An Analysis of Causation

The ancients understood causation in terms of causal powers. The reason for this is because of how they understood change. Change for the ancients was just the reduction of potency into act - see analysis on act and potency here. Hence, the change of water from cold to hot is the potentiality of hotness in the cold water to become actual (and the actuality of the coldness in the water to become potential). But what brings about this actuality had to be something external to the thing undergoing change, otherwise there would be no explanation as to why the thing has not already undergone the change in question; however, the ancients clearly understood that not any old thing could bring about a change in something: ice can bring about change in water in a way that smoke cannot; a magnet can bring about change in metal in a way that it cannot with wood, &c. There are limits to the effects that efficient causes can produce. The ancients understood these limits by referring to the causal powers that efficient causes have. So ice cools water because it has the power to cool objects it touches. Smoke does not cool water because it does not have this power to cool. 

Now, this understanding of causation has been mocked as trivial and vacuous. French playwright Molière says that to say ‘opium causes sleep because it has the power to cause sleep’ is to say a mere tautology, and hence explains nothing. But as pointed out by Ed Feser, this is not a tautology: a tautology would be to say that ‘opium causes sleep because it causes sleep’ rather than ‘opium causes sleep because it has the power to cause sleep’. To say that opium has the power to cause sleep is to say that falling to sleep after taking opium is not some mere accidental feature of taking opium, but belongs to the nature of opium itself. The fact that this is not a tautology is evidenced by the fact that critics of causal powers don’t say to this statement, ‘yes, we know, and this is too trivial to be worth mentioning.’ Rather, they say that ‘No, Opium does not have the power to cause sleep as things do not have causal powers.’

It would also be wrong to imagine the idea of causal powers as going against what science tells us. So, in the case of opium, the idea of powers does not contradict the chemist who tells us that opium causes sleep because it is composed of a such and such chemical compound that reacts with the body when digested. As noted by Anthony Kenny, we need to distinguish the possessor of power, the power itself, the vehicle of power, and the actual exercise of power. In the case of opium, it is the specific chemical properties that are the vehicle by which its casual power is exercised; other substances will have other vehicles by which their powers are exercised. The difference, thus, between the chemist and the metaphysician is that the former is concerned with vehicles while the latter is concerned with powers. 

David Hume questioned this way of analysing of causality. In his An Enquiry Concerning Human Understanding, Hume writes that the ‘constant conjunction of two objects’ in our experience is what leads us to regard these two objects as casually related, but that objectively ‘all events seem entirely loose and separate’ rather than being necessarily connected. In principle, any effect or none might follow from any cause. The efficacy we think we perceive in things is really just a projection of our expectations onto the world. So, a cause is analysed by Hume as follows: ‘We may define a cause to be an object, followed by another, and where all the objects similar to the first, are followed by objects similar to the second.’ (Section VIII). What Hume is espousing here is a regularity theory of causation. Regularity theories of causation argue that causation can be captured entirely in terms of the regular correlation that exists between two events - say, A, and B. No reference is required to explain the relationship between A and its effect B, such as the power A has over B. The relation between them are loose and separate. 

However, there are several objections that face regularity theories. First, regularity theories have difficulty in accounting for the asymmetry between causes and effects. A regular correlation between A and B does not entail that A is the cause of B rather than B the cause of A. Adding a condition that a cause must precede its effect B will not help since, as proved by Kant, causes can be simultaneous with their effects, such as a ball on a cushion making an indent. A second problem is that there are cases of regularity that are not cases of causation. This is illustrated by the following example adapted from Ed Feser. Suppose that I drop a cube of sugar into my tea; this is followed by a splashing sound, and then ripples, which move a floating tea leaf to the side of the cup. Events like the moving of the tea leaf are regularly preceded by events such as ripples and the splashing sound. So, on the regularity analysis, the splashing sound and the ripples are equal candidates for being the cause of the leaf’s motion. But, of course, it is only the ripples, and not the sound, that is the cause. A third difficulty can be seen if we add the example of someone else trying, before the splash is made by the first cube, to add a second cube of sugar to my tea but say it is caught before it can make contact (I have already put enough in my tea). The regularity theory would seem to entail that the dropping of the second cube of sugar is the cause of the splash in the tea. But, of course, it is only the dropping of the first cube of sugar. 

There is a second definition of causation that Hume offers. Well, Hume did see this as a further explanation of the first, but it is really a second definition. According to this second definition, we can say that an object caused another object ‘if the first object had not been, the second never had existed.’ Here Hume is developing a counterfactual definition of causation. We can formalise it as follows. To say A caused B means that the following is fulfilled:

1) If A had not occurred, B would not have occurred.

2) If A had occurred, B would have occurred.

3) A and B both occurred.

This is a better analysis of causation as B depends counterfactually on A in a way that A does not depend on B; hence, a solution to the asymmetry problem. Neither does the account seem to be threatened by examples involving the sugar cube: the leaf would not have moved had the ripples by the sugar cube not been made, through it would have still moved had the splashing sound somehow been prevented. The counterfactual account therefore captures the fact that it was the ripples and not the sound that moved the leaf. It also captures the fact that it was the first cube of sugar and not the second that moved it. 

Nevertheless, this account of causality is still Humean in that it views causes and their effects as loose and separate, in that they have no intrinsic or necessary connection to one another. The causes under this view cannot be said to have an active tendency to bring about their effects. This leads to problems, which have been well noted; particularly influential has been C.B Martin’s ‘electro-fink’ example. Consider a live wire, which if touched by a conductor, will cause electricity to flow into it. If the counterfactual analysis is correct, anything we might want to say about the causal relation in question here would be captured in a conditional such as the following:

c) If the live wire is touched by a conductor, then the electrical current flows form the wire to the conductor. 

But now imagine that we connect to this wire an electro-fink, which is a device that renders a live wire dead when touched by a conductor and dead wire live when touched by a conductor. The conditional above will now no longer be true. If a live wire is touched while the electro-fink is attached, current will not flow to the conductor, because it will be prevented from doing so by the electro-fink; hence the conditional fails to give the necessary conditions for the wire’s being live, since a wire could be live even when it is not true that it will transmit current to a conductor. And when the wire is dead, current will still flow from it to the conductor, because it will be made live by the electro-fink; hence the conditional fails to give sufficient reasons for the wire being live, since a wire could be dead and yet still transmit current to a conductor. The proper way to characterise the wire, in Martin’s view, is to say that it has power when live which is hindered by the electro-fink, and lacks power when dead but is given power by the electro-fink when touched. 

Additionally, the counterfactual analysis has trouble in adequately accounting for masks and antidotes. So, in the tale of King Midas, King Midas had the power to nourish himself, but this power was masked by his power to turn everything he touched into gold. And we can imagine a case where someone takes a poison that has the power to kill that person, but the poison is hindered by an antidote that changes the immune system of the person so that he can resist the poison. 

One could try and reformulate the counterfactual definition as follows. A causes B if:

1*) If A had not occurred, B would not have occurred.

2*) If A had occurred and C did not occur, B would have occurred.

3*) A and B both occurred and C did not occur.

The trouble with this new definition is that we can easily think up examples where A and C both occur and yet B does occur. For instance, someone takes poison and is given an antidote to stop the effects of the poison but is then given another poison to counteract the antidote to allow for the original poison to take effect. And similar revisions and counter-examples can kept being made on the counterfactual definition. It seems, then, that talk of powers is a much better way of analysing causation. 

Indeed, talk of powers fits well with how we describe events involving change. To again appeal to Feser, let us consider a case of a boat being pulled at an angle by two horses, one each side of the canal. The combined pulling of these horses causes the boat to head northwards up the canal. What we see here is a combination of powers at work. The power exercised by horse A and the power exercised by horse B leads to the boat moving down the canal, but these powers would result in the boat moving in a different direction when exercised in a different context. Moreover, we would not describe what is going on here as a set of loose and separate events but just one event that is the exercise of multiple powers; an effect can be the result of a polygeny of powers.    

The talk of powers, therefore, seems to be a good way to analyse causation. The very nature of these powers and how they relate to philosophy of science will be discussed in a later blog post.