under a single mechanism: memory-driven pattern matching. This reframing suggests that many forms of intelligent behavior can be explained as pattern matching over stored experience, rather than abstract computation.
Intelligence is one of the most used—and least understood—concepts in science and technology. It’s invoked to sell products, justify theories, and predict the future. But what is it, exactly? The term is often overloaded, vague, or circular. Many use it as a placeholder for whatever humans or machines seem to do well. Others overhype it, assuming intelligence is a singular essence that will soon be replicated.
But real understanding begins with constraints. By focusing not on what intelligence can do, but where it consistently fails, we shift from speculation to structure. Biases and breakdowns aren't noise—they are the visible seams of an otherwise hidden system. This perspective reframes intelligence not as mystery or magic, but as a pattern-matching engine with traceable limits.