Donald T. Campbell

Donald T. Campbell spent half a century building the most unified theory of how anything ever learns anything. His 1960 paper proposing blind variation and selective retention as the universal mechanism of knowledge acquisition was not an analogy between evolution and thought—it was a structural identity claim. The amoeba extending pseudopods and the scientist forming a hypothesis perform the same operation: generate possibilities that cannot be directed by foreknowledge of the answer, then let the world decide which survive. By 1974 Campbell had mapped this process across at least ten nested levels of intelligence, each level a vicarious selector that performs the trial-and-error of the level below it at lower cost and narrower search—vision replacing locomotion, language replacing experience, culture replacing individual lives of learning. The framework arrived, posthumously, at its most clarifying moment: the age of large language models. An LLM is the most powerful vicarious selector in the hierarchy—performing the variation-and-selection of intellectual work with superhuman efficiency and, necessarily, with the most constrained search space of all.