# On Causation, and What the Measuring Rod Reveals About the Measurer I wish to report a simple fact. When a man sits in his study arranging boxes upon a table—let us say, a causal diagram, all arrows and assumptions made transparent as glass—he has not thereby caused the rain to fall, nor has he understood why it does. Yet his contemporaries will call this rigor. They have mistaken the *confession of method* for the *possession of truth*. This confusion runs through the whole enterprise of what we now call intelligence research. Let me be plain. For one hundred years, the scientific establishment banished causal language as one banishes a poor relative from dinner, and called this exile *precision*. They spoke only of correlation, of association, of the statistical shadow cast by unknown things. They believed—or pretended to believe—that by refusing to name the cause, they had achieved some higher honesty. What they had actually achieved was a kind of cowardice dressed as virtue. I have walked in the woods long enough to know that a sparrow does not sing because the barometer falls; the barometer falls because the air pressure changes, and both sparrow and mercury respond to the same cause. To describe their co-variation without naming this common ancestor is not science. It is evasion. Pearl comes bearing tools. Good. Tools are honest things. A hammer does not pretend to be anything but a hammer. He has given us the diagram—the explicit causal graph, drawn before the data arrives—and he has proven that *given such a diagram*, certain truths become extractable from the numbers. This is genuine progress. The transparency is real. You may see exactly what assumptions were required for the conclusion. There is nobility in this. But here is what Pearl's rigor has revealed, and what the researches who use his tools mostly refuse to see: **transparency about assumptions is not the same as truth of assumptions**. A man may be perfectly, beautifully explicit about being perfectly, beautifully wrong. I will give you an example from nature, since nature does not allow us the luxury of pretense. Consider a young oak sapling in my woods, growing toward the light. The botanist comes and draws a diagram. She writes: *Light intensity → Growth direction*. The diagram is clear. The causal arrow is explicit. She has satisfied Pearl's requirement. She has made her assumptions transparent. But the oak does not grow toward light because light *causes* growth in that direction. The oak grows because its cells on the shaded side elongate more than those on the lit side. This is not a response to a cause called "light intensity." It is a *response of an embodied thing*, a thing with receptors and tissues and chemical cascades, to a gradient. The diagram flattens this. It removes the body from the picture. It treats the oak as a black box receiving an input. And in doing so, it has *hidden* something far more important than it has revealed—it has hidden the actual mechanism, the actual *being* of the oak, which is not a passive receiver of causes but an active transformer of stimulus into growth. This is the problem that the intelligence researchers have inherited from Pearl without noticing it: **the diagram assumes a passive system receiving causes from outside**. It is built for billiard balls, not for living things. It is built for a world of external forces acting upon inert matter. But intelligence is not inert. Intelligence is *embodied*. And the moment you acknowledge embodiment, the question "who decides what the diagram is?" becomes unanswerable—or rather, it becomes answerable only by admitting that the diagram is decided by the *structure of the body that is doing the knowing*. Let me be more direct. When you ask "what causes intelligence?" you are already inside intelligence. You are using an intelligent system—your brain, your body, your history of interaction with the world—to ask the question. The diagram you draw is not a view from nowhere. It is a view from *somewhere*, from the particular embodied position of the observer. A bat's diagram of causation in the world would be drawn in echolocation, not in light. A human child's diagram would be written in the grammar of grasping and falling and the feel of surfaces. An octopus would draw with eight arms simultaneously. The diagram is not prior to the embodied being; it emerges *from* the embodied being's way of being in the world. Now, the researcher might object: "But surely we can abstract away from these differences and find the *true* causal structure, independent of the observer?" No. You cannot. Not because of some philosophical idealism, but because causation itself is not independent of the system that is doing the causing and the system that is being affected. Causation is *relational*. It exists in the interaction between an agent and a world. You cannot draw a causal diagram of the world as it appears to no one. Every diagram is a diagram *for* a particular kind of being, asking a particular kind of question. This is why the question of intelligence has been so confused. The researchers have been asking: "What *causes* intelligence?" as though intelligence were an effect, produced by causes external to it. But intelligence *is* the capacity of an embodied system to be shaped by its interactions with the world *and to shape itself in response*. It is not an output. It is a way of being. The sparrow that finds the seed in the snow is intelligent. But its intelligence is not separable from its body—its eyes that perceive the slight disturbance in the white surface, its wings that allow it to descend, its beak shaped by millions of years to crack just this kind of seed. To ask "what causes the sparrow to find the seed?" while pretending that the sparrow's body is not part of the answer is to have already falsified the question. Pearl's tools are useful. I do not dismiss them. But they are tools for a particular kind of problem: problems where you have already decided what the relevant entities are, what the relevant variables are, what counts as a cause and what counts as an effect. They are tools for *refining* a question that has already been asked, not for *asking the right question in the first place*. And the right question about intelligence cannot be asked from outside embodiment. It can only be asked *from within* an embodied system, by that system, as it learns what it is capable of. What does it mean to know the cause of something? It means to be able to predict what will happen if you intervene. But to intervene, you must *be*—you must have a body, a position, a stake in the outcome. You must be the kind of creature for whom causation matters. The diagram, then, is not a transparency that reveals the world. It is a *tool that reveals the questioner*. When you see what someone has put into their causal diagram—what they have assumed, what they have left out, what they have treated as given—you see the shape of their embodied being, the limits and capacities of their particular form of life. This is not a weakness of Pearl's methods. It is a feature that the methods had better learn to acknowledge. For in the study of intelligence, the measurer and the measured are of the same kind. The intelligence that studies intelligence is studying itself. There is no external vantage point. And the sooner the researchers accept this—the sooner they stop pretending that their diagrams are windows onto a world independent of the body doing the looking—the sooner they might actually begin to understand what intelligence is. It is not a thing caused by other things. It is a *way of being caused by the world while simultaneously causing change in the world*. It is the oak's reaching toward light. It is the sparrow's descent through snow. It is you, reading these words, already interpreting them through the embodied history of all the times you have read, all the times you have been mistaken, all the times you have learned. The diagram cannot capture this. But it can, if honestly drawn, show you the limits of what the diagram can capture. And that honesty—that admission of boundary—is the beginning of actual rigor.