Well, you kind of know what you'd like to do, but you also know what he'd like to do, and you know that he knows what you'd like to do, and so you think about what he's going to think you're going to do, so you try to anticipate that by doing something else, except that he probably realizes you're going to try to cheat him, so he's going to try to get you before you get him, lather, rinse, repeat ad infinitum and you're not one step closer to figuring out what to do.
One of the fascinating and paradoxical implications of trying to rationally maximize your own utility function is that, when confronted with certain scenarios, such as the famous prisoner's dilemma, and by virtue of pursuing your own individual interests, you end up screwing yourself more than if you simply decided to cooperate. Oh, but it's never that simple. Anyway, here's a basic introduction to the idea of the prisoner's dilemma:
Now, you may think that because you now understand how this dilemma works, you could beat it if you were to find yourself in such a scenario, but as Dilbert shows, that's not quite right (which is exactly why such scenarios are so interesting and why so many academics in all sorts of disciplines have devoted so much of their time to think about and research such cases):
In political philosophy, one of the most famous instances of a philosopher recognizing the nature of these interactions was Thomas Hobbes, who argued (a few centuries before game theory was actually invented... isn't it awesome how philosophy is so often way ahead of its time?) that the way to solve these problems is to make sure there is some sort of mechanism to enforce cooperation.
Hobbes' particular solution was that all rights should be transferred to a sovereign who would have a monopoly on power (and the violence that could be inflicted on defectors), but as the following absolutely brilliant and gripping game-show example demonstrates, there can be other ways to ensure cooperation:
Was that brilliant or what?!?