Second, recast that maxim as a universal law of nature governing all rational agents, and so as holding that all must, by natural law, act as you yourself propose to act in these circumstances. Third, consider whether your maxim is even conceivable in a world governed by this law of nature. If it is, then, fourth, ask yourself whether you would, or could, rationally will to act on your maxim in such a world. If you could, then your action is morally permissible. Throughout his moral works, Kant returns time and again to the question of the method moral philosophy should employ when pursuing these aims. A basic theme of these discussions is that the fundamental philosophical issues must be addressed a priori, that is, without drawing on observations of human beings and their behavior. Once we “seek out and establish” the fundamental principle of morality a priori, then we may consult facts drawn from experience in order to determine how best to apply this principle to human beings and generate particular conclusions about how we ought to act. Kant's insistence on an a priorimethod to seek out and establish fundamental moral principles, however, does not always appear to be matched by his own practice. The Groundwork, for instance, makes repeated appeals to empirical facts (that our wills are determined by practical principles, that various motivations are variable in producing right actions, and so on). Later ethical works rely even more heavily on empirical generalizations. Kant did not take himself to be employing these assumptions in seeking out and establishing the fundamental moral principle, only in applying it to human beings. Nevertheless, it is not always easy to tell whether Kant's arguments gain their plausibility only by relying on ideas established by observations of human being and the world they inhabit.
Kant's example of a perfect duty to others concerns a promise you might consider making but have no intention of keeping in order to get needed money. Naturally, being rational requires not contradicting oneself, but there is no self-contradiction in the maxim "I will make lying promises when it achieves something I want". An immoral action clearly does not involve a self-contradiction in this sense (as would the maxim of finding a married bachelor). Kant's position is that it is irrational to perform an action if that action's maxim contradicts itself once made into a universal law of nature. The maxim of lying whenever it gets what you want generates a contradiction once you try to combine it with the universalized version that all rational agents must, by a law of nature, lie when it gets what they want.
Here is one way of seeing how this might work: If I conceive of a world in which everyone by nature must try to deceive people any time it will get what they want, I am conceiving of a world in which no practice of giving one's word could ever arise. So I am conceiving of a world in which no practice of giving one's word exists. My maxim, however, is to make a deceptive promise in order to get needed money. And it is a necessary means of doing this that a practice of taking the word of others exists, so that someone might take my word and I take advantage of their doing so. Thus, in trying to conceive of my maxim in a world in which no one ever takes anyone's word in such circumstances, I am trying to conceiv