In Options and Agency, I defend what I call the Performance Principle:
(PP) If an agent performs an act of type A, then A is an option for S
In a thoughtful and generous review of that book, Sophie Kikkert and Barbara Vetter observe that this principle encounters some difficulties. Consider an unskilled player who hits the bullseye on a dart board. From (PP) it follows that she had the option of hitting the bullseye on the dart board. This is so even though, prior to her throwing the dart, her hitting the bullseye was an exceedingly unlikely event (indeed we may make the probability of her hitting the bullseye conditional on throwing the dart arbitrarily close to 0). It was not, in some intuitive sense, something that was under her control. It was just something that she happened, luckily, to do.
As the authors observe, the claim that this was an option for her appears to conflict with the constitutive connection between options and deliberation. Our options are what we properly deliberate about. But hitting the bullseye was not something the agent should have deliberated about, before the fact. Our deliberations and intentions tend to be, as I emphasize elsewhere in Options and Agency, more coarse-grained than this. So (PP) appears to say that we should deliberate about acts that, in fact, we ought not deliberate about.
Perhaps one can make the theory work with some special pleading for these particular cases. But even if one can do that, the authors point out that a deeper worry looms. Options are supposed to be, on my account, something that is basic and familiar to us. But something having these strange properties – normally coarse-grained, but then maximally fine-grained in cases where we actually act – cannot be something that is basic and familiar to us, even if we grant it to be formally consistent.
I want to pull back and consider the rationale for (PP). (PP) is not motivated, in the first place, by reflection on cases (though I grant that it must ultimately be tested against these). Rather, it is motivated by a deep claim about agentive possibility, which is that it has the features of what is sometimes called an alethic modality. Claims about epistemic and deontic possibility in a certain sense float free of what actually happens. This is clearest in the deontic case: from the fact that someone performs an act, it does not follow that it was permissible that she perform it. Claims about agentive possibility, on the other hand, do seem in some sense so constrained: from the fact that someone performs an act, something does seem to follow about what she could in fact do.
On its classical development, alethic possibility is understood in terms of a schema in modal logic on which the possibility of a proposition is implied by its truth (or, equivalently, on which the truth of a proposition is implied by its necessity). This is the so-called T axiom. Since I, like others, reject the analysis of agentive modality in terms of classical modal logic, I cannot say that agentive possibility obeys the T axiom. So I need a different way of expressing the core truth that agentive modality is in a certain way constrained by what agents actually do. That is where (PP) comes in. It is, as it were, a way of recapturing the T axiom within an act-based (as opposed to proposition-based) framework for understanding agentive possibility.
That said, I grant the force of the objection raised by Kikkert and Vetter. I especially acknowledge their point about fineness of grain. The coarse-grainedness of options is central to many of the arguments of the book, and it is not plausible to make options go suddenly fine-grained in these cases, just to make the (PP) work. So we are left with a challenge: how do we acknowledge the way in which actuality constrains agentive possibility, without admitting the strange claims about options that this statement of (PP) seems to license?
To begin, note that acts in general are not produced fully formed. They tend to occur as the result of a chain of acts, some more fine-grained than others. I do not simply have a cup of coffee. I put the kettle on, pour the water over the grounds, raise the mug to my mouth, and so forth. In the book I attempted to prescind from this aspect of action in order to stay neutral on various views of the metaphysics of action. But this neutrality cannot be maintained. To answer the authors’ challenge to (PP) we need to say at least something about the metaphysics of action.
Let us define an ancestor of an act as follows. If S performs an act of type A, let an ancestor of that act be any act that causes or constitutes the act of type A. For example, if I hit the bullseye, then an ancestor of my act may be an act of the type: throwing the dart thusly. Let us also stipulate that ancestry is reflexive, so that every act is an ancestor of itself.
Note the background ontology here, as in Options and Agency generally, is one on which acts are actual particulars (act-tokens in one standard lexicon) whereas act-types are properties of acts that may or may not be instantiated. Act-types are the objects of deliberation and the relata of options; acts are actual substratum by which the objects of deliberation and options are realized. The performance principle concerns, as it were, the nexus between act-types and acts.
I propose that we then revise (PP) as follows:
(PP*) If an agent S performs an act x of type A, then S performs some act y of type B such that (i) y is an ancestor of x and (ii) B is an option for S
Informally, what (PP*) says is this: if someone does something, then she has the option of doing that thing, or of doing something that causes or constitutes it. (PP*) simply regiments this informal and I think plausible thought into the metaphysics of acts and act-types.
How does this go? Sometimes, we need only appeal one act and one act-type, and the complexities of (PP*) are redundant. I raise my arm. Which is to say I perform an act x of raising my arm. I also perform an act y of raising my arm, and raising my arm is an option for me. That is the case where x=y and A=B.
Other times, the relationship will be more indirect. I perform an act x of hitting the bullseye. But my act was caused by a prior act y of throwing the dart thusly, and throwing the dart thusly was an option for me. Here x and y and A and B are distinct, and we need to appeal to (PP*) rather than (PP) to properly capture the more complex relationship between what is an option for me and what I in fact do.
Thus (PP*) captures the core connection between action and options in virtue of which, as I have said, agentive possibility is alethic. The introduction of ancestry in no way trivializes the demand that is being proposed. Note, for example, that an analogue of (PP*) is clearly false when applied to deontic possibility: if an agent performs an act, it does not follow that it is permissible for her to perform an ancestor of that act. (PP*) is true of agentive possibility, but not of (for example) deontic possibility, precisely because agentive possibility is constrained by an agent’s actions in a way that deontic possibility is not.
What of Kikkert and Vetter’s concern about the alleged obscurity of the notion of an option? This concern, as I described it, was brought against one way of defending (PP). Does it apply also to (PP*)? I think it does not. My revision of (PP) and defense of (PP*) notably did not turn on any tendentious claims about options or on introducing a distinction in options. The notion of an option remained univocal throughout. The issue turned on an orthogonal issue in the metaphysics of action, one on which the framework of options is neutral.
Finally, (PP*) is fit to play the role of (PP) in the subsequent arguments of the book. For example, in my defense of compatibilism in Chapter Seven, I argue that (PP) is compatible with the existence of unexercised options in a deterministic world, since it concerns only the acts that an agent actually performs. The same is true of (PP*), since it too concerns only the acts that an agent actually performs. In general, the appeal to ancestry in (PP*) does not impact the dialectical work that the Performance Principle is subsequently asked to do.
