Incompatibilists generally demand the power to A or not-A, where A-ing is some broadly conceived action (a stealing, a helping of an old lady across the road). It is virtue of this alternative possibility that the agent is supposed to be morally responsible. This thought has been subject to major objections: (1) if it is possible for the agent to A or not-A, at t, with everything about the agent held fixed, then isn't it just a matter of luck whether the agent As or not-As? (2) don't Frankfurt-style cases show that agents can be morally responsible despite lacking these alternative possibilities?
Steward's response is to concede both points. If agents need to have alternative possibilities like these, she thinks, in order to be morally responsible for their actions, then we should understand these possibilities along the lines of some compatibilist proposal. But we can be morally responsible agents only if we are agents at all, and we can be agents only if determinism is false.
Agency, she argues, is the power to settle how certain things will be. A mere one-way power is not the power to settle how things will be because if I have such a power, and I am determined to exercise it, then how things will be is already settled prior to my action. So agency requires the exercise of a two-way power: to A or not-A. But because what is at stake is agency, the A-ing in question here need not be some broadly conceived action. Instead, we are agents if we exercise a two-way power over some finely individuated action tokens; eg., to A at t, rather than at t1 (where t and t1 are very close in time), or to A by means of one's left hand rather than one's right, or by flexing at the wrist, rather than an at the elbow, and so on. Agency incompatibilism is incompatibilism with regard to these finely individuted action tokens; Steward combines it with compatibilism with regard to broadly individuated actions.
This allows her to respond to the Frankfurt-style case. Even if, due to the presence of a counterfactual intervener, I do not have the two-way power to A or not-A, for some broadly individuated A, I do not lose the power to A or not-A where A-ing is finely individuated. One way to think of this view, in fact, is as claiming that though flickers of freedom may not ground moral responsibility, they can ground agency. It also allows her to reply to the luck argument. She claims that it is more plausibly up to us whether we A or not, for finely individuated A-ing, than for broadly individuated A-ing.
How should we respond? I am not convinced that whether we A rather than not, for finely individuated A-ing, is up to us in a way that renders it invulnerable to luck (it is just as true for finely individuated A-ing that in some possible worlds we A and in some we do not, and nothing about us explains this fact). I do think, though, that the view avoids the luck objection, just because finely individuated A-ing is essentially trivial: it fails to be significant enough to count as lucky.
The thought that some kind of significance condition is required for luck is the least controversial element of extant analyses, defended by Pritchard, Coffman, and me. So it is hard to see how we can develop an account of luck on which these flickers of freedom are lucky, but are trivial. But surely this is no way to avoid the luck objection! If these flickers are trivial, they may not be lucky, but nor do they do any real work.
That is, it is hard to see how I lose anything of significance in losing the power to scratch my nose by flexing at the wrist rather than at the elbow. Indeed, it is hard to see how such a power could arise in evolutionary history (here it is important to stress that Steward thinks that animals are agents too). Evolution is insensitive to this power, since it is only broadly individuated actions that are of adaptive significance.
So a dilemma: either A-ing is broadly individuted, and the view falls to luck. Or it is finely individuated and it fails to buy us anything of significance (and it is mysterious how we came to have this power). If agency requires such a trivial power, why should we care whether we are agents or not?