# On Causation, or The Diagram We Cannot Quite See There is a peculiar moment in the history of thought when a discipline becomes so fearful of error that it mistakes the refusal to speak for wisdom. This happened to science, or to much of it, during that long century when to say "caused" was to confess oneself unrigorous, when causation became the opium of the unsophisticated, and correlation—chaste, measurable, stripped of all presumption—took its place at the altar. We called this progress. We called it rigor. It was, perhaps, a kind of shyness. Pearl's work—and here one must linger over what seems at first merely technical—amounts to something more interesting than a rehabilitation. He has not simply restored causal language to respectability. He has shown that causation, when treated with sufficient care, with sufficient *transparency* about one's assumptions, need not be the enemy of precision. The diagram—this is the heart of it—becomes a way of making one's ignorance visible. Before the data can speak, we must *first* declare what we believe about the world. We must commit ourselves. But here is where the matter grows troubling, and where intelligence itself—that shimmer we are always chasing—becomes difficult to locate. ## The Transparency That Does Not Guarantee Truth There is a seductiveness in explicitness. When a researcher lays out their causal diagram with perfect clarity—every variable named, every arrow justified—we experience something like relief. We can *see* the thinking. We can audit it, critique it, understand exactly where the assumptions lie. This feels like honesty. This feels like we have come closer to knowing. And yet. A model can be completely transparent and completely wrong. The researcher might be, in fact, *more* wrong for having been so clear about it. There is something almost more deceptive about a false assumption that has been rendered explicit than about one that remains vague and trembling in the background. When we articulate something, we give it a kind of authority. We make it *seem* settled. Consider: a researcher studying intelligence itself—that most elusive and culturally saturated of concepts—might draw a diagram showing how a particular test measures "general cognitive ability," and that ability "causes" success in school. The diagram is clear. The assumptions are stated. The mathematics that follows, given that diagram, are provably correct. But the diagram itself—where did it come from? What work was it doing *before* it became a diagram? What alternative architectures of the mind were quietly discarded in the process of choosing this one? Intelligence, after all, is not a thing that sits in the world waiting to be measured. It is a story we tell about certain kinds of thinking, and that story is never innocent. It carries within it—embedded like light in crystal—assumptions about what matters, what can be counted, what ought to be valued. The diagram appears to be a mirror held up to nature. But it is more like a portrait, and every portrait is also an omission. ## The Social Dimension, or What the Diagram Cannot Hold This becomes urgent—becomes almost unbearable—when we consider the social. Intelligence research has always had a social life. It has always moved out from the laboratory into the world, where it became a tool for sorting, for justifying hierarchies, for deciding who was fit and who was not. That history is not separate from the science of intelligence; it is woven into it. The very concepts we use—the very idea of measuring intelligence as a single, heritable, stable quantity—emerged from a social need to rationalize existing inequalities. Now, when Pearl's framework is applied to questions about intelligence, about its causes and consequences, we face a particular difficulty. The diagram must include not only neurological or cognitive variables, but also the social ones: family background, educational access, wealth, discrimination, opportunity. But here the diagram itself becomes a problem in a new way. Who decides what belongs in the diagram? Who decides what counts as a "cause" and what as a "confounder"? These are not merely technical questions. They are political questions wearing the clothing of mathematics. Consider a researcher investigating whether intelligence "causes" income. The causal diagram must include not only intelligence itself, but the social structures through which intelligence gets rewarded or ignored. Yet those structures are not discrete variables that sit neatly in boxes. They are vast, historical, changing. They operate through mechanisms that are not measurable in the usual sense—through bias, through expectation, through the accumulated weight of centuries. How does one place on a diagram the fact that the *same* intelligence, demonstrated by a person from a marginalized group, may be read differently, may be rewarded differently, may be permitted to unfold differently? The diagram assumes a kind of stability, a kind of separability. It assumes that we can isolate variables, that we can say: *this* part is intelligence, *that* part is social context, and they interact in these specific ways. But the social and the cognitive are not separate. They are not things that interact at the boundaries. They are, at some fundamental level, inseparable. The very capacity we call "intelligence" is always already social—shaped by language, by culture, by what one has been invited to think, by what kinds of thinking have been made available. ## The Question of Authority There remains, then, a deeper question about knowledge itself—about what it means to "know the cause of something." When Pearl's tools are used correctly, they produce results that are provably correct *given the diagram*. This is a form of validity: internal consistency, logical coherence. But validity of this kind is not the same as truth. A provably correct derivation from a false premise is still a derivation from a false premise. The question of who decides what the diagram is—who has the authority to say what variables matter, what causal structures obtain, what counts as rigorous enough to be included—this is a question that cannot be answered by mathematics alone. It cannot be answered by transparency alone. It requires judgment. It requires wisdom. It requires, one might say, a certain kind of intelligence. But intelligence of what sort? Not the kind that can be measured and placed in a diagram. Rather, the kind that recognizes the limits of diagrams, that can hold in mind simultaneously the precision of a mathematical tool and the irreducible complexity of the thing it purports to represent. The kind of intelligence that says: *I am being clear about what I assume, and I am doing this not because my assumption is certainly true, but because you deserve to know what I am assuming, so that you can decide whether to trust me.* This is a more modest intelligence than science sometimes claims. It is also, perhaps, a more honest one. ## Toward a Different Knowing What, then, does it mean to know the cause of something? Perhaps it means something less than we have supposed. Perhaps it means: to be able to describe a plausible mechanism, to trace a path of influence, to say how one thing might lead to another—but always with the qualification that other paths exist, that the world is more intricate than our diagrams, that the very act of diagramming involves a kind of violence, a kind of cutting away. To know the cause is not to possess a truth that sits outside interpretation. It is to participate in an ongoing conversation about what we believe to be true, and why, and what we stand to gain and lose by believing it. This seems particularly important when the question is intelligence—which is to say, when the question is, indirectly, about *who we are* and *what we value*. In such cases, transparency is not the end of the matter. It is only the beginning. We must ask not only: is this diagram clearly drawn? But: who drew it? Toward what end? What lives does it touch? What possibilities does it open, and what does it foreclose? The social dimension of intelligence research cannot be eliminated by rigor. It can only be acknowledged, examined, held in view. We can make our assumptions explicit—we must—but we cannot escape the fact of assumption itself. We remain, always, inside the human world, trying to understand the human mind, using tools and concepts that the human world has created. We are not observers standing outside. We are participants, and that participation shapes everything. Perhaps the deepest intelligence lies in knowing this—in recognizing that the most rigorous science is the science that remains aware of its own limits, that does not mistake clarity for truth, that remembers always that the diagram is not the territory, and that some of the most important territories cannot be drawn.