Kant recognized the prima facie conflict between the idea that the world is causally determined and the idea that we are free agents, but he employed his novel epistemological theory in order to resolve the conflict. Kant's solution is flawed, I think, because his epistemology is flawed, but his recognition that the fact that we appear to ourselves to be free does constitute indisputable evidence that we are free is a key insight in the history of the development of the debate about free will. I'll discuss his solution here, which will necessitate diving into his strange vocabulary of technical terms. Kant was an obsessive system builder, and the interrelations between terms in his system is complicated and messy, but I'll do my best to untangle the mess enough so that his version of free will is explicable.
In a previous post I discussed a shorthand way to understand some of the key concepts of Kant's Critique of Pure Reason. One of those key concepts is the idea that the mind supplies the framework for understanding the world, and is not merely a passive receptacle of information. Because of this all of our cognitive experiences, including our perceptions, are constructed out of materials received from the world and materials supplied by the mind. These cognitive experiences make up what we can call the "world of appearances." But a consequence of that is that we are screened off from what Kant calls the "thing-in-itself" or the "world in itself." We have no direct access to the world via perception; instead the objects of perception are already modified by our own minds.
This distinction between the world in itself and the world of appearances allows Kant to assert that the world in itself exists, however. Otherwise there would be the absurdity of an “appearance without an object that appears.” Kant also holds that this distinction makes human freedom and universal causation compatible because the will as an object of appearance is subject to causality but the will as a thing-in-itself is not (Bxxv-Bxxvii). Kant can think himself free only on the "critical distinction;" this critical distinction requires that the scope of pure reason be limited. And since morality requires freedom, morality also requires this circumscription of reason. He claims that other concepts necessary for the practical use of reason (God and immortality, for example) also cannot be put to use without reason being deprived of its “pretension to extravagant insights.” Kant had to “deny knowledge in order to make room for faith” (Bxxviii-Bxxx). This denial of knowledge results from the fact that Kant takes as an axiom the existence of moral responsibility, which means he has to assume that freedom of some sort exists. Kant rightly claims that the idea of moral responsibility is closely connected with the idea of freedom: if we ought to have done something, then it makes sense that we were free to have done it. He never seriously entertains the thought that moral responsibility does not really exist.
The ideas of freedom and responsibility are understood as being part of Kant's broader theory of causality. Causality can be thought in reference to either nature or to freedom. Natural causality is the connection between a state of affairs and the states of affairs that preceded it according to a rule or law of some kind. This cause has to occur in time; an atemporal cause of events is not possible. Causality derived from freedom occurs when a faculty begins a causal series or brings about another state of affairs from itself without being preceded by a cause. This transcendental notion of freedom is the ground for the practical notion of freedom. Practical freedom is the power of choice exercised without being necessitated by sensibility. The power of choice can be affected by sensibility, but if it is to be free (as opposed to mere animal choice) then it must not be determined by it (A532-4/B560-2). In other words, the final arbiter of whether an agent acts is the agent herself, not some series of non-agential causes.
It is clear that universal natural causation would mean that every occurrence, including the exercise of choice, would be fully determined, eliminating transcendental freedom and practical freedom. Kant argues that the solution to this problem lies in properly formulating the notions of causation and freedom within the context of Transcendental Idealism. For if we do not make the distinction between appearances and things in themselves, and appearances are things in themselves, then the conditions for an event will always occur in the same temporal, causal series as the conditioned event. In that case, freedom is impossible, because under natural law all the conditions of an event are contained within the series of appearances of which it is the effect. But if we distinguish between them the the question becomes “whether freedom is possible anywhere at all.” Because under this consideration, the appearances have their grounds in things that are not appearances (“intelligible causes”). These intelligible causes lie outside the series of appearances, even though their effects can be seen in the series as the outcome of other appearances via empirical laws (A534-7/B562-5). The will capable of practical freedom must be one of these "intelligible objects."
The intelligible object is not encountered in experience, but we can think this object as both intelligible in action and as a cause of effects that are sensible in appearance. We can attribute causality to this object; there is nothing that prevents us from doing so. On this view of the causality of intelligible objects there is thus a dual character of objects of sense: every subject has an empirical character, under which it is determined by preceding states of affairs and determines proceeding states of affairs through causal laws and it has an intelligible character, under which it is not subject to the conditions of sensibility. In its intelligible character the subject is thought of (in general terms) as free of the conditions of sensibility and thus is not determined by appearances. This subject can thus act in the world and demonstrate both freedom and nature, at the same time and without contradiction (A538-41/B566-9).
It is a law of nature that everything that occurs has a cause that precedes it in time, and thus that all events are part of the empirical order. This law is a firm law of the understanding which does not permit exceptions in the world of appearance. Within this orderly presentation of all appearances, each cause is itself an effect of a prior cause and so it is not expected that there would be within a time series an “original action.” Kant claims, however, that even though empirical causality dictates that every appearance have an appearance as a cause it is possible that empirical causality “could nevertheless be an effect of a causality that is not empirical, but rather intelligible” (A542-4/B570-2).
The principle of causality is necessary for the ordering of appearances by the understanding, but this function is not impaired by supposing there is among natural causes an intelligible causality that does not rest on empirical principles. The intelligible cause “does not touch the empirical questions;” the effects of the intelligible cause in nature are perceptible, but must be entirely explicable by reference to empirical causes (A544-6/B572-4).
Kant applies this principle to human experience. A human agent stands in the world of appearance, and as such must possess an empirical character. But this human agent also knows himself (through pure apperception) to be an intelligible object in the sense that he is not fully determined in his actions by any appearances. Instead, his actions are determined by reason, which is why we can assert imperatives as guides to human action. The “ought” of an imperative suggests that the ground of an act determined by reason is a concept, whereas the ground of an natural act is an appearance. This ought cannot be derived from any combination of sensory inputs or natural grounds (A546-8/B574-6).
But assuming that reason possesses causality in the world of appearances, the power of choice according to reason must then have an empirical character, which means human action must follow from appearances as their effect. This means that according to its empirical character there is no freedom, because all human actions would be predictable “if we could investigate all the appearances.” But if we consider the same action in relation to practical reason we can see that it conforms to a different sort of ordering pattern, according to which it can be seen as being undetermined by preceding causes. We are not immediately acquainted with the intelligible cause (thus hiding the true moral nature of actions from us) but can infer that it is efficacious in relation to human action (A549-51/B577-9).
This intelligible condition, freedom, does not operate in time, otherwise it would be part of the series itself and belong to the world of appearances. Accordingly, it is assumed that there must be an unconditioned condition of an empirical series outside the series itself. However, in another sense, this condition does exist in the series of appearances: both the human agent and his power of choice are members of the series. Human reason, however, is not an appearance and is thus not subject to the conditions of sensibility; as a result reason is the “persisting condition of all voluntary actions under which the human being appears.” In its intelligible character reason is atemporal and immediately causes all action; this freedom is negative (free from determination by empirical conditions) as well as positive (able to begin a causal series from itself) (A551-4/B579-82).
As an example Kant analyzes the case of a malicious lie. If we were to investigate the causes of another person’s lying, we would first look into the empirical conditions, in this case types and sources of the character flaw of the offender. We might speculate that poor upbringing, bad temper, carelessness, etc. might have contributed to the act. But we nonetheless find the offender blameworthy, a judgment grounded in the law of reason. Despite the empirical determination of the action we find that the lie can be understood as an act unconditioned by previous states that begins a series of consequences. This judgment relies on the independence of reason from sensibility. Thus, even though we can have no insight into the intelligible character of freedom, we can know that our actions could be free (A554-8/B582-6).
Kant's solution to the problem of freedom is complex, though because it is dependent on his epistemology-driven metaphysics, it's frankly pretty weird. But weirdness doesn't mean it is untrue, so, since it is a version of compatibilism, might it offer us legitimate option avoiding determinism?
The answer to that is "No," I think. Even setting aside an evaluation of his full theory, which I think is problematic, Kant makes two errors that render his account less than coherent. First, Kant takes the lack of knowledge about "intelligible objects" to warrant speculatively assigning whatever properties he wishes to them. But why should we think that merely because we don't know anything about these mysterious objects provides us with the warrant to make any claims about them? Kant should take Wittgenstein's advice--we should keep quiet about whatever we can't talk intelligibly about.
Second, Kant claims that we can assume that intelligible objects exist outside of time and yet still understand them as somehow causally connected to our temporal experience of the world of appearances. But this is unintelligible, even on Kant's own terms. The spatiotemporal structure of our experiences is given a priori, which means it acts as an organizing principle for all experiences and all our cognitions about experience. Which means it is probably not even coherent within his system to talk about objects outside the spatiotemporal schema causing things at all.
In the end, Kant's valiant attempt to rescue freedom from the clutches of determinism falls short, which might make us despair of ever finding a formulation of compatibilism we can live with. Then again, maybe Kant wasn't all that bright after all, so we can just plug away. On to the next one.