I think what it is that Anscombe can’t bring herself to do precisely is what it is that generates tensions within Davidson’s own account of reasons as causes. And part of that is the principle of the nomological character of causality. This is the idea that causal relations are a matter of subsuming events under strict laws. And this principle seems to be of a piece with Davidson’s anomalous monism, or the idea that mental states/events (and actions) are token-identical with physical states/events.
Davidson explicitly credited Anscombe’s 1957 monograph Intention with reviving the view that actions are intentional under specific descriptions, and he indeed saw his own 1963 paper “Actions, Reasons, and Causes” as vindicating the everyday idea that reasons explain actions by being their causes. So, where is it that he diverges from Anscombe in that respect?
I think it is the event-event metaphysics of causation that Anscombe justifiably resists in favor of an account of rational causation. Hence, when the reason why an agent acted (by their own lights, and hence their reason for acting in this way) is appealed to in order to explain the action, then the agent herself is disclosed as intentionally causing things to happen. As Jennifer Hornsby puts it in Simple Mindedness, (to paraphrase,) when an agent intentionally brings it about that P, her action isn’t the cause of P. Rather, her action is her causing P to be the case. (Hence Thompson’s claim in Life and Action that the paradigmatic rational explanation of an intentional action isn’t an abstract reason but rather another intentional action that the first one is a means to accomplishing, by the agent’s own lights, as an exercise of her powers of practical deliberation). This is very different from appealing to “causes” construed as events happening within the agent, as it were.
Davidson’s own account runs into the problem of deviant causal chains that he himself seemed to acknowledge to be a massive headache for his theory. Consider his own famous example of a climber holding another climber by a rope. The first climber wants to rid himself of the weight and he knows that loosening his grip will do it. This desire and belief unnerve him so much that they cause his hand to tremble and loosen its grip. According to Davidson’s strict causal criteria, the belief and desire “caused” the action. But intuitively the climber did not act intentionally. It was an accident caused by nervousness.
Davidson struggled to explain how a reason could cause an action “in the right way” without falling back on teleological language. For Anscombe, this problem doesn’t even arise in the first place. That’s because she doesn’t treat the reason as a separate antecedent event pushing the body. She doesn’t have to worry about the “mechanism” missing its target. The intention is in the action structured by the agent’s practical knowledge.
The idea that actions are caused by antecedent mental events fits neatly with what Susan Hurley called the sandwich conception of the mind, or the idea that mental acts (and the formation of intentions) are events happening in the head of an individual temporally sandwiched between sensory inputs (incoming information) and behavioral outputs (bodily movements). By contrast, embodied and enactive accounts of the mind, and Thompson’s account of action, which Foot draws upon and which owes much to Anscombe, dispense with this picture entirely. They reject the idea that actions are triggered by antecedent mental events like beliefs and desires that simply happen to an agent. Instead of viewing action as the passive output of a psychological sandwich, descriptions of intentional actions disclose them as being under the continuous active control of the agent. The paradigm here is someone pushing a crate along an inclined plane rather than someone hitting a billiard ball. The action is a protracted event whereby the agent is causing (progressive aspect) the crate to slide up against friction, and of driving the action/movement towards its completion. The characterisation is irreducibly teleological.
So, I’d like to say that when Anscombe can’t bring herself to do as Davidson does, it’s precisely because treating reasons as “efficient causes” effectively obliterates the agent as a rational self-moving subject, turning her instead into a passive container for causal happenings.
By the way, Anscombe’s main responses to Davidson’s conception (and to other people as well) can be found in “Causality and Determination” (1971) and “Under a Description” (1979)