# On Knowing Causes, and the Diagrams We Pretend We Did Not Draw Intelligence is commonly defined as the ability to solve problems. This is nonsense. A trained dog solves problems. A chess computer solves problems better than any human being. Yet we do not call these things intelligent in the way we mean when we speak of an intelligent man. Intelligence, I suspect, is the ability to see what is actually there. Not what we wish were there. Not what our theory says should be there. What is actually there. This is harder than it sounds, and rarer. For a hundred years, scientists studying the mind adopted a peculiar strategy. They would not speak of causes. They would speak only of correlations, of variables that moved together, of statistical associations. This had the appearance of humility. It was in fact an elaborate escape. By refusing to name causes, researchers avoided the necessity of being wrong about causes. They could hide behind equations. Then Pearl came along and said: you have been drawing diagrams all along. You simply refused to look at them. Every statistical model contains assumptions about causation. Every regression contains hidden arrows. You might as well draw them openly. This was useful. It is better to see what you assume than to pretend you assume nothing. But Pearl's tools—and here is the trap—can only work if you get the diagram right. The diagram must be assumed before the data can speak. And here is what no one wants to say clearly: *the data cannot tell you whether your diagram is correct*. This is the crisis of modern intelligence research. We build models. We make them explicit. We run them against data. We get numbers. The numbers are often very clean. They look like truth. But they are only the truth of the diagram we drew before we looked at anything. We have merely confirmed our assumptions with mathematical elegance. A researcher can be completely transparent about his model and completely wrong. Transparency and validity are not the same thing. I have seen papers with the clearest diagrams, the most rigorous equations, the most honest acknowledgment of assumptions—and the conclusions were simply false. The researchers were not liars. They were not even careless. They had simply drawn the wrong picture. So the question becomes: who decides what the diagram is? Not the data. The data is mute until you speak to it. It cannot defend itself against a well-drawn lie. Not the equations. The equations only work within the logic of the diagram. The answer, uncomfortable as it is, must be: the person who draws it. The researcher. The human being with his particular experience, his prejudices, his access to the world. This is where wisdom enters. Wisdom is not intelligence. An intelligent man can see the logical structure of a problem. A wise man knows which problems are worth solving and how the answer will change the world. More importantly, a wise man knows the limits of what he understands. He does not mistake clarity for truth. The researcher with wisdom knows this: I have drawn a diagram. It is explicit. It seems to match the data. But I could be wrong in ways that no amount of statistical rigor can prevent. I could have left out a crucial variable. I could have drawn the arrows pointing the wrong direction. I could have mistaken a symptom for a cause. A wise researcher does not boast of his transparency. He worries about it. He asks: what am I not seeing? What would change my mind? What did I assume so early that I no longer notice I am assuming it? This is not paralysis. It is the only kind of rigor that matters. The problem with modern intelligence research—and I mean this precisely—is that it has confused the ability to build explicit models with the ability to understand reality. We have become very good at the first. We have become worse at the second. We draw our diagrams with such confidence, we present them with such clarity, that we forget they are diagrams. We forget they are made by human beings looking at a small corner of the world from a particular angle. The question "What is intelligence?" cannot be answered by building a better model. It can only be approached by a man who has looked at many kinds of intelligent behavior, who has failed at understanding it several times, who has sat with his confusion long enough to see its shape. Such a man will not give you a diagram. He will tell you what he has noticed. He will be suspicious of his own noticing. He will refuse to pretend that clarity and truth are the same. That refusal—that willingness to live inside uncertainty while still thinking clearly—is closer to wisdom than any model can be. And wisdom, finally, is the only form of intelligence that matters.