Non‐committal Causal Explanations

Abstract

Some causal explanations are non‐committal in that mention of a property in the explanans conveys information about the causal origin of the explanandum even if the property in question plays no causal role for the explanandum. Programme explanations are a variety of non‐committal causal (NCC) explanations. Yet their interest is very limited since, as I will argue in this paper, their range of applicability is in fact quite narrow. However there is at least another variety of NCC explanations, causal orientation explanations, which offer a plausible model for many explanations in the special sciences.

Acknowledgements

Research leading to this work has been partially funded by the Spanish Government, research grants FFI2008‐01580/FISO, HUM2007‐61108 and CSD2009‐00056. Previous versions of it have been read in conferences in Barcelona, Freiburg (Switzerland) and Kansas state. I thank the audiences in all these places for useful comments and suggestions. I’m also grateful to two anonymous referees of International Studies in the Philosophy of Science.

Notes

[1] It is usual in the literature to distinguish between events and states as follows: the former involve change and the latter do not. This distinction is completely irrelevant for my purposes and I will ignore it.

[2] In giving this definition I deliberately remain non‐committal about whether events should be individuated finely, as for example in Kim (1980), or coarsely, as in Quine (1960). Again nothing essential in our discussion depends on taking a definite stand about the right way of individuating events. For a useful discussion of the metaphysics of events see Bennett (1988).

[3] I have in mind here something akin to Armstrong’s conception of laws as a relation between universals. In the case of causal laws, the antecedent properties would be the causing properties and the consequent properties would be the caused properties. As I said at the outset, this assumption is just a way of defining a metaphysical background for the discussion that follows. I do not think that any of the claims I wish to defend depends on something like Armstrong’s conception of natural laws.

[4] Fodor (1987b, 142) explicitly endorses all these assumptions about causality.

[5] For the purposes of this paper I will take effects, that is to say, the explananda of causal explanations, to be instantiations of a property by an object or of a relation by a sequence of objects. This is also what should be expected given the metaphysical background described in the previous section.

[6] This is not an assumption shared by everybody. Some theorists, for instance, consider that a causal explanation ought to mention only causally efficacious factors. Under such a view, of course, requirement (NCE) looks oxymonoric. I will not argue here for the plausibility or implausibility of such a restrictive view about causal explanations. I shall be happy if the reader finds reasonable my less restrictive usage, namely, that causal explanations are statements which transmit causal information about the explanandum. I think it is reasonable to regard a causal explanation of E as a statement which gives information about the causal origin of E. How this information is conveyed, whether by mentioning a causally efficacious factor or otherwise, as is the case with NCC explanations, does not jeopardize in my view the status of the statement as expressing a causal explanation of E.

[7] I say that this is part of the information conveyed by (E2) since most philosophers are prepared to regard the property of the elevator of having more than ten people inside as being causally efficacious as well for the property of the elevator of not moving: see for example Yablo (1992) and Shoemaker (2000, 2007). However, it is crucial for our purposes to observe that it is not part of the counterfactual information actually conveyed by (E2) that the property of having more than ten people inside is causally responsible for the property of the elevator of not moving.

[8] In spite of this, one should not see (E1) and (E2) as expressing two competing explanations of the same effect, since the causal information provided in each explanation is different in kind, counterfactual in the case of (E2) and non‐counterfactual in the case of (E1).

[9] The problem of causal exclusion has been advanced by Jaegwon Kim for the last twenty years. A recent and particularly clear statement of it can be found in Kim (1998, ch. 1). The argument of causal exclusion purports to show that physically multiply realized properties are causally inert.

[10] This clause (a) in the definition (J&P) looks unnecessary and in fact not relevant for the sort of reasons just stated in the main text. I’m here simply following Jackson and Pettit original analysis which does include this clause. Later on, when I offer my own analysis of programme explanations, this clause will be dropped.

[11] Given how the notion of causally efficacious property has been elucidated in the first section, clause (b) implies that there are causal laws with the realizer or subvenient properties in the antecedent part of the law and the property individuating the effect type in the consequent part; presumably one such law for each realizer. Of course a particular causal explanation has as explanandum an effect token, namely, a particular instantiation of the property individuating the effect type. For instance, in our example, a particular instantiation of the property of not moving by such and such elevator at such and such zone of the space‐time. Of course only the instantiation of one of the realizers brings about the effect token. We may therefore distinguish a causally efficacious property for the effect type from a causally effective property for a token of the effect type. All realizers are causally efficacious for the effect type, just as (J&P) states, but only one of them is causally effective for a particular token of this effect type. I’m grateful to an anonymous referee of ISPS for calling my attention to this point.

[12] Throughout this paper I will assume Quine’s conception of predicates as open sentences for expository convenience. I think, though I’m not quite sure, that this usage comes from Frege. Frege thought that predicates refer to concepts and concepts are functions which assign truth values to their arguments. Frege also held that functions are incomplete or unsaturated entities which demand one of their arguments to become complete. This is why he thought that any linguistic expression referring to a function, as natural language predicates do according to him, should conspicuously include a sign or mark of the incompleteness of its referent. This is why a variable such as ‘x’ should occur in such an expression, according to Frege (1892), to signal this incomplete character of its referent.

[13] Actually, metaphysical condition (MC) is only required in order for the explanation to convey true or veridical causal information.

[14] Thanks to an anonymous referee of ISPS for raising this issue.

[15] I’m grateful to Bruce Glymour and to an anonymous referee of ISPS for raising this important point and the need to discuss it.

[16] I’m of course assuming a realist analysis of dispositions (as defended in Armstrong 1968 or in Lewis 1986b) rather than a Rylean account (Ryle 1949, ch. 5). I lack the space here to justify this assumption. See however Armstrong (1968, 85–88) for some compelling reasons for the realist interpretation.

[17] The functionalist position which is relevant for our discussion is of course so called ‘analytical functionalism’, the claim that mental predicates are a certain kind of dispositional predicates. Some other functionalists, frequently labelled ‘psychofunctionalists’, hold that the essence of mental properties is the causal role they play in an empirical theory belonging to cognitive science. This latter position, to the extent that it remains silent about the nature of mental predicates, is not relevant for our discussion. On the distinction between analytical functionalism and psychofunctionalism, see Block (1994, 325).

[18] Some will be inclined to think that (E3) tells us also that solubility in water itself is causally responsible for the dissolution of O. But we can leave aside these controversial issues about the causal efficacy of functional properties since this is in any event not part of the disjunctive causal information allegedly conveyed by (E3).

[19] The assumption made in this paragraph—namely that it is not a priori true that the property referred to through the predicate ‘x is SOL’ is a realizer of solubility in water—is based on the idea that the fact that the property being SOL is a realizer of solubility in water is something that depends on the causal powers of being SOL, that is to say, on a contingent fact about being SOL, its power to cause dissolution in water, presumably known a posteriori. However, a suitable analysis of a microphysical predicate such as ‘x is SOL’ might show that this contingent fact is part of the correctness condition for ‘x is SOL’ (I’m not endorsing such an analysis, but suggesting it as a mere possibility). If this were so, then it would be a priori true that to apply the predicate ‘x is SOL’ to an object implies applying to it the predicate ‘x is soluble in water’, and so our assumption would be false.

I do not think that such an analysis of microphysical predicates such as ‘x is SOL’ is correct, but even if it were, it is important to observe that it would not suppose an objection to the main point defended in this paper. This is so because it would still be true that our competence in a dispositional predicate such as ‘x is soluble in water’ does not require our competence in microphysical predicates such as ‘x is SOL’. The epistemic condition (EC) of the analysis of programme explanations would then not be met, since ‘x is soluble in water’ would not C‐entail the set formed by ‘x is SOL’ and ‘x is LOS’.

[20] The use of the subjunctive expresses the already mentioned fact that the disposition does not require that the triggering circumstances occur, it only requires that if they do occur then the suitable manifestations should follow (in normal conditions).

[21] Of course, if there were some sort of miracle or disruption of the causal laws thus preventing the categorical basis of the disposition together with the triggering circumstance to bring about the relevant manifestation we would say that the conditions are not normal.

[22] Notice that ‘x is soluble in water’ is not conceptually entailed by ‘x dissolves in water’ since an object O may be dissolved in water under conditions which are not normal (for example, if it were divided in very tiny pieces). Notice also that for dispositions with more than one manifestation, each manifestation predicate will be equally D‐entailed by the dispositional predicate though of course through different triggering circumstance predicates in each case. Finally, as is well known normal conditions are hard to unpack. Yet, a competent user of a dispositional predicate can in general discern when normal conditions are not normal in every particular case. Still, it is not preposterous to hold that there might be a certain margin for conceptual revision here as knowledge about relevant normal causal processes grows.

[23] I borrow this example from Jackson (1996, 396–397).

[24] Content C need not be a propositional content, as seems to happen in our example. I’m ignoring this complication here.

[25] When discussing the intensionality of causal explanations at the end of Section 2, I mentioned the interesting case in which we use in the explanans a special science predicate referring to a property the realizers of which are completely unknown. I said that in this case we would not be offering a programme explanation, since the predicate used in the explanans does not C‐entail any set of predicates of the realizers and therefore no disjunctive causal information is offered about the effect E. Still, it seems correct that at least in some cases we would regard such explanations as good enough. Introducing causal orientation explanations might help us to see why. As I have just argued in the main text, we may provide a correct causal orientation explanation about E just by using in the explanans a dispositional predicate even if we are completely ignorant of the categorical bases of the functional property mentioned. The explanation would tell us that the causal path leading to the E crucially involves a particular manifestation property of the functional property mentioned in the explanans, which in turn is caused by an unknown or unspecified categorical basis.