The cycle’s central challenge is to see the river of intelligence clearly—without the narcotic of hype or the paralysis of fear—and to understand what kind of intelligence it carries. Pearson’s framework reveals the mathematical nature of that intelligence with uncomfortable precision: it is correlational intelligence, first-rung in Pearl’s hierarchy, supremely powerful at the level of describing regularities and systematically unable to answer the questions that require causal reasoning. The system that tells you what tends to follow what cannot tell you what would happen if you intervened to change something. It can describe the world as it has been. It cannot reason about the world as it could be.
Pearson also explains the borrowed authority of objectivity that the cycle identifies as the central hazard of algorithmic decision-making. His Grammar of Science gave statistics the philosophical status of the universal method of knowledge, and that status has descended—unexamined—to the machine-learning systems that are now sorting people for jobs, loans, parole, and medical care. The “data shows” carries the same authority, and conceals the same buried values, that “the statistics prove” carried a century ago. Pearson’s eugenics was more dangerous than a bigot’s prejudice precisely because it wore the lab coat; and algorithmic systems wear the same lab coat, better tailored.
The cycle’s question “are you worth amplifying?” receives from Pearson’s career its most direct cautionary answer. Pearson was worth amplifying in the mathematical sense—his methods were genuine, his rigor was real, his contribution to the tools of knowledge was enormous. He was catastrophically worth amplifying in the moral sense—the amplification of his methods by institutional support and public authority produced forced sterilizations across two continents and intellectual cover for the racial pseudoscience of the Third Reich. The lesson is not that powerful methods should not be amplified. It is that the amplification of powerful methods without a robust account of cause, without commitment to human dignity, and without the kind of conscience that Newman described as the aboriginal authority, produces harm at the scale of the method’s reach.
In the cycle’s gallery of thinkers, Pearson stands opposite Judea Pearl: the founder of the correlational paradigm facing the thinker who spent his career arguing that the paradigm was incomplete and that the incompleteness was not a minor technical limitation but a century-long detour away from the questions that matter most when intelligence is applied to consequential action.
Karl Pearson was born Carl Pearson in London on March 27, 1857. He changed the spelling of his given name to Karl in his early twenties—an homage, reportedly, to Karl Marx and to the German intellectual world that had reshaped him during years of study in Germany. He came up to King's College, Cambridge, on a mathematics scholarship, sat the Tripos examination, and emerged Third Wrangler in 1879—one of the three highest mathematics graduates in his year, in a cohort that included students taught by James Clerk Maxwell himself. He came to statistics not as a clerk of numbers but as a mathematical physicist trained to believe that the universe yields to equations.
His engagement with statistics began through biology. In 1890 the zoologist W.F.R. Weldon arrived at University College London with a question—how do you measure evolution?—and Pearson had the mathematical power to answer it. He read Francis Galton’s Natural Inheritance (1889) and saw what Galton had glimpsed: that the statistical description of populations could be made exact. Between 1893 and 1912 he published eighteen papers under the title “Mathematical Contributions to the Theory of Evolution” that effectively invented the toolkit of modern statistics. He coined the term “standard deviation” in 1893. He established and edited the journal Biometrika from 1901 until his death. He held the Galton Professorship of Eugenics from 1911 to 1933.
His philosophical manifesto, The Grammar of Science (1892), articulated the positivist doctrine that science describes regularities and must remain silent about causes—the doctrine that the young Albert Einstein recommended to his reading circle and that, a century later, every large language model embodies by construction. Pearson died in 1936, seven years before Alan Turing published the paper that would eventually lead to the machines that most completely instantiate his philosophy.
The correlation coefficient as the engine of AI. Pearson converted the intuition of “things going together” into exact geometry: the product-moment correlation coefficient, the cosine of the angle between two centered variable vectors, a number between negative one and one that measures co-movement with precision. When a modern transformer computes attention—asking which earlier tokens are relevant to the current one—it performs cascades of scaled dot products between vectors. This is Pearson’s operation in everything but name. Contemporary AI is a Pearsonian intelligence: the correlation coefficient, computed at planetary scale, billions of times per second.
The grammar of description. Pearson’s Grammar of Science held that science describes regularities and must remain silent about causes: “A scientific law is nothing more than a brief description in mental shorthand of as wide a range as possible of the sequences of our sense-impressions.” This positivist doctrine—description without explanation, sequence without cause—is exactly the philosophy that every large language model embodies. The model classifies, recognizes sequences, and acts on them with no model of cause whatsoever. It cannot tell the difference between a correlation that reflects a fair mechanism and one that launders historical injustice into present harm.
Correlation is not causation—but Pearson thought it was. The proverb that is attached to his name as a warning was, for Pearson himself, a metaphysics: causation is the limiting case of very tight correlation. This move locked statistics—and by inheritance, machine learning—to the bottom rung of Pearl’s ladder of causation, the rung of association and pattern, where systems that can only describe cannot answer the interventional questions that consequential action requires. A model trained on data in which asthmatic pneumonia patients survive at higher rates will recommend against aggressive treatment for asthmatics—because the correlation is real and the causal story it conceals is lethal.
Biometrics and the measured human being. Pearson’s project of biometrics—reducing persons to measurable attributes and studying the statistical structure of those attributes—is the intellectual ancestor of every modern system that represents people as feature vectors. The reduction is simultaneously the source of these systems’ usefulness and their most insidious harm: it encodes a prior decision about what a person is for the system’s purposes, a decision that disappears into the apparent neutrality of the numbers and then does moral work that the cloak of quantification conceals.
Eugenics and the catastrophe of ranking people. Pearson did not separate his mathematics from his eugenics. He believed they were the same project: applying the statistical method to the measurement and improvement of human populations. His 1925 paper purporting to demonstrate the inferiority of immigrant Jewish children, his admiring reference to Nazi proposals as “a vast experiment” in 1934—these were not aberrations from his scientific commitments but applications of them. The catastrophe was not a misuse of a neutral tool. It was the direct expression of the tool’s blind spots about cause, about value, and about what a person is.