Monday, March 5, 2012

Week: Meta-Theories of Explanation

Weeks 4 and 5 (and beyond!): What is an explanation?

If we’re going to be able to assess IBE’s reliability as a pattern of inference, then we better be able to answer the question, “What is an explanation?” This is more or less a reprise of our earlier question, “What are the dominant views about explanation in the literature?” We’ll approach this in two stages. First, we’ll approach the more abstract question about what we should expect from our theories of explanation (this is a kind of “meta-theory of explanation”). Then, we’ll look at specific theories of explanation.
Week 4: Meta-Theories of Explanation
We’ll begin by looking at Lipton’s quick review of the explanation literature, in Chapter 2 of his book. (I stress that this is a very broad review; the literature gets more detailed very quickly.) Here, Lipton shows that several models of explanation (reason, familiarity, deductive-nomological, unification, and necessity) fail to satisfy several important criteria of adequacy: the why-regress, distinguishing evidence from explanation, asymmetries (i.e. that A explains B typically entails that B does not explain A), etc.
I’ll pair this with two readings that take opposing views about what we should be expecting from a theory of explanation. Nickel (2010) argues against explanatory skepticism.
Explanatory skeptics deny the existence of any distinctively explanatory information within or across domains. In other words, skeptics hold that whether A explains B depends on the background beliefs, commitments, practical interests, etc. of the person who is inquiring about A or B. By contrast, anti-skeptics claim that whether A explains B depends on certain features common to all explanations, regardless of inquirers' background beliefs, commitments, and interests. 
While Nickel rejects explanatory skepticism, my collaborators and I show that his arguments are inadequate (Díez,Khalifa, and Leuridan forthcoming). While this does not show that explanatory skepticism is correct, it shows that it is defensible.
And what follows if explanatory skepticism is defensible? Nickel (p. 308) does a nice job discussing three implications of this view: (1) "the special status of understanding is a chimera"; (2) "IBE is undermined," and (3) reduction's status as a "metaphysically important relation" is put into question. Let's also add that Lipton's quick dismissal of the five aforementioned models of explanation becomes far too quick. The skeptic can argue that the criteria of adequacy he uses to knock out these five models at best show that these models aren't universal templates of explanation, but then again, the skeptic will also claim that no such templates exist. Consequently, we are free to adopt the reason model in one context, the familiarity in another, etc., and worries about their susceptibility to why-regresses, asymmetries and the like are also constrained by context-specific interests, background beliefs, etc. Now, it may be the case that these worries crop up more frequently when deploying these kinds of explanations than, e.g. causal explanations, but the skeptic argues that this is simply a contingent fact emerging from most people's attitudes towards explanations, but is not thereby written into the "nature" of explanation.


  1. I found the Nickel and Diez et al. articles rather confusing, so I want to add some musings to Meredith’s question 2, on the DN model and generalism vs. skepticism. In particular, I believe that the DN models reliance on lawfulness begs the question of explanation.

    A successful DN explanation includes three steps: we have a law, we have ancillary/contextual information, and we then reason deductively to an explanation which follows logically from the first two factors. Unfortunately, as Meredith noted, it is unclear what a law is. A law could be something as fundamental as the logical laws (e.g., law of non-contradiction or identity) or various physical laws (e.g., law of refraction, Newton's laws, etc.). I’m guessing we cannot just include the logical laws, as we really cannot derive anything from them.

    So, what constitutes a law? I think – and I want to flesh this out tomorrow – that possessing any law requires a certain understanding, which in turn entails a certain explanation. Now, the question is, is this a pernicious or benign circular?

    1. Oh and Andrew, I didn't fully read your post before posting mine, but you put what I began to talk about and then stopped myself better than I. The circularity of law as explanation. Let's talk about this today.

  2. I'm glad not to have been the only person to have found confusion at some of Nickel's points. Though there are a number of things I'd care to pick at there, I'd first like to take a look at skepticism. It seems a far stretch to believe that, in any context, the statement "Jupiter's great red spot has persisted for more than a century" would explain "Alden has an allergy to peanuts."

    More specifically, then, I'd wonder what particularly strong arguments there are for skepticism. It seems like a bold assertion, given patterns that seem to arise among explanations in almost any given context. Perhaps an interesting connected topic would be to explore principles of explanation generally used in mathematics and ethics, since these are commonly cited as resisting causal explanation. Perhaps others understand these well already, but they're beyond territory well familiar to me.

  3. I too found the readings rather difficult, but I want to focus in on the points made on asymmetry of explanation. I agree with the Diez paper that Nickel's example about the moon just isn't a very good explanation to begin with, because we also need to have a background of Newtonian mechanics.

    But, let's assume, arguendo, that the moon example is an acceptable explanation, and moreover that the universe is deterministic. This means that we're ignoring any sort of randomness at the quantum level, free will, etc. If the universe is deterministic, and if Newtonian mechanics is true everywhere, why is the explanation that the moon is here because it was at location a earlier better than the explanation that the moon is here because it will be at location b later? Are there asymmetries of explanation that do not have to do with time? If not, isn't the "problem" of asymmetry just a contingent fact regarding our knowledge of the universe?

  4. I have the same problem. My inclination (like Diez et al. and Mr. Taylor) is to say that even Nickel's 1a is not an acceptable explanation. Unless, of course, you offer it with the understanding that some kind of law is operating (e.g. Newtonian mechanics), which then is the thing that's really doing the explaining, right?

    But a law is not the same thing as a cause or a force, yes? A law is a description of what occurs. To say "The moon appears there because it was at location l earlier" seems to only make sense if it's hashed out as "The moon appears there because a law exists which stipulates that an object like the moon (with such and such a mass and velocity and such and such gravitational attractions operating upon it and so forth) which is at location l at time t will be at its present location now." But that law itself...

    Well, maybe I've just devolved into the big problem of induction. I was going to say something silly about how laws aren't good explanations unless we can describe their mechanics - e.g. Kepler had a law about how the planets moved, but Newton described the mechanism - but, a., that's just me showing my cards as liking the only-causal-explanations-for-events camp, which is controversial, and b., at a deeper level I think you can probably dissect any mechanism proffered down until you hit Humean madness. (How does gravity work? Well, maybe there are gravitons. How do those work? infinitum.)

    But anyway. If you stay up at the level Nickel's at, I don't see why 1a is actually preferable to 1b. And if you drop to the causal level (where it clearly is preferable), the asymmetry business seems to go away.

  5. I also had trouble with understanding Nickel's main reasoning behind how Generalization is "presumed" correct only because both Skepticism and Moderation "are mistaken." Like James, I wanted to get more insight on both Skepticism (and Moderation) before it's just dismissed. I found my initial tug towards being skeptic (maybe because I've studied too much Philosophy) so perhaps that's a plausible theory of explanation.