The Problem of Interference and Prevention
Russell’s problem of action at a temporal distance is actually a special case of the problem of interference and prevention. His argument is designed to show that the general model of causation in terms of temporally
Causal Necessity 89 successive entities, of which the first exerts an influence on the second, cannot in principle exclude that an interference or prevention occurs in the unavoidable ‘gap’ between the end of a cause and beginning of an effect. The only difference really is that Russell’s argument isn’t formulated in terms of a concrete example, only on a highly abstract temporal characterisation of causal connections.
However, the problem of interference and prevention, as it is usually formulated, typically refers to a family of concrete counterexamples that share the following general form. Given the occurrence of some kind of cause C, whose occurrence is supposed to reliably yield an effect E, then it seems always possible to assume the presence of some additional item (/> such that it would be the case that £ would not follow even if C occurred, or, in the case of pre-emption, that even if £ actually does follow C then it isn’t as a result of C but instead of the additional item >. The argument presupposes a certain understanding of necessity, notably the kind of unconditional necessity of which J.S. Mill spoke:
If there be any meaning which confessedly belongs to the term necessity, it is unconditionalness. That which is necessary, that which must be, means that which will be whatever supposition we make with regard to other things.
This is the type of necessity that Mumford & Anjum take to be relevant for causal necessity, and which they use to construe what they call ‘the test of antecedent strengthening’; a criterion they think any successful account of causal necessity must fulfil:
If A necessitates B, then: If A plus >, for any >, then B.
It is important to note that the test is to be applied to particular theories. That is, it should be understood as saying that whatever a certain theory specifies as a cause C that necessitates a specified effect £, then it had better be the case according to that theory that if C happens we always get £ ‘no matter whatever supposition we make with regard to other things’; i.e. the test must be true to the theory being tested both in the kind of cause and effect posited, and the kind of interference > added to the situation. In other words, an argument operating with neo-Humean causes and effects, as well as interfering factors, will be irrelevant as a critique of a powerful particulars view. And, of course, it will not do to postulate as interferences/preventions anything that is actually forbidden by the theory in question, say, interferences by a deity, wizards, or other occult phenomena (which are forbidden on any naturalist account). A theorycan hardly be expected to be impervious to objections invoking phenomena that the theory excludes as unreal.
So, in sum, the problem of causal necessity is this: is there an account of causation that can establish that whenever a certain kind of cause C occurs, then whatever suppositions we make with regard to anything else not already excluded by the theory, the effect £ always and invariably follows? Today it is taken for granted that for any putative cause C ever suggested in the history of philosophy, it has been all too easy to construe counterexamples involving some interfering factor (/> that prevents the putative effect £ from following.
Over the years different types of counterexamples involving prevention and/or interference have come to be roughly divided into four main categories: (i) prevention, (ii) pre-emption, (iii) finks and (iv) antidotes. Prevention occurs when some factor > in the circumstances in which C occurs prevents £ from occurring. For instance, suppose that an assassin fires a shot and the bullet travels the distance to the victim but is deflected out of harm’s way on the very moment of impact, by a second bullet fired by another assassin; leaving only a scratch, as a mark of them having achieved contiguity. This is supposed to illustrate that C can occur— the assassin fires a bullet, the bullet travels to the victim and becomes contiguous with it—and yet something intervenes to prevent the intended effect from occurring.
Pre-emption is when (/> does not prevent £ from following C, but prevents C from being the actual cause to £ simply by producing £ instead of C. In that case, > is really another token cause of the same kind as C, which competes to produce £, and wins. Consider assassins Duke and Drew shooting at the same victim and both hitting the victim in the heart, except that Drew is a tad closer wherefore her bullet kills the victim an instant before Duke’s bullet arrives (alternatively, renders the victim as good as dead). The assumption is that a cause C occurs to completion but its designated effect, the death of the victim, is pre-empted by >; the victim is already dead or as good as dead. There is a counterfactual version of this in which both bullets hit the heart simultaneously, but each pre-empts the other in the sense that even if one wouldn’t have hit, the other would.
In the case of prevention and pre-emption, the interfering factor is external both to the cause and the effect, or, rather, external to the object that acts (Agent) and the object that is acted upon (Patient). Finks and antidotes, on the other hand, are instead peculiar kinds of interfering powers that are intrinsic to the Agent and Patient, respectively. Finks are located in the kind of object that counts as an Agent in the standard view, and they work their interfering magic by defusing the Agent’s own supposed power to produce a certain effect. Using a suitably modified assassin scenario, consider a bullet fired at a victim and reaching the victim’s heart but causing no damage because in addition to having momentum,
Causal Necessity 91 impenetrability, and velocity (powers to produce death), the bullet has the ‘finkish’ power to instantaneously evaporate when coming into contact with striated muscle cells, such as the heart is made of. C.B. Martin’s famous finkish live wire—deadly until touched—is another example of the same kind (1994). Consequently, the bullet penetrates the outer layers of tissue, but evaporates as soon as it touches the heart; the victim is maimed but doesn’t die.
Antidotes are powers of the Patient either to defuse its own liability to be affected by the action of the Agent or defuse the Agent’s power to produce an effect in the Patient. Thus, if the victim is disposed to instantaneously become perfectly permeable (i.e. ghost-like) when coming into contact with a particular metal present in the bullet, then on contact the victim becomes ghost-like, and the bullet passes through the victim without causing damage. Antidotes to poisons work in a similar way, either by directly neutralising the toxic abilities of the poison (e.g. active charcoal) or affecting the ability of the Patient to be affected by the poison (e.g. calcium channel blockers). Of course, real-life antidotes are extrinsic to the Patient and so really preventive factors.
It is indeed very difficult to come up with realistic cases of ‘finks’ and ‘antidotes’, since normally we would consider anything that has an intrinsic ‘fink’ or ‘antidote’ to already be immune to the action of the Agent in question, or unable to act on the Patient in the way required to produce the intended effect. An assassination-victim who is intrinsically immune to poison will simply not have the power to be poisoned that is prevented by a second power of immunity. This is why discussions about interference and prevention are full of oddly contrived powers, such as becoming perfectly permeable. Sometimes they even invoke magic (e.g. Manley & Wasserman 2008: 62) or divine intervention (Reeder 1995: 143). The supernatural is already excluded by causal realist commitment to the genetic principle but let us see how the standard view might respond to the more realistic counterexamples.
In sum, arguments invoking cases of prevention and interference aim to show that some cause C occurs to completion and yet does not lead to E, because some > interferes to stop £ from happening, or to make C superfluous to the production of £, as in the case of prevention and preemption, or cancel the occurrence of £ by making the Agent impotent or the Patient impervious.