PAP: An agent is responsible for an action if and only if the agent can act otherwise with respect to that action.
The reason for doubt about PAP is the Frankfurt-style counterexample, which was first put forward in Frankfurt's Alternative Possibilities and Moral Responsibility. John gives an example of a Frankfurt-style counterexample in his post. If you think, as many people do, that Frankfurt-style counterexamples refute PAP, then you will want to find an alternative account of moral responsibility and perhaps an alternative account of free will.
In the second he proposes the first elements of a US account of LFW. his suggestion is as follows:
An agent is free with respect to some action only if (1) the agent responsible for the action is the original cause of the action; (2) the causal sequence that causes the action to occur cannot originate outside the agent.
I'm all for LFW, but I think US accounts are untenable. My reasons go back to the very beginning: I deny that Frankfurt-style counterexamples tell us anything about PAP. In my view they tell us that the model of free will that most people work with is simplistic. For instance, it is common to conflate two distinct elements of free will: free decision and free choice. The first, the non-coerced conclusion of practical reason, is not affected one way or another by Frankfurt-style counterexamples. However, both elements presuppose alternative possibilities. All Frankfurt-style counterexamples really show, then, is that you can suppress free choice without suppressing the alternative possibilities presupposed by free decision. Indeed, most Frankfurt-style counterexamples seem to depend crucially on this fact: the original characterization of the scenario is in terms of alternative possibilities. And it is unlikely one could build a Frankfurt-style counterexample that would eliminate the alternative possibilities of free decision: any such scenario would be straightforwardly deterministic, since it would effectively mean that Jones's thoughts as well as his choices are rigged. What Frankfurt-style counterexamples are picking up on is not that PAP is false but that there is more than one point where alternative possibilities can enter into the picture. (I think there are lots of other problems with Frankfurt-style counterexamples, since I think that people have been insufficiently critical in their acceptance of them, but I won't go into them now. Suffice it to say that I think them a laughably weak basis for denying PAP. You can see a small handful of my other objections to Frankfurt-style cases here.)
So I am unconvinced by attempts to claim that PAP is not a necessary condition of LFW or moral responsibility. Even if it weren't, however, I don't think US accounts of LFW (or moral responsibility) will work. The big issue, of course, is what is meant by something's being the 'original cause' of an action. One way of glossing it is to read it as simply first cause. On such an interpretation, the US account would read:
US(1): An agent is free with respect to some action only if (1) the agent responsible for the action is the first cause of the action; (2) the causal sequence that causes the action to occur does not begin outside the agent.
I think US(1) is obviously false. It is certainly false on any view that holds that nothing can happen without God's permission; and it's clearly the case that the causal sequence always in our ordinary experience begins outside the agent. The causal sequence for any action includes the sine qua non causes. For instance, all my free actions presuppose that I am alive; this life has causes that are certainly outside me (the sun, the air, etc.).
But sine qua non causes are necessary conditions. Perhaps the idea is that of sufficient conditions:
US(2): An agent is free with respect to some action only if (1) the choice of the agent responsible for the action is the sufficient condition of the action; (2)there is no sufficient condition for which the agent is not responsible.
But US(2) seems to imply PAP. If there is no sufficient condition other than the agent's choice, that means that the conditions of the choice itself cannot determine the choice to one alternative (otherwise they would also be sufficient conditions) unless they are conditions due to the agent, and those conditions cannot determine the choice to one unless they are due to the agent. Since we are finite agents, the chain cannot infinitely regress; we must eventually reach a point at which there are no sufficient conditions other than the agent's own choice (the so-called 'self-forming action'). In such a case, the possibilities cannot be determined to one by any conditions (otherwise these conditions would be sufficient conditions for which the agent is not responsible), and the choice will therefore take place in an environment in which there are alternative possibilities. If this is true, however, then PAP is vindicated, since there will be no case of responsibility that does not involve alternative possibilities, somewhere along the causal chain. If PAP is untenable, so is US(2).
Is there another account of US that might be given? US can't be formulated in terms of necessary conditions, since then it is obviously false. If we formulate it in terms of sufficient conditions, it concedes that there are alternative possibilities.