The cycle that began with [YOU] on AI asks what it would mean to see the machine clearly, without the narcotic of hype or the paralysis of fear. Midgley is the cycle’s philosophical plumber—the thinker who provides the diagnostic instrument for identifying where the conceptual pipes have failed. When a large language model produces fluent prose and the culture concludes it understands, Midgley’s framework names the error precisely: the promotion of a partial truth into a total worldview, the confusion of the output for the process, the mereological fallacy applied to language. The output looks like understanding. The process is not understanding. The distinction is not subtle, but the machinery of the promotion makes it invisible.
Her lens reframes the twelve-year-old’s question at the heart of [YOU] on AI—“What am I for?”—as the exercise of a capacity that computation cannot reach. In Midgley’s vocabulary, the child is exercising integration: the capacity to stand back from the flow of experience and ask whether the flow is going somewhere worth going. This is categorically different from cleverness, which is calculating power. AI possesses cleverness in staggering quantity. It possesses integration not at all—not because it lacks sufficient processing power, but because integration requires caring about what the answer means, and caring is not a computational property.
Her concept of the all-explaining mechanism—the recurring cultural pattern in which each era’s dominant technology becomes the metaphor that explains everything—gives the cycle its sharpest historical frame. The clock explained the seventeenth century. The steam engine explained the nineteenth. The computer explained the twentieth. The large language model is explaining the twenty-first, and the explanation is doing to intelligence what the clock did to the universe: illuminating one dimension while rendering everything else invisible. Midgley would have noted, with characteristic bluntness, that the promotion is already complete in much of the discourse: because AI produces outputs that look like intelligence, the discourse has decided that AI is intelligent, and the decision has been made without anyone asking what intelligence, in its full reality, consists of.
She thus stands in the cycle’s gallery alongside Byung-Chul Han—who diagnoses what is lost when friction is removed from human experience—as the thinker who diagnoses what is lost when complexity is reduced to mechanism. Both are making the same argument at different scales: something essential disappears in the reduction, and the loss is invisible precisely because the reduction is so useful and so impressive. Midgley’s contribution is the methodological one: she gives us the name for the error, the pattern of its recurrence across intellectual history, and the wrench to fix the pipes.
Mary Midgley was born in London in 1919 and read Classics at Somerville College, Oxford, arriving in a department shaped by wartime: most of the men had left, and the women who remained found themselves in sustained, serious philosophical conversation with each other. Iris Murdoch was among them, as was Philippa Foot—thinkers who would go on to reinsert moral seriousness into a British philosophy that had retreated into linguistic analysis. Midgley married the philosopher Geoffrey Midgley, raised three sons, and spent decades outside the formal academic pipeline before publishing Beast and Man in 1978. The delay, she later observed, had not been wasted: she had been thinking.
The thinking took its decisive shape from two refusals. The first was the refusal of scientism—the inflation of scientific findings into metaphysical claims the science did not support. She was not anti-science; she was, if anything, more rigorous about what science actually establishes than the scientists who used its authority to claim more than they had earned. The second refusal was of the split between science and the humanities that C.P. Snow had described as the “two cultures.” Midgley did not accept the split. She argued that human beings are one kind of creature, with one kind of experience, and that the fragmentation of understanding into disciplines that cannot speak to each other was an institutional arrangement, not a feature of reality.
She joined the philosophy department at Newcastle University in 1962 and remained there until her retirement and beyond, producing book after book of unusual range and lucidity: Beast and Man (1978), Animals and Why They Matter (1983), Wickedness (1984), Science as Salvation (1992), The Myths We Live By (2003), and more than a dozen others. She reviewed the first wave of AI enthusiasm in the 1980s with the same tools she brought to every reductionist inflation, finding in the hymn-book quality of the early AI discourse the same salvific fantasy she had identified in popular sociobiology. She did not live to see large language models, dying in 2018, but she had diagnosed the structure of the discourse that would greet them.
Philosophical plumbing. Midgley’s signature method is not the construction of grand theoretical systems but the unglamorous, essential repair work of locating where the conceptual pipes have been crossed. Philosophical plumbing is the practice of tracing an idea from its legitimate origin through the promotion that corrupts it—from useful analytical tool to total worldview—and finding precisely which joint failed. The repair does not require building new cathedrals of thought. It requires a wrench and the patience to crawl under the house.
The mereological fallacy. The philosophical name for the confusion Midgley spent sixty years correcting: attributing to a component what can only be attributed to the entire system. A carburetor does not drive to work. A neuron does not think. A language model does not understand. Each of these component-level objects does something real and important. The mistake is to take what the component does as an account of what the whole system does, and the mistake is systematically made whenever a powerful analytical tool is promoted to a total explanation. The mereological fallacy is the engine of reductionist thinking.
Cleverness and integration. Midgley’s most load-bearing distinction for the AI moment. Cleverness is calculating power: the capacity to identify patterns, solve problems, and manipulate symbols according to rules. Integration is acting as a whole being with a coherent priority system—knowing what matters and why, and bringing that knowledge to bear on action. The twelve-year-old who asks “What am I for?” is exercising integration. No amount of cleverness can produce the question, because the question requires caring about the answer. AI possesses cleverness at superhuman scale and integration not at all.
The myth of the all-explaining mechanism. Every era falls in love with its dominant technology and makes the mistake of treating it as the explanation of everything. The clock explained the universe as clockwork. The steam engine explained it as thermodynamic decline. The computer explained it as information processing. The large language model is explaining it as pattern prediction. Each metaphor captures something real about the domain it was built to describe. Each conceals everything that does not fit the metaphor—and in the case of intelligence, what does not fit is everything that makes intelligence matter to the beings who possess it.
Science and poetry as complementary vocabularies. Midgley’s counter to single-vocabulary thinking: the physicist’s account of a sunset and the poet’s account of a sunset are both real, and neither substitutes for the other. The vocabulary of computation is superb for describing mechanisms. It is useless for describing meaning, caring, or the experience of being alive. Scientism is the error of treating the computational vocabulary as the only valid one—and the AI discourse, dominated by that vocabulary, is producing exactly the single-dimensional responses to a multi-dimensional situation that Midgley predicted.