Monday, March 12, 2012

An Expressivist Theory of Explanation?

So you’re now in a position to see the first of the papers that I’ll be working on this semester. This paper aims to take my “Contrastive Explanations as Social Accounts” and turn a rather clumsily formulated social epistemology of explanation into an expressivist theory of explanation. I’ll go through this in four stages. First, I’ll give you expressivism in a nutshell. Second, I’ll give you the three core ideas from my “accountabilist” approach to explanation. Third, I’ll briefly sketch how to convert these into an expressivist account, and fourth, what benefits it holds.

Before proceeding, let me add that all of these thoughts are brand spankin' new. I literally began piecing together these ideas Sunday afternoon, and have had a steady diet of interruptions throughout their formulation. What I really want from you is a full battery of questions about this. I also want you to switch hats, from being altogether critical of this idea to really trying to inhabit and see it to its proper development. 

For all I know, this is a hair-brained idea! This is actually a good thing to learn about the writing process. For every published paper that I have, there are tens of thousands of words that I throw to the flames.

1. Expressivism in a nutshell. Traditionally, accounts of explanatory claims are committed to descriptivist semantics. In other words, explanatory claims were taken to describe real relations between explanantia and explananda, and assertions thus express cognitive mental states. Cognitive mental states, paradigmatically beliefs, purport to represent the world in a particular way. By contrast, expressivism claims that explanatory claims express a kind of mental state, and more specifically a non-cognitive state. Unlike cognitive states, non-cognitive mental states (e.g. intentions) purport to make the world become one way or another, and figure first and foremost in our actions.  
Different expressivists deploy different kinds of non-cognitive mental states.  Gibbard (1990),  the foremost expressivist in metaethics, favors a complex (i.e. a combo of both cognitive and non-cognitive) mental state of ‘‘norm-acceptance’’. Concerning attributions of rationality, Gibbard (1986, 479) writes,
Thinking X rational…is a combination of a normative state and a state of factual belief. It is accepting a system N of norms such that one believes the subject to be in circumstances for which N permits X.
On Gibbard’s expressivism, the mental state expressed by claims of the form “X is rational” is complex, involving both a factual belief—that the subject is in circumstances for which N permits X—and a non-cognitive state—the acceptance of the system of norms N.

2. Contrastive Explanations as Social Accounts. I’ll discuss the core ideas from this older paper, offering critical commentary along the way:
·      An audience A demands from a person S an explanation of p rather than q if and only if:
(1)  A undertakes commitment to the topic p;
(2)  A undertakes commitment to not-q;
(3)  A takes S as committed but not entitled to p and not-q; and
(4)  Given the uncontroversial commitments u that entitle one to p, a member of A’s epistemic community would be committed to q if she were committed to a controversial claim c.
Problem: controversy is not necessary for explanation. Example: it’s unclear why controversy is necessary for explaining “Why is the sky blue rather than green?” Would anybody be committed to the sky being green under any controversial commitments of which we can conceive? Minimally it appears ad hoc to insist on this.

·      S purports to explain p rather than q with h relative to controversial commitments c and uncontroversial commitments u if and only if:
(1)  S undertakes a commitment to h, p, and not-q; and
(2)  S takes h to entitle the strongest variant of p, not-q, c, and u to which S is committed.
Here, the strongest variant of a set of propositions will be the largest subset of those propositions closed under deduction.

·      S correctly explains p rather than q with h if:
(1)  S is entitled to h; and
(2)  h actually entitles the strongest variant of p, not-q, c, and u to which S is committed.
Problem: not clear how to articulate “actual entitlement.” I offered three different stories (dialectical, intersubjective, and objective), but find none to be satisfying.


3. From accountabilism to expressivism. I suggest the following expressivist amendments to my accountabilist view.

·      An audience A demands from a person S an explanation of p rather than q if and only if:
(1)  A believes p;
(2)  A believes not-q;
(3)  A accepts epistemic norms N;
(4)  A believes that S is in circumstances C;
(5)  According to N, if S is in C, then
a.     S is committed but not entitled to believe that p and not-q; and
b.     S is entitled to believe that p.
(6)  There exist an alternative set of epistemic norms N* such that according to N*, if S is in C, then S would be entitled to believe that q.
Note that there is no longer any reference to controversy or commitments, so the problems with the first formulation have evaporated. However, we will need to index all explanations to a system of norms N, a rival system of norms N*, and a set of circumstances C.

Next, let’s look at purported explanation.
·      S purports to explain p rather than q with h (relative to N, N*, and C) if and only if:
(1)  S believes that h, p, and not-q;
(2)  S believes that S is in C+, where C+ is the strongest variant of C that S believes;
(3)  S accepts a system of norms N+, where N+ is the most comprehensive intersection of N and N* that S accepts; and
(4)  According to N+,
a.     If S is in C+, then S is entitled to h
b.     If S is in circumstances C+ and h, then S is entitled to believe that p and not-q.

Now we get to the kicker: correct explanation is just purported explanation with the additional caveat that all of S’s beliefs are true. (Here I’ll bracket issues of realism about explanantia.) But observe what this means: take any four claims that could stand for p, not-q, h, and C+. There are no further true beliefs about whether these things could stand in a “correct” explanatory relationship; only norms that different people accept that then dictate these explanatory relationships. (Which isn’t to say that everything explains everything; such a reckless dictum is curtailed on this account largely by C+ and by any contingent overlap in our epistemic norms.)


4. Implications of this view.
A. Dialectical Intuitions. This is all very speculative, but I think the main upshot of this view is that it inherits Chrisman’s (epistemic) expressivist solution to what he calls the “dialectical intuitions problem.” As he writes:
…by maintaining that knowledge claims express states of norm-acceptance rather than relativized factual beliefs, the epistemic expressivist …gains a second axis of possible opposition or agreement. That is, he can account for the intuition of cross-context opposition and agreement by claiming that the speakers are expressing pragmatically opposed or concurring states of norm acceptance, rather logically contradictory or identical descriptive beliefs (244).
In other words, there are still debates to be had about whether something is explanatory or not, but the debates might be factual (if they deal with the beliefs in question) and/or they might be pragmatic (if they only deal with how the contents of these beliefs are “explanatorily relevant” to one another).

B. Explanatory Skepticism. This gives us a new and strong version of what Nickel called explanatory skepticism. What counts as a suitable explanatory contrast, as well as what counts as satisfying a relevance relation with respect to such a contrast, is not determined by substantive semantic constraints (at least if semantics must be descriptivist), but must now involve pragmatic considerations about the epistemic norms that we accept. Its major upshot is that it provides a more precise account of how context affects explanatory claims. The specific contextual mechanisms involve the norms that the explainer and her audience accept.
Note that we need to distinguish explanatory skepticism from explanatory pluralism. If it turns out that everybody accepts the same norms when explaining, then although the view will still be skeptical (because the major constraints on explanations will be non-semantic and hence non-substantive), but it will not be pluralist. There will only be one way of explaining things: whichever way the preferred set of epistemic norms dictates.
However, we also have further reason to think that the relevant norms exhibit a good deal of variation. Lipton already countenances geometrical and non-contrastive explanations in addition to his preferred causal model. Furthermore, “Contrastive Explanations as Social Accounts,” particularly its appendices, also provides independent support for explanatory pluralism. There are lots of different kinds of explanations within science (since then I’ve found even more!), so we can only imagine that adding math and ethics into the mix will make explanation all the more motley.

C. IBE. As a skeptical view, explanatory expressivism is not terribly congenial to IBE. In particular, it’s rather odd that a canon of inference would counsel not only the adoption of beliefs but also the acceptance of norms. Alternatively, it may point to the fact that IBE only makes sense if you accept certain epistemic norms. The question then becomes what norms they might be, and what non-acceptance of these norms entails. Consequently, in being entitled to the explanans—the signature inferential benefit of IBE—we may find ourselves entitled to far less than belief in the explanans’ truth, since a more modest set of norms would provide entitlement to something which falls short of truth, e.g. empirical adequacy.

As I said, this is all very speculative and new. But the exciting bit is that we get to think about these ideas together and try and hammer them out.


  1. Perhaps it's the juxtaposition of Lipton's chapter on causal explanation with the other two articles that's causing what will amount to nothing more than a bout of confusion, but I'm wondering whether accepting an accountabilist or expressivist view of explanation has any effect on what traits we look for in a good explanation.

    It seems to me that, although accountabilism does and an expressivist theory would involve different explanatia with different qualities depending on context in addition to contrast, each is nevertheless compatible with the generalist, moderate, and skeptical views that we discussed last time. After all, is there any reason why certain features should be expected of our explanations across all standards that we do accept, regardless of context?

    Am I missing some key detail here, or are these two qualities that I'm trying to compare simply on different levels and essentially non-interactive?

  2. I want to make a number of points, which are probably not related.

    The first is that I remain stubbornly, fundamentally, and possibly stupidly skeptical of the bundle of sticks example of a non-causal explanation. The reasoning that because there are two horizontal dimensions and only one vertical one means that sticks are more likely to be aligned along a horizontal axis strikes me as akin to reasoning that because I could come to class or not come to class today (2 options), each is equally likely. It seems to me that you need a whole host of additional information. In my example, you would look at my past pattern of class attendance and probably conclude that it's not a 50-50 chance. For the bundle of sticks, it does turn out that the sticks are more likely to be aligned horizontally, but this has to do with the way the sticks move through the space, not the mere existence of a 2:1 ratio of dimensions in favor of the horizontal.

    The second point gets at Andrew's question 1. The reason that he might prefer the superficial account is that framing all constrastives as incompatible will lead to a multiplicity of possible framings. Take the paresis example. If we frame "Why did Jones rather than Smith contract paresis?" as an incompatible constrastive, it could take three forms:
    1) "Why did Jones get paresis and not Smith, rather than Smith and not Jones?"
    2) "Why did Jones get paresis and not Smith, rather than both?"
    3) "Why did Jones get paresis and not Smith, rather than neiher?"
    That said, I think that all why-questions are fundamentally expressing non-compatible contrastives. First, I think that 1) is irrelevant, because that's not really the question that I think one is asking in this situation. But 2) and 3) can be combined in the following way: "Why is their difference between Jones and Smith regarding the attribute having-paresis, rather than sameness?" Our background beliefs make this clear. Jones and Smith are, to our eyes, physically similar individuals. There is no discernible reason why one should get paresis and not the other. The additional fact of Jones' syphilis explains why apparently similar individuals were actually different, and in a crucial way.

    I also want to tackle Andrew's second question. Khalifa and Lipton are not incompatible on this point. Khalifa is making the (correct) observation that believing that the fact p is true and the foil q is false is, necessarily, believing the conjunction p and not-q. That's not really a fundamental point of his argument as I take it, just a clarification for his readers.
    Lipton is making a more complicated point; he is saying that to explain "p rather than q" is not always to explain "p and not-q." His argument is this: I can explain why "Jones rather than Smith has paresis" by reference to Jones' syphilis and Smith's lack of syphilis because only someone with syphilis can get paresis (this is not how the term "paresis" is used in the medical community today, btw). But that's not a very good explanation of the "p and not-q" formulation because it doesn't explain Jones' paresis as a standalone fact. This is because even for someone who has syphilis, the chance of getting paresis is fairly small. That seems basically right to me, but I don't think Khalifa is arguing that point.

    1. Please ignore my grammatical error in paragraph 3 (of 4). I clearly meant "there" and not "their," and it pains me that Blogspot does not have an edit feature.

  3. I share Josh’s skepticism for the fundamental reason that I really hold that probability is not something in the world – it’s not a force that does anything. Someone would need to convince me that probability is not merely a description of frequency (either observed or anticipated.) Otherwise – if you use probability to explain why more sticks are aligned horizontally, what do you do when by some freak coincidence more sticks are aligned vertically? It seems that a causal explanation is really what is called for in both cases. Probability is a property emergent from underlying causes.

    Josh, I think your point about the p & ~q conjunction is spot on.

    I’d like to address Andrew’s very last question as well, since it’s something I’ve thought a lot about in the moral realm and so it also came to mind for me while reading Chrisman. Expressivists (and other sorts of non-cognitivists) talk a lot about how moral (or whatever) type statements ARE TO BE understood, as if they might have some sort of meaning of which their speakers are UNAWARE. So whether and how what we THINK we are doing with an explanation can be separated from what we ARE doing is an important point. My own view is that people are often, but certainly not always, very unaware of what they really mean or could mean by certain speech-behaviors – that often they are simply playing by the intuitively learned (but not necessarily explicitly understood) rules of some language-game. So if we take something like accountabilism to be ultimately be a good account (ha), does that mean people need to feel like, or believe, that their explanations boil down to accounts? I think I say no, but in some ways that feels odd.

    “What do you mean?” “Well, I don’t know, I’m just saying – “

    I think “I’m just saying,” is a telling expression. I think it reveals, on a deep level, “I just said that because it’s the sort of thing one says in that situation.” It’s very possible we can develop a sort of explaining-behavior which we do automatically in certain conversation types without having any sort of theory, sophisticated or otherwise, about what KIND of information we’re actually conveying.

    But I have digressed a bit.

    PS. I’m sorry for shouting, I’m used to a somewhat promiscuous use of emphatic italics and caps-lock seems to be the only useful analog here.

    PS. I think about this probably at least once a class:

    Relevant part starts around :54. Be warned of a bit of language.

  4. As we are getting exposed to different theories of explanations, i am getting confused about their differences. For example, how is Lipton's Causal Model fundamentally different from Mill's Method of Difference? Or how does Litpon's Causal Model represent a real advance over Mill's model which is basically a Causal-Inference Model of Induction?

  5. I'm going to echo my shared agreement with Taylor on the impact probability has on explanation. I also agree that it is merely a representation of a statistic, and (outside of a 100% probability which truly isn't probability in any sense) not a causal reason for why things occur.

    My question is though: What other forms of "probability" type concepts exist in our current culture? What other things do we use to explain certain situations, but upon closer analysis, they don't actually present a causal relationship with what they explain?