In a previous post we looked briefly at an argument put forward by David Hume that the concept of cause can be meaningful only if it is directly observable. He argued that when we pay careful attention to our direct observations, we do not see any necessary connections between the objects which are said to be causes and effects. Rather, what we see are regular successions between kinds of objects. Thus, to say that A causes B is just to say that objects of kind-A are regularly followed by objects of kind-B and that we never (or rarely) observe B’s without their being preceded by A’s. Or more specifically, to say that a moving brick caused the glass window to shatter is just to say that these two objects (or events) regularly co-occur, and we never observe windows with similar characteristics not shatter when bricks are propelled at them.
Hume also argued that his analysis could make sense of causation within the context of human action. When we experience ourselves causing things through acts of trying, we infer a causal connection because our minds anticipate a correlation. We anticipate a correlation between our acts of trying and the outcomes of our tryings because we have regularly observed them together in the past. Similar tryings have led to similar outcomes, and that is our basis for knowing we have caused things by our actions.
But is Hume’s analysis of causation adequate? I will consider three reasons why his analysis should be rejected.
First, Hume’s analysis fails to distinguish accidental regularities from true causes. Sometimes A’s are always followed by B’s without causation being involved, as when “one clock pointing to the hour [is] followed regularly by a different clock striking the hour” (Swinburne, 1997, p.80). In this case, we know that the one silent clock pointing to the hour does not cause the sound made by the other clock at the same hour.
Second, Hume’s analysis fails to distinguish non-accidental regularities from true causes. Even if causation is involved when A’s are always followed by B’s, A’s can still fail to cause B’s if they both stem from a common cause. For example, suppose the formation of one chemical is regularly followed by the formation of another chemical inside a test tube. Does this mean that one caused the other to form? Not always. Perhaps they both successively formed due to a common chemical element being added to the test tube.
Third, Hume’s analysis of causation by human agents is inadequate. By claiming that our knowledge of this form of causation stems from observed regularities between (A) acts of trying and (B) the outcomes of trying, he assumes that agent causation is best analyzed as a relation between two events, namely events A and B. But there is good reason to think that agent causation does not yield to such analysis. Why? Because as Swinburne (2011) argues, the exertion of causal influence (at least when I perform basic actions) just is the event of the agent trying. I experience the influence as a single event, not as a relation between it and some prior or subsequent event (p.8). Certainly, the event of trying can be followed by a second, third, or fourth event, but my awareness of exerting influence does not come from observing correlations between such events. I already experience my trying as a single event of causal exertion before I can correlate it with another event. Any correlation I do discover already presupposes my prior awareness of exerting influence by trying. Even when my trying fails – as when I endeavor to lift a rock too heavy for me – I still experience my failed endeavor as an event of causal influence, no matter how weak it happened to be. Thus, Hume’s attempt to reduce agent causation to a relation of correlation between events falls short.
** important clarification: I am not denying that correlations can count as evidence of causation. Clearly they can. Rather, my claim is that correlations do not constitute what causation is or serve as our most basic experience of it. **
A SKEPTICAL WORRY
But what if my experience of exerting influence is illusory? Presumably I do feel myself trying, but how do I know that my trying is really making a causal difference to things? Perhaps (unbeknownst to me) I exist in The Matrix where these experiences are just effects of a computer simulation, without any connection to the real world. Or perhaps the mental events I experience as “tryings” are just epiphenomena – that is, mental effects which are generated by physical events in the brain, but which themselves make no difference to the physical world. But if an illusion of making a difference is indistinguishable from an experience in which I truly affect things, how can my awareness of trying be regarded as evidence of causation?
That’s an important question. But notice how it assumes that agency can only be evidence of causation if it can withstand skeptical worries like the ones described. Is this a reasonable assumption? Surely not. As Michael Rota (2009) argues, the assumption comes at too high an intellectual price. Why? Because if it is true, then in addition to rejecting the evidence of agency, we would also have to reject our belief in the external world or in the general reliability of our senses. Presumably we can check the reliability of a particular sense by comparing it with input from the other senses (as when I check my vision of a bent stick in water by touching it with my hands) but we have no way of checking if all of our senses are functioning reliably; any argument to that effect would be circular because we “would have to depend on at least one of our perceptual faculties to furnish some of its premises” (p.23).
So what can we learn from Rota’s (2009) argument? Simply this: if skepticism of such grand proportions is required in order to reject evidence from agent causation, then surely we should reject skepticism in favour of the belief that our acts of trying really do make a difference in the world. This belief certainly appears to be true, so in the absence of counter-evidence, we are reasonable in holding it.
 I am assuming throughout this paragraph that the agent is intentionally performing (or trying to perform) a basic action.
 There are two additional points of clarification worth mentioning here: first, this argument can be leveled against any analysis that tries to reduce agent causation to a relation (or set of relations) between events, whether they are regularities, counterfactuals, probabilities, or continuous processes. Second, it is also worth noting that agent causation favours a metaphysic in which substances and their properties/powers are the most basic features of reality, with events derivatively understood in terms of substances having, gaining, or losing properties, etc.
 The modest principle that things probably are as they appear to be – in the absence of counter-evidence – is a principle which Swinburne has cogently defended in his many technical writings.