# On the Nature of Causation in Knowledge, and What It Profits a Man to Understand The human mind, that restless and ambitious faculty, perpetually labours under a singular delusion—that the rigour of its methods may be mistaken for the rigour of truth itself. We have witnessed, in the recent annals of natural philosophy, a most instructive spectacle: the banishment of causal language from the discourse of science, an exile maintained with such severity and for such duration as to persuade many learned men that they had thereby achieved a species of purity. Yet what they called rigour was in truth but the evasion of difficulty; and in their flight from the treacherous ground of causation, they had merely constructed a more elaborate shelter of ignorance, a fortress of negation within which they might dwell in safety. Then came Judea Pearl, bearing the tools of graphical representation and mathematical formalism, and demonstrated—with a clarity that admits of no dispute—that causal language need not be banished to restore it to utility. His diagrams, his calculus, his methods of stratification and adjustment, have restored to science a language that men had possessed since Aristotle, yet had been persuaded to abandon as unbecoming to their elevated station. This is no small matter. For science without causation is but cataloguing; and cataloguing, however minute and however organized, remains forever imprisoned in the particular, unable to speak to the universal condition of mankind. Yet here we encounter a difficulty more profound than any that mere methodological quarrel might produce. ## The Diagram as Assumption, Not Discovery Pearl's great gift carries within it a poison of its own manufacture. His instruments—invaluable as they are—require of us a preliminary act of faith. Before the data may speak, we must first draw the diagram. We must decide, prior to observation, which variables shall influence which; we must establish the skeleton upon which all subsequent inference shall hang. And here we have not gained knowledge; we have merely made explicit the ignorance we began with. The researcher sits before his diagram as Prometheus before the fire—believing himself to have stolen divine knowledge, when he has in fact only rendered visible the particular shape of his own assumptions. He may articulate these assumptions with perfect clarity. He may defend them with elaborate reasoning. He may even subject them to the most searching scrutiny. And yet—and this is the point which cannot be too earnestly pressed—transparency in error is not the cure for error. A falsehood stated plainly remains a falsehood still. Consider the physician who constructs a diagram showing that some treatment causes improvement in his patients. He has been explicit about his model. He has declared his assumptions. He has applied Pearl's methods with technical perfection. And yet if his diagram omits some variable—some aspect of the patient's constitution, some circumstance of their life, some quality of their spirit that he had not the wit to measure—then his entire edifice, however transparent, however rigorously defended, rests upon a foundation of sand. The diagram he drew before the data could speak has determined what the data shall be permitted to say. This is the perpetual tragedy of human knowledge: we cannot know what we do not know. The diagram is not a discovery; it is a confession. It confesses what the researcher believes to be true. But belief, however reasonable, is not truth. ## The Question of Authority Yet there remains a more vexing question still, and one which touches upon matters of moral consequence: *Who decides what the diagram shall be?* The statistician, armed with Pearl's methods, may claim a species of technical authority—that he understands the mathematics, that he can verify the internal consistency of his model, that he can demonstrate its correctness given the diagram he has assumed. This is a narrow and defensible claim. But it is precisely this narrowness which renders it dangerous. For the diagram itself is not a mathematical question; it is a substantive one. It requires knowledge of the domain, understanding of the mechanisms at work, familiarity with the particular circumstances of human life that the abstract variables are meant to represent. Who, then, shall determine the diagram? Shall it be the statistician, whose training has made him expert in the formal properties of inference but who may know nothing of the actual world those inferences concern? Shall it be the subject-matter expert—the physician, the economist, the social investigator—whose knowledge of reality may exceed his mathematical sophistication? Shall it be some committee of the learned, deliberating in the manner of the ancients? Or shall we admit, with the honesty that becomes a man of sense, that the question admits of no answer that does not immediately raise further questions? For in this matter, authority is not a thing to be possessed but a responsibility to be borne. The person who draws the diagram—whether they acknowledge it or not—has assumed a power of considerable magnitude: the power to determine what shall count as a cause, what shall be deemed a consequence, what shall be regarded as merely incidental to the matter at hand. This is not a power that resides in mathematics. It resides in judgment. And judgment, that most necessary of human faculties, is precisely what cannot be formalized, cannot be taught in the seminar, cannot be verified by any diagram whatsoever. ## The Restoration of Wisdom Here, then, we come to the question which the learned have increasingly neglected, to their great impoverishment: *What is wisdom, and how does it relate to the knowledge that our diagrams might afford us?* Intelligence—if I may be permitted to return to the original inquiry—is not synonymous with wisdom, nor should we confound them. A man of intelligence may perceive the relations between things; he may manipulate symbols with great dexterity; he may construct arguments of admirable complexity. But wisdom is a different faculty altogether. Wisdom is the capacity to know not merely what is true, but what is *good*; not merely what causes what, but what *ought* to be done in light of that knowledge; not merely what the diagram permits us to infer, but whether that inference serves the genuine interests of human flourishing. Wisdom requires, first, a knowledge of human nature—not the abstract human nature of philosophical discourse, but the particular, suffering, complicated human nature that we observe in the streets, in the sickbeds, in the courts of law, in the cottages of the poor. It requires, second, a humility before the limits of one's own understanding. The wise man knows that his diagram is incomplete, that his assumptions are fallible, that the world exceeds the categories by which he attempts to comprehend it. And it requires, third—and this is the most difficult requirement of all—a willingness to act despite this uncertainty. For life does not permit us the luxury of perfect knowledge before we must choose. The physician must decide whether to administer the treatment though his causal understanding remains imperfect. The statesman must frame policy though the consequences cannot be fully foreseen. The ordinary person must conduct their affairs though the causes of their condition remain obscure to them. ## The Diagram and the Moral Life This is where the restoration of causal language by Pearl and his successors becomes not merely a technical matter but a moral one. For in restoring to us the language of causes, they have also restored to us the possibility of moral responsibility. If we can speak of causes, we can speak of agency. If we can speak of agency, we can ask: who is responsible? And this question, once asked, cannot be evaded by retreat into methodological purity. The researcher who draws a diagram has implicitly claimed that this diagram captures something true about the world. He has not merely said: "Here is a mathematical structure that is internally consistent." He has said: "This is how things actually work." And if his diagram is wrong—if it omits some crucial variable, if it misrepresents some fundamental relation—then he bears some responsibility for the consequences that follow from his error. He cannot hide behind the transparency of his assumptions. Yet here we encounter an irony that should humble us all. The more carefully we articulate our assumptions, the more we may be tempted to believe them. Explicitness breeds confidence; and confidence, in the presence of hidden ignorance, is the most dangerous of all conditions. We are worse off when we know precisely what we are assuming than when we dimly apprehend that we are assuming something without being able to say what. ## Conclusion: The Limits of Knowing What, then, does it mean to know the cause of something? It means, I submit, to understand the mechanism by which one event produces another; to perceive the chain of dependence that connects them; to be able to say with some assurance that if the cause were absent, the effect would not occur. But this understanding is always provisional, always incomplete, always subject to revision in light of further experience. And who decides what the diagram shall be? The answer is uncomfortable but unavoidable: we all do. The researcher decides when he draws the diagram. The reader decides when she accepts or rejects it. The subject of the research decides when his or her actual life either confirms or confounds the predictions the diagram permits. Authority over the diagram is distributed throughout the community of inquirers, and this distribution cannot be otherwise. What remains, then, is not the pursuit of perfect causal knowledge—that is a phantom, a will-o'-the-wisp that has led many learned men into the marshes. What remains is the cultivation of wisdom: the habit of asking not merely "What causes what?" but "What should I do about what I know?" The habit of holding our diagrams lightly, as useful fictions rather than captured truths. The habit of attending to the particular and the concrete, even as we theorize about the general and the abstract. For in the end, intelligence without wisdom is merely the capacity to err with greater sophistication. And this, I fear, is a capacity we have cultivated to excess.