How Does Kant Respond to Hume’s Problem of Causation?

This is the fourteenth article in the series Understanding Philosophy: Figures, Problems, and Ideas. The tenth article explained why Hume thinks experience supplies constant conjunction without necessary causation. This article examines how Kant accepts Hume’s criticism while transforming causality from a habit formed by experience into an a priori condition under which experience itself is possible.

Hume’s problem is powerful because it is simple. We see one event follow another, but never see a bond saying that the second must follow. Past repetition cannot logically guarantee the future. Causal belief therefore comes from custom rather than rational insight into objective necessity. If this conclusion stands, laws of nature may be exceptionally reliable, yet it becomes difficult to explain why scientific judgement takes a universal and necessary form.

Kant accepts that Hume has established a crucial point: causality cannot be extracted from experience. However many sequences we observe, experience can only add further instances. It cannot suddenly generate necessity. Kant refuses, however, to conclude that causality is therefore merely habitual. He reverses the direction of the question. Perhaps the category of causality is not learned from already constituted objects of experience. Perhaps it must operate before a manifold of sensations can become objective experience at all.

The answer belongs to what Kant calls synthetic a priori judgement. A causal principle is not analytic, because the concept of an event does not contain, simply by definition, the concept that it has a cause. Yet the principle claims a universality that goes beyond any collection of particular observations. If causal order is valid, it must extend knowledge without depending upon inspection of every possible event. The question is how such knowledge is possible, not how it can be generalised from repetition.

Kant’s central thought is that experience does not arise through the automatic accumulation of sensory materials. A sound, a flash, a change of temperature and a movement may enter consciousness successively without yet forming events in one shared world. The subject has to synthesise the manifold of representation, identifying objects, temporal order and relations. Categories of understanding are not labels added after experience is complete. They are basic rules through which synthesis occurs.

Causality is especially connected with objective temporal order. We can look first at the roof of a house and then at its windows, or reverse the order. A change in the sequence of observation does not mean that the parts of the house have undergone an event. When a boat moves downstream, however, its upstream and downstream positions cannot be reversed within the event itself. To say that the boat has actually moved, rather than that the order of my attention has changed, is to understand the later state as following the earlier according to a rule.

The rule does not provide ready-made knowledge of a particular cause. The category of causality cannot tell us whether fire will melt a particular material, and it cannot replace experiment in discovering the mechanism of a disease. Its demand is more basic. If a change is to be experienced as an objective event, the sequence of states must be capable of falling under a rule that determines their order. Particular rules are discovered in inquiry; the requirement that events be causally orderable is a condition under which inquiry can possess objective events.

This is the deepest disagreement between Kant and Hume. Hume begins with repeated complete events and explains how expectation forms. Kant asks how a subject can first experience a complete event and its repetition. Without unified objects, determinate time and rule-governed relations, we could not know that two perceptions belonged to the same kind of event. Habit presupposes that experience already has a recognisable structure, and that structure cannot in turn be explained entirely through habit.

Kant does not mean that an individual consciousness deliberately projects causality onto the world. The categories normally operate without becoming an object of attention and are not selected by personal choice. They belong to common conditions of human experience and can therefore support objectivity. Objectivity no longer means that cognition has escaped every contribution from the subject. It means that an object is unified according to rules available to any possible human experiencer. Subjective conditions and objective knowledge are not simply opposites.

Kant’s transcendental deduction is intended to establish the legitimate application of the categories. It does not trace the psychological or historical origin of a concept. It asks by what right that concept applies to objects. The answer is that without the synthetic unity supplied by the categories, representations could not belong to one unified consciousness and could not form knowable objects. Causality is valid not because we see inside things in themselves but because it participates in the constitution of possible objects of experience.

The argument also imposes a strict limit. Categories have cognitive validity only for appearances within possible experience. They cannot be extended unconditionally to things in themselves. From the principle that every empirical event has a cause, we cannot immediately infer that the world as a totality has a first cause beyond experience. Traditional metaphysics repeatedly projects concepts valid within experience beyond experience’s boundary. Kant’s response protects natural science while restraining such projection.

Hume explains the feeling of necessity through a transition produced in the mind by repetition. Kant emphasises the normativity of causal judgement. When we claim that two events are causally related, we do not merely report what we have become accustomed to expect. We assert that under relevantly similar conditions the events should be connected through a shareable rule. An individual expectation can be mistaken, and further evidence can correct a causal judgement. The demand for an objective rule allows us to distinguish psychological association from valid causal explanation.

Kant does not, however, establish the necessity of each particular scientific law. The causal principle requires events to have rule-governed connections, but it does not determine which forces and laws exist in nature. The specific form of Newtonian mechanics still depends upon theory and observation. Later scientific revolutions have shown that laws treated as fundamental in one period may be revised. Kant seeks to protect the lawful intelligibility of experience, not to provide an eternal guarantee for every current theory.

His solution also faces challenges. If the categories belong to human cognitive structure, how can we know that every possible rational being must experience in the same way? Could cognitive science or cultural difference alter the status of the categories? Do probabilistic relations in quantum theory undermine a traditional conception of cause? Such questions lead later thinkers to understand the a priori either as a completely fixed universal form or as the revisable conditions under which a particular practice of knowledge can operate.

Even with these difficulties, Kant’s transformation remains important. Philosophy no longer has to choose between two options: causality as an observable force in objects or causality as an association inside an individual mind. A third possibility appears. Causal structure belongs to the conditions under which objects can become objects of experience. It is neither private invention nor mysterious insight into things in themselves, but a basic way in which subject and object become related within experience.

This changes the concept of knowledge. Cognition is no longer a passive copy of a world already organised in every respect. Through space, time and the categories, the subject participates in the structure of experience. Participation does not mean creation of all content. Sensory material cannot be decided at will, and particular laws must be found through public inquiry. Kant tries to preserve a division of labour between what experience gives and what a priori organisation makes possible.

Kant responds to Hume not by restoring the old metaphysical causal power Hume dismantled, but by relocating the source of necessity. Experience cannot hand necessity to us as a sensation. Yet if experience is to be experience of objective events, it must be organised by rules capable of determining temporal order. Causality is not merely a summary produced after observation. It is a condition under which observation can refer to a world of events.

The disagreement between the two philosophers remains alive. Hume warns us not to mistake expectation for a guarantee supplied by reality. Kant warns us that experience is not an unstructured collection of sensations. Causal knowledge requires evidence, but also rules through which evidence can be understood as relations among objects. Holding these levels together prevents scientific law from being inflated into an absolute viewpoint without reducing objective judgement to long-established habit.

If Kant’s answer is kept strictly separate from my further judgement, his own argument remains centred on the transcendental unity of the subject. The category of causality has objective validity because it belongs to the conditions under which possible human experience can become unified experience of objects. It would be an anachronism to rewrite this as the claim that any rule-governed system therefore possesses a Kantian transcendental structure.

What I retain is a more basic insight: knowledge is not the unconditional reception of data. For experience to form relations among objects, organisation, rules and constraints are required. My extension is that these conditions need not all be concentrated inside one human knowing subject. Scientific institutions, measurement procedures, databases and artificial intelligence can distribute parts of organisation, validation and error correction across a system. I therefore keep Kant’s question about conditions of formation while broadening where those conditions may be realised. That is my conclusion, not Kant’s.