According to Sider, theories based on non-joint carving classifications may be unexplanatory even if true (23). Consider, for example, two true theories T and T* where T is more joint-carving than T*. According to Sider, T is more explanatory than T* because of how much better it carves at the joints than T*.
I’m skeptical of this proposal for largely Hirschian reasons (cf. Hirsch 2017, ms). I take Sider to be saying that T* is less explanatory than T for objective reasons: that to describe the world using T* would result in an objectively inferior description. But why isn’t T* just less explanatory for practical reasons?—i.e., that it’s just easier for us to formulate our theories in T rather than in T*? Imagine two linguistic communities, call them L and L*: L-speakers speak our language—English—and L*-speakers speak some strange and bizarre version of it. Now compare two theories offered by L-speakers and L*-speakers respectively,
TL: The glass shattered only if the baseball shattered it.
T*L*: The glassable shattered only if the bageball shattered it.
where ‘glassable’ means ‘the glass or the table’ and ‘bageball’ means ‘the baseball or the bagel’.
Now, as theories, both TL and T*L* are true. In particular, T*L* is true because its strange disjunctive predicates are all true. But, so says Sider, TL is the more explanatory theory. Why? Because its predicates, namely ‘glass’ and ‘baseball’, somehow carve better at the joints than ‘glassable’ and ‘bageball’. But what joints? Of course, us English speakers would and should prefer a theory formulated in L rather than L*, but why is this a mark of objectivity? Why should this make TL the more objective theory? Perhaps, with Sider, we ought to say that T*L* is unexplanatory (and hence less objective) because its carvings seem arbitrary: glasses and tables and baseballs and bagels don’t go together in any natural way. Well, prima facie maybe they don’t; maybe they don’t go together in any meaningful, worldly way, but regardless, it seems the explanatory power of the theory remains unchanged; T*L* doesn’t leave anything out: the glass or the table shattered only if the baseball or the bagel shattered it. And that’s just true!
Consider another example. Imagine a species of humans with extraordinary computational ability; surely any joint-carving discrepancies in the logical equivalence below would be negligible to these extraordinary humans:
Indeed when these extraordinary humans see the formalism flanking the left side of the equivalence symbol, they process it immediately as p. Now, would Sider say that the more complicated formalism is less joint-carving then p? More precisely, if both formalisms were invoked in two true theories—one theory which used only p and the other which used the more convoluted formalism—would the theory using only p be more explanatory? Perhaps we would, at first, be skeptical of the extraordinary humans strange preference for the more complicated formalism: why use it to describe a theory when p is just more simple, more elegant, more easier to parse (especially when the formulas get really long and complicated)? But to the extraordinary humans, the complicated formalism isn’t hard to parse at all; to them it’s read just fine without any hesitation; to the extraordinary humans, the more complicated formalism is more beautiful and that’s why they prefer to use it in their theories.
Now, if I’m understanding Sider correctly, I believe he would say that the extraordinary humans are just getting something wrong: a theory’s syntactical simplicity is a guide to more better, more objective theories; the theory written in the more convoluted formalism, though true (and equivalent to the simpler theory) is explanatorily inferior. If this is really what Sider is suggesting, then I’m not sure I can buy it—at least not just yet. In the two examples above, the explanatorily robustness of a theory seems only to be a feature of our preferences rather than a feature of its objective virtues. As equivalent and true theories differing only in syntactical complexity (or lack thereof) both encode the very same information. One might be more simple to use and handle than the other, but does this really entail that one may be more explanatory and, hence, more objective than another?
Should we not find it suspiciously anthropocentric for ‘dog’ to just carve better (but not perfectly) at the joints than, say, ‘trog’, i.e., ‘the tree or the dog’?