Kantian ethics begins with maxims. Maxims are subjective principles of action. In Kantian ethics, one uses the CI-procedure to test whether one should or should not act on the maxim that s/he has formulated.
For example, if I wanted to know whether to act on the maxim, "whenever I can get away with it, I should cheat my business partners," I subject it to the CI-tests (i.e., universalizability and treating humanity as an end in itself). CI1 tells us to "act only on those maxims that you can at the same time will it to become a universal law." I cannot will my maxim to become a universal law because every business person and government official would be suspicious of business activities. We would think that everyone would be trying to get away with something. Since it cannot pass the first CI-test of universalizability, I should not act on the maxim.
But there is a problem lingering in the background over maxims. Maxims can be described in many different ways. Since there are different descriptions of action, one way of devising the above maxim could pass the CI-test of universalizability. If it can pass the first CI-test, then we might have reason to think we should act on that maxim. Given that there are multiple descriptions of the same action, we could think that we should act on it and we should not act on it. This is a contradiction, so we need a way out of the problem.
One way around the problem is to explore action individuation. If actions can be individuated to show that there are different descriptions of action and if these descriptions designate different actions, then the contradiction does not follow. You're acting in each of these situations is different.
This is a preliminary look at a discussion of the problem of relevant descriptions and its relation to action individuation. I have to read Mark Timmons's piece on the problem of relevant descriptions this weekend. After that, I hope to have a better take on the problem and its relation to individuation of action.