Monday, November 15, 2010

Various Positions on Analogical Inferences

A while back, in talking about the fallacy of false analogy, I noted that the fallacy, as typically understood, seems traceable to Mill, and that Mill makes it a fallacy because he has a particular view of analogical inference that is controvertible. In discussing Mill's account of analogical inference I called him a minimalist, and contrasted him with Hume, who I called a maximalist. When Alan asked some questions about this, I gave an answer, but I think both my original classification and my response were made unnecessarily murky by a failure to distinguish two key questions a good account of analogical inference would have to answer:

(1) Can good analogical inferences oppose each other?

(2) Must analogies meet some condition beyond real resemblance in order to be good?

Let's call a position that requires a Yes to the first question, inclusivism (it includes opposing inferences as legitimate), and the negative complement, exclusivism. And let's call a position that requires a Yes to the second question, restrictivism (it restricts the conditions under which analogies can be good in the first place), and the negative complement, generalism. We can then better see the position between Mill on the one hand and Hume and myself on the other. Mill is an exclusivist restrictivist. He thinks that any analogical inference that gets an incorrect conclusion is fallacious; and he thinks that good analogical inferences, in addition to the resemblance of the analogy itself, also need to build on an established causal connection. Hume, on the other hand, is an inclusivist generalist (as am I). He thinks that even good analogical inferences are defeasible, and he thinks that all an analogical inference needs in order to be a good inference (albeit one that can be defeated by better or stronger inferences) is resemblance. Indeed, he explicitly says that no matter how imperfect the resemblance may be, the inference "may still retain as much as may be the foundation of probability, as long as there is any resemblance remaining".

Putting it this way raises the question of whether the history of philosophy also includes inclusivist restrictivists and exclusivist generalists. An inclusivist restrictivist would have to hold that (1) good analogical inferences can oppose each other and also hold that (2) something more is needed for a good analogical inference than just some sort of resemblance or analogy. This is obviously a coherent position. I can't think of any significant name in the history of philosophy that holds it, but there really aren't that many significant names in the history of philosophy who discuss analogical inference at length. An exclusivist generalist would have to hold that (1) good analogical inferences cannot oppose each other and yet that (2) all that is required for a good analogical inference is resemblance. This seems like it would be an extraordinarily difficult and unintuitive position to hold, and perhaps an impossible one. At the least one would have to say that a lot of things that seem to resemble each other don't actually resemble each other at all (not 'at all' in the sense of 'in all the most important ways', but 'at all' in the sense of at all). This would take some doing, and I doubt anyone has ever taken the exclusivist generalist line on analogical inferences.


  1. Ocham5:26 AM

    Interesting. How does this relate to 'argument by equal reasoning' which i discuss here?

    (I don't have a copy of Mill to hand right now).

  2. branemrys1:12 PM

    I don't really know, but I think in an odd way it depends on things like whether one is a nominalist. I'm inclined to take parity arguments as arguments based on genus-and-species (i.e. sharing exactly the same kind of argument); under most non-nominalist accounts of what it is to belong to a kind they would have to be different from analogical arguments. Most Aristotelians would be committed to that, for instance; the topos of genus and species is a completely different topos from that of similarity. But there are other views of what it takes to belong to a kind that make this purely a matter of similarity, in which case there is no fundamental difference here: parity arguments would just be a subset of analogical arguments. I suspect (although I do not know) why there is this sort of confusion that you noted in your post: parity arguments are based on the way species within a genus, or subkinds within a kind, are related, but some people are inclined to read these more fundamental relationships purely in terms of similarity while others are not.

  3. Vance Ricks1:32 PM

    Sorry, but I'm getting stuck on your characterization of the "exclusive"/"inclusive" distinction.  When you talk about good analogical inferences "oppos[ing] each other", do you mean opposition in the sense of conclusions that are contradictories?

  4. branemrys2:29 PM

    The term 'opposition' in this context is simply taken from Hume. In his sense, conclusions need not be strictly and directly contradictories in the logical sense, but they need to be either inconsistent somehow or make each other very improbable. Hume, for instance, is very much committed to the claim that if you have analogies both for and against a conclusion that the inferences on both sides can be entirely reasonable and legitimate inferences. He deliberately uses this in the Dialogues to argue the limitations of analogical design arguments: the apparent goodness of the argument doesn't rule out the possibility that analogical inferences to (probably) inconsistent conclusions are also good. It's not Philo's argument (and, in Humean terms couldn't be) that the existence of these contrary analogies explaining the origin of the world show that Cleanthes's analogical inference is bad (despite the way it is sometimes described); rather, the point he is making is that, while Cleanthes's inference is entirely reasonable, you can get very different answers with inferences that are about as reasonable, or perhaps even more reasonable. That fits the Treatise account of analogy exactly: all inferences based on real resemblance are reasonable, although all are defeasible and some are more reasonable (being based on more extensive resemblances) than others.

    This contrasts with Mill, since Mill regularly identifies the fact that a conclusion is false as a reason to think the analogical inference is itself fallacious: and from the other side, if you have two opposed conclusions from analogy, that is itself proof that at least one of the two is based on an irrelevant similarity and thus a bad analogical inference. This is how Mill always analyzes arguments that he claims commit the fallacy of false analogy, for instance. The particular causal relevance conditions he requires of analogical inferences (as part of his retrictivism) make it either impossible for there to be non-fallacious analogical inferences that are opposed to each other or at least very difficult and improbable (the classificiation is only a practical guideline for sorting through accounts of analogical inference, so if Mill turned out not to be a pure exclusivist, it wouldn't really change the way the classification applies much).


Please understand that this weblog runs on a third-party comment system, not on Blogger's comment system. If you have come by way of a mobile device and can see this message, you may have landed on the Blogger comment page, or the third party commenting system has not yet completely loaded; your comments will only be shown on this page and not on the page most people will see, and it is much more likely that your comment will be missed.