In this article I argue that the objective probabilities of free choice which some agent-causal incompatibilists appeal to as a way of explaining how it is that free choices can be subject to the influence of factors of which the agent is not aware are, in fact, incompatible with the agent-causal relationship. In section 2 I motivate the issue by noting that there is reason to think free choices can be so influenced, such that it would be desirable to have an account which could explain this influence. In section 3 I argue that the agent causal relationship invoked by Timothy O’Connor’s account of free choice involves a form of determination (between the agent and the intention formed) that is incompatible with objective probabilities of free choice. I show that other versions of agent-causation and two-way powers-based accounts of (free) agency also posit such a determination relation, with the result that those accounts too are incompatible with objective probabilities of free choice (or free action). In section 4 I sketch a type of agent-causation account which allows for exercises of the agent-causal power to be indirectly influenced, such that the apparent influence of free choices can be understood without appeal to objective probabilities.
I develop two Rawlsian original position-style scenarios with which to assess three accounts of agential, freedom-level control. I suggest that by assuming these accounts have been shown internally coherent and had all major objections answered, the Rawlsian original position provides a novel way of assessing the value of the quality of the agency these accounts bestow. By exploring some possible trains of thought which reflect on these scenarios, I suggest that it is not implausible that many readers, including some committed compatibilists, might find themselves judging the incompatibilist, agent-causal control to be more desirable – more worth wanting – than either a compatibilist reasons-responsive account of control or a compatibilist hierarchical mesh account of control. Such compatibilists should, then, recognise an extra cost to the position they endorse: it is not the most valuable form of control. For compatibilist readers who do not share the intuitions I hope to elicit, the scenarios may nevertheless allow them a greater insight into why many incompatibilists find the agent-causal control valuable.