Understanding

Thus far, we've explored the idea that explanations guide us to the truth. To the extent that we've entertained challenges to that view, it's the explanatory skeptic's claim that explanations lack any uniform properties that could reliably underwrite inferences. However, the explanationist has three ways of sidestepping that worry.

The first is a direct attack on skepticism, as Nickel does. Similarly, even if Nickel's arguments don't work, an explanationist can start to look at more concrete skeptical views--such as my accountabilist and expressivist ideas--and find faults with them. Second, even if explanationists concede a good deal to explanatory skepticism, they can pick out a class of explanations that do underwrite inferences. Here, Lipton hones in on the most plausible candidate: causal explanation. Yet a third explanationist tactic is to claim that it is not the concept of explanation that underwrites Inference to the Best Explanation, but rather the notion of the best that guides us to truth. This invariably leads to the question, "What makes one explanation better than another?" Lipton's answer is "Loveliness," or that which provides the greatest potential understanding.

Lipton has already told us a bit about understanding--specifically it is knowledge of an explanation. (Recall his claim on p. 30: "Understanding is not some sort of super-knowledge, but simply more knowledge: knowledge of causes." Following Lipton's move in this week's chapter, we'll abstract the causal bit.) This would suggest that the best explanation is the one that would provide the greatest potential explanatory knowledge. Here "greatest" might mean "most," suggesting that we infer the explanation that would allow us: (a) to know more about it than its rivals, and (b) to know more explanations than would its rivals. However, "greatest" might also mean "best," in which case we need to have some way of judging what makes one body of explanatory knowledge better than another. Undoubtedly, (a) and (b) figure in such judgments, but others include theoretical virtues such as simplicity, conservatism, analogy, and consilience. (Arguably, consilience is exhausted by (a) and (b).)

However, this all presupposes that understanding is a species of knowledge, which brings us to squarely back to the crash course in epistemology from Week 1 (Scott Sturgeon's article). Recall the Gettier problem: one may have justified true belief, yet if one's belief easily could have been false, then a person does not know what she believes. For instance, suppose that I see an animal with white fluffy fur at Duclos Sheep Farm down the road from the College. Since I have seen sheep there quite regularly, I believe that there is a sheep in the field. Hence, my belief is justified. However, unbeknownst to me, I am seeing a shaggy dog who is perfectly occluding my view of a sheep. So although my belief is justified and true, it appears that I do not know. That is a Gettier case.

Some, most notably Jonathan Kvanvig, have argued that understanding is not a species of knowledge precisely because it is unaffected by "Gettier Luck." Kvanvig's case runs as follows: suppose that you pick up a library book at random on Comanche history. You read the book diligently, grasping how various explanations make sense of a large and well-studied body of evidence. When asked about Comanche history, you fire back a bevy of true answers. Wouldn't you say that you understand a good deal about Comanche history, including particular events in that history, e.g. why they dominated the Southern plains? Now, Kvanvig adds, suppose that the book you pulled from the library happened to be the only one on the shelf that had accurate information about Comanche history. Had you grabbed any other book on the Comanche, your beliefs would have been false. So, just like the sheep example, you don't know, because your belief could have easily been false. Yet your understanding seems unblemished by this fact.

(The analogy is sheep:shaggy dog::good Comanche book:shoddy Comanche book.)

Grimm critiques, and Morris defends, Kvanvig's stance on lucky understanding. [Where do you stand? Why?] Obviously, Grimm's view is more in line with Lipton's, but it is an interesting (and relatively unexplored) question as to how the possibility of lucky understanding affects IBE. Minimally, it would appear to lower the standards on what counts as the best explanation, as one would not need evidence that one's belief in the explanation could have easily been false. However, this presumably increases the riskiness of using IBE as a rule of inference. So if there are reasons for thinking that understanding is compatible with luck antecedent of any considerations of IBE, then perhaps the proper lesson is that IBE is not a reliable rule of inference.

Update on my Expressivism Paper
Our conversations from last week have prompted me to take a step back, and ask a more elementary question, "What is a pragmatic theory of explanation?" I'm also looking at the kind of argument that could provide good reasons to favor such a theory. It's turning out to be a fascinating exercise. I'll share once I'm further along.