Strevens’s account of ad hocness includes two different types of ad hoc hypotheses – the ones that are a part of a glorious rescue and those that are a part of a desperate rescue. Howson and Urbach’s account of ad hocness also includes two
different types of ad hoc hypotheses – the acceptable and the unacceptable. And it looks like any similar Bayesian account of ad hocness will also include two different types of ad hoc hypotheses, whatever their labels. I will use Strevens’s account as the illustration for the problem found in so doing, but the critique will apply to all of these Bayesian accounts of ad hocness.
The problem in so doing is twofold. The first problem is that ad hocness is used as a pejorative in science and in (most of) philosophy. The Bayesians will have to give a good reason as to why the glorious rescues are actually ad hoc moves at all. Strevens, for example, differentiates between the introduction of good ad hoc hypotheses and bad ad hoc hypotheses – that is, the types of ad hoc hypotheses that are justified and those that are not. This is the distinction identified in the distinction between glorious and
desperate rescues. However, this sort of heterogeneity of ad hocness runs strongly counter to how scientists use the term ‘ad hoc’. Having part of one’s theory labeled ‘ad hoc’ is considered a detriment to the theory, not a form of approbation. Something that philosophers of science must be sensitive to is to how terms are actually used in science. If we are to differ markedly in our use of terminology, we must give a good reason for so doing. I do not think that Strevens has done this. I will argue this point more both in the second, related problem in this section and also in the criticism found in the next section. The second, related, problem is that, on occasions where certain types of acts can be clearly distinguished, and there is a reason to so distinguish them, it is a mistake to place such acts under the same umbrella term. For example, Strevens’s label ‘glorious rescue’ in the Uranus perturbation case seem to be another way of labeling an instance where it was rationally permissible (maybe even encouraged) for scientists to postulate the existence of another planet to account for these perturbations. To label some act a desperate rescue seems to be another way of labeling an instance where it was not
rationally permissible for scientists to postulate such a hypothesis. To label the former as a good ad hoc move and the latter as a bad one seems to unnecessarily complicate the issue.
In sections VII-X of Chapter 1, I discuss instances where there does not seem to be any principled way to make a distinction between two types of hypotheses or two types of acts. In contrast, there seems, here, to be a major and clear distinction between a move that is rationally permissible to make and one that is not rationally permissible to make. Why not respect this distinction, and the way in which scientists use the term ‘ad hoc’, and label the rationally impermissible moves ‘ad hoc’ and the rationally permissible moves as good scientific methodology?
After all, an important part of scientific methodology is to modify one’s theories in the light of disconfirming evidence. To do otherwise would be either to throw over a theory the instant disconfirming evidence arose, no matter the theory’s virtues, or to maintain theories that will be empirically inadequate. These modifications are being made because of disconfirming evidence, which seems to point to the possibility of their being ad hoc. To label indiscriminately all (or most) of these types of theory
modifications as ad hoc is either to misrepresent what is going on or to render the term ‘ad hoc’ rather meaningless.19 It will become meaningless if it embodies too many actions, especially if the actions that it embodies are very heterogeneous, or if the term
19
In “The Assessment of Auxiliary Hypotheses”, Jarrett Leplin puts this point quite well: “If in fact ‘ad hoc’ is used in science to mark a particular methodological liability, if it has the univocity I have claimed in making this liability a necessary condition of its application, then a neutral epistemic analysis, which must distinguish as many senses of ‘ad hoc’ as there are forms of empirical deficiency, radically underestimates the scientific importance of the concept. […] In so far as ‘ad hoc’ is subjected to distinctions of sense, there is a natural inclination, exhibited by many along the slippery path to Holton, to attribute such differences to differences in usage whose only legitimate significance is biographical. In this situation, if it can be shown that a common judgment was made in even a small number of importantly different cases by use of the concept, then univocity should be presumed pending evidence to the contrary. The distinguishing of sense so popular from ordinary language philosophy is a valuable tool only if not exploited ad hoc to convert philosophical problems into historical ones.” (Leplin 1982 240-241)
goes too far afield from scientific practice. If the term is meaningless, why use it at all? I don’t find any of these consequences acceptable.