ON THE STRATEGY oF Methodological SIMILARITY
I will now critically examine arguments that defend explanationism in metaphysics by reference to methodological similarity with science (e.g., Swoyer 1999, 2008; Paul 2012). I will put aside various differences in details and focus on a common gambit of methodological unity, which can be expressed in general terms as follows.
The gambit begins with the premise that explanationism is truth-conducive in (some relevant area of) science, and hence justified in that context. It is then argued that (a given area of) metaphysics is methodologically continuous with science in the following sense:
MC1 Both metaphysics and science employ inference to the best explanation.
MC2 We have no reason to think that if explanationism is truth-conducive in science, it is not so in metaphysics.
MC3 We have a positive reason to think that if explanationism is truth-conducive in science, it is also so in metaphysics.
On the basis of these assumptions, the gambit concludes that explanationism in (the given area of) metaphysics is truth-conducive, and hence rational and justified also in that context. (Note that MC2 does not entail MC3: even if we cannot see why explanationism would fail to be truth-conducive only in metaphysics, we may not be able to positively argue for its truth-conduciveness either. Both MC2 and MC3 are required to tackle a skeptic who demands a positive argument for the use of IBE in metaphysics.)
As far as I am concerned, the conclusion follows if MC1-MC3 can be established. But the problem is that MC2 and MC3 have not been established, and there is reason to think that they cannot be established. It is fairly obvious that without support for MC2 and MC3, the gambit reduces to a more-or-less trivial recognition that theory choice in different contexts and disciplines can be described in similar explanationist terms at some level of abstraction. It is clear that this kind of purely descriptive continuity does not in itself carry any justificatory weight. It is comparable to a foolish attempt to justify any old enumerative induction merely on the basis of it being of the same form as some licit enumerative inductions (cf. Norton 2003).
Before we analyze the problems with MC2 and MC3 in detail, it is worth noting an ill-considered worry about the gambit. It has to do with an appropriate reading of ‘truth-conducive’ and ‘justification’ in the schema above. One might think that the gambit is problematic if we have reason think that inferences to the best explanation are much less reliable in metaphysics than they are in science, or even if we lack a positive reason for thinking that they are equally reliable in both. For example, one might point to differences in the disciplines’ track records—physics vs. metaphysics, say—or argue that our best reasons for thinking that scientific IBEs are reliable do not carry over to metaphysics, raising the specter of potential unreliability (cf. Ladyman 2012). And one might think, in particular, that any such worry about the relative reliability in different domains is ipso facto a worry about either MC2 or MC3.
Although there is something to this worry—one cannot wholly divorce a method’s reliability from its justification—it is quite difficult to square it with the fact that metaphysics is admittedly inherently speculative in a way that our best science arguably is not. I take it that theorizing in metaphysics is generally not taken to be progressive in the way science is. For the realist at least, science is systematically latching onto unobservable reality in an ever-better way, and scientific theorizing is guided or constrained by “correspondence principles” that are grounded in the ideal of discernible continuity in theoretical development. Whatever progress metaphysical theorizing makes, it appears to be compatible with the likelihood of much more significant theoretical discontinuities. Accordingly, our degree of confidence is admittedly significantly lower to any particular output of explanatory reasoning in metaphysics. But this need not mean that in metaphysics explanatory loveliness is not in any sense functioning reliably as a guide to inductive likeli- ness. For instance, it could be that in metaphysics we are much less able to think of a pool of potential alternative explanations from which we choose the “best,” so that we often end up debating over the “best of a bad lot.” Still, it could be that inference to the best explanation is quite reliable in picking out the right theory assuming that it happens to be included in the pool. In this way the method could be reliable, for example, “in the long run” (assuming that we eventually manage to conceive of the right theories) and also rational, despite not being reliable in the sense of engendering a high degree of confidence in any particular explanatory inference.
Hence, some difference in the method’s relative reliability is compatible with the continuity gambit, because such a difference need not affect the method’s justification, which in inherently speculative disciplines such as metaphysics could be based on the notion of reliability “in the long run,” or reliability conditional on not having a “bad lot,” or reliability in mere relative ranking of alternatives. Unfortunately we have little reason to think that explanationism is reliable even in such a qualified way. As I will next explain, there are other differences between science and metaphysics that give us reason to think that the (assumed) justification of explanationism in science does not carry over to metaphysics.
-  Typically explanationists only offer vague tu quoque references to science, scientific realism, and inference to thebest explanation therein. I construe them as aspiring to the schema presented here.
-  There is reason to be skeptical about MCi as well. Cf., e.g., Ladyman 2012.
-  See Lipton (1999, 56ff.) for an explication of the explanationist slogan that “explanatory loveliness guidesinductive likeliness.”
-  The “bad lot” objection against explanationism is due to van Fraassen (1989), who employed it against theidea that it is rational to believe that the best explanation is more likely to be (approximately) true than not.The objection loses its bite if explanationism is construed more cautiously, pertaining only to a hypothesis’sepistemic probability relative to its rival hypotheses (cf. Okasha 2000).