Scholars have almost universally found fault with the argument of Kant’s deduction of the categorical imperative in Groundwork III. No one seems to endorse it without reservation; and only a few think that it is in any way successful. There is even evidence, quoted above (Kant 1997b, 41-2/5:47), that Kant himself came to see that an argument such as he had attempted in the Groundwork could not succeed.

The deduction’s framework. The prospect of a hidden circle in Kant’s reasoning in Groundwork III preceded the argument of the deduction. It appeared as though freedom had been assumed in order to secure the possibility of moral action in the natural world. Yet it also seemed as though freedom had been derived from the possibility of moral action. But that would be circular reasoning. So an independent argument for freedom would be required. Only then, with the reality of freedom established, could the close relation between (positive) freedom and the principle of autonomy establish the validity of the categorical imperative. The deduction seems to supply the argument required. Freedom is established in the deduction not through the concept of morality, but by reference to rational beings’ membership in an intelligible world. Then that world’s being the “ground” of the sensible world effectively subjects sensible rational beings to the categorical imperative. Hence, the crucial role of the deduction seems to be to supply a non-moral basis for freedom. That way, the reality (validity) of the moral principle can follow from the reality of freedom, without circular reasoning. Yet skepticism reigns over Kant’s apparent non-moral argument for freedom of the will.

Free will in the intelligible world. The deduction supposes, first of all, that rational beings in an intelligible world would enjoy freedom of the will. But this seems to be a synthetic a priori proposition itself, and so one that would require a prior deduction, which Kant does not supply (Paton 1948, 245). And even if he may have proved in the first Critique that rational beings are members of a timeless, intelligible world, this falls short of proving that they would have free will in that world, so that they would be subject to the moral law (Ameriks 2003, 171-2). There may also something problematic about one world’s being the ground of another, or so Allison claims. Suppose that God were in the same way the ground of the natural world, and so of its laws—a proposition that enjoys wide acceptance. Kant would be forced to conclude from this prospect that moral agents are heteronomous. So it does not seem to follow that, if something else is the ground of the sensible world, and of its laws, then the moral law is a law of autonomy (Allison 2011, 335-6). Even so, ignoring all of the preceding difficulties, Kant’s appeal to the intelligible world as the ground of the laws of the sensible makes it hard to see how any rational agent could ever disobey that law in the sensible world (Timmermann 2007, 143).

The deduction of freedom. Is it possible that the standard reading of the Groundwork deduction is mistaken? Could the point have been instead to show how morality provides an argument for freedom, and then from there to show the validity of the categorical imperative? In an earlier passage Kant made the point that proof of a synthetic a priori proposition requires reference to some third idea that effectively joins its subject and predicate. He wrote:

What this third cognition is, to which freedom points us and of which we have an idea a priori cannot yet be shown here and now; nor can the deduction of the concept of freedom from pure practical reason, and with it the possibility of a categorical imperative as well, as yet be made comprehensible.... (Kant 1997a, 53/4:447, emphasis added).

The above suggestion about the point of the deduction gains some plausibility when it is acknowledged that Kant sometimes distinguished between the moral law and the categorical imperative. Suppose he assumed that subjection to the moral law is known intuitively, through what in the Groundwork is called “common rational cognition of morals.” Suppose also that this is why he wrote in the Critique of Practical Reason that “the objective reality of the moral law cannot be proved by any deduction”; and that it is given as a “fact of pure reason.” The idea of freedom, and of membership in an intelligible world, could then be deduced from intuitive consciousness of moral obligation; and from those ideas it would be possible in turn to deduce the validity of the categorical imperative. This procedure seems to describe what Kant meant when he referred to “the deduction of the concept of freedom from pure practical reason, and with it the possibility of a categorical imperative as well.” It does not entirely evade the circle, however; though there may be reasons why the circle interpreted in this way turns out not to be vicious (cf. Henrich 1998, 329-35; and Beck 1960, 173-4).

Autonomy and the intelligible world. The preceding suggestion will probably not satisfy those raising problems, above, about free will, autonomy and membership in the intelligible world. But I think those worries can be addressed, even apart from the “deduction-of-freedom interpretation.” Let us begin with Allison’s complaint that another world’s “grounding” the laws of this, sensible world would not guarantee moral autonomy. It is clear that it would if this world were the appearance of that world, and if the members of that world were members of that community through their own free will. To the others skeptical about free will in the intelligible world, it need only be pointed out that membership in such a world community would imply action, hence willing. A world requires connections among its members who, if they are free, connect themselves into a whole. Finally, as Kant explains in the Groundwork deduction, those who may be members of the intelligible world only cannot deviate from the moral law. But there is no inconsistency in the idea that those who are members of both worlds, and so who are subject to both the moral law of autonomy, and the natural laws of inclination, can deviate from the moral law. (On this point see Kant 1998, 52-73/6:29-53.)

Intelligible-world skepticism. The distinction between appearances and things in themselves implies at least that something exists that is unconditioned by space and time, and is that of which the sensible world is the appearance. Kant assumes, without any argument, that what exists apart from space and time is plural. I do not think this plurality is ever established, nor that it can be established by reasoning and argument. As far as I am able to know, there is only a single thing in itself, and it is I. In other words, there is no possible refutation of noumenal solipsism.

A new deduction? Yet I cannot get myself to believe that I am all there really is, and that the sensible world is therefore the appearance of myself. I cannot help believing that the reality behind the sensible world is plural, and that it includes me and every other rational being. For this reason, I take it as a “fact” that there is an intelligible world, and that it is populated by rational beings like myself, who act freely there to establish the connections required for the unity of that world, without which there would be no unity and so no lawfulness in the natural world. Extending my insight into the intelligible world at least this far, I do not then see why the connections we willfully establish in common are not moral relations. That is why I take it also as a “fact” about the sensible world that rational beings here owe each other duties of justice and virtue. Of course, having no further insight into the intelligible world than these “facts,” I cannot determine anything about the moral relations of intelligible and freely acting rational beings than conditions of their equality and reciprocal action. Hence, these can be the only rational conditions that could determine the contents of duties of justice and virtue. And from these conditions it is not hard to infer the content of morality as expressed in the formula of universal law of the categorical imperative. The preceding seems to me therefore to be a satisfactory, “fact-based” deduction of that principle’s validity.