(The difficulty arises if UDT B reasons logically that there should not logically exist any copies of its current decision process finding themselves in worlds where A is dependent on its own decision process, and yet A defects. I'm starting to think that this resembles the problem I talked about earlier, where you have to use Omega's probability distribution in order to agree to be Counterfactually Mugged on problems that Omega expects to have a high payoff. Namely, you may have to use A's logical uncertainty, rather than your own logical uncertainty, in order to perceive a copy of yourself inside A's counterfactual. This is a complicated issue and I may have to post about it in order to explain it properly.)