Week 3: Explanation and Induction

In addition to questions about realism, Lipton aims to settle issues about inductive inference (Chapter 1). In short, his claim is that despite the apparent diversity of inductive inferences, many of them can be captured as special cases of a pattern of reasoning called Inference to the Best Explanation (IBE). Lycan (2002) lays out a more general program, explanationism, of which Lipton’s can be seen as a special case. We’ll also read Laurence Bonjour’s short review of the major problems of induction.

            Neither Lipton nor Lycan take themselves to be tackling the classic, Humean problem of induction—what Lipton calls the “problem of justification.” The problem of justification asks, “How are we ever entitled to believe things on the basis of inductive inference?” Instead, they take themselves to be addressing the “problem of description,” in which the problem is to devise an elegant classification of the kinds of inductive inferences we take as warranted.

            Solutions to the two problems can be separated. The problem of description—at least as construed by Lipton and Lycan--only assumes that some inductive inferences appear justified, and aims to explain this appearance in terms of explanatory roles played by the conclusions of these inferences. It could turn out that this is merely an appearance of justification, in which case the problem of description would be entirely distinct from the problem of justification.

            Conversely, solutions to the problem of justification can proceed without a very detailed classification/description of our inductive practices. All inductive inferences share certain features—e.g. that the premises of an inductive inference can be true whilst its conclusion be false. Furthermore, it’s precisely these common features that drive the problem of justification. Consequently, if we can justify the assumptions concerning these features, then we don’t need any further description of the kinds of inductive inferences that we employ.

            Of course, there’s nothing that prohibits concerns about description and justification from informing one another. In particular, if we think that the “meta-philosophy” that underwrites Humean skepticism unreasonably privileges philosophical reflection over other kinds methods, then explanationism might be justified in part by its ability to provide a good classification/description of our inductive practices. (Compare: our theory of atomic structure is justified in part by the periodic table’s ability to provide a good classification of the elements.) Indeed, Lipton gestures towards this maneuver later in his book.

So why should we care about the relationship between explanation and truth? In addition to the issues about realism raised last week, considerations about explanation also promise to offer a powerful description of our inductive practices. Since induction figures prominently in nearly all of our inquiries, this means that explanation would figure prominently in most of our inquiries. Consequently, if the connection between explanation and truth is tenuous, then we have to adjudicate between several scenarios: (1) that our inquiries properly aim for explanation/understanding but only incidentally aim for truth, (2) that our inquiries properly aim for truth but only incidentally aim for explanation/understanding, or (3) that our inquiries only incidentally aim for truth and only incidentally aim for explanation/understanding. So our self-conception as inquirers is at stake! That’s why we should care.

Week 2: Explanation and Realism

Last week, I asked (among other things), “Why should we care about the relationship between explanation, understanding, and truth?” One answer to this question is that explanatory considerations are frequently used to argue for the objectivity of our discourses Specifically, IBE is used to justify various forms of realism. Broadly construed, realists claim that some philosophically controversial part of our discourse (e.g. about unobservable entities in science, about the external world, other minds, morality) refers to mind-independent truths. Realists who do this via IBE claim that these mind-independence truths best explain the relevant evidence. Hence, realists who deploy IBE claim that there is an intimate relationship between our best explanations and the truth. Indeed, anybody who uses IBE seems committed to this claim. After all, a good inference transmits the truth of its premises to its conclusion. So, endorsing IBE entails accepting that good explanations are a reliable guide to truth. (People might disagree about whether mind-independent truths best explain the relevant evidence, as Harman and Frost-Arnold do in our readings.)

            There are at least three related domains in which IBE is used to justify realism. First, scientific realists argue that claims about the unobservable posits of scientific theories (e.g. subatomic particles) are true because otherwise the success of science would be unexplained. This is called “The No-Miracles Argument,” and it has received its fair share of criticism, e.g. (Frost-Arnold 2010).  Second, we’ll skim Thagard’s (2000)’s chapter, “Reality,” to get a sense of how IBE can justify realism about certain metaphysical issues. Finally, we’ll read Harman’s “Ethics & Explanation,” which denies moral realism on the grounds that it does NOT provide the best explanation of the relevant phenomena. Finally, we’ll read Sturgeon’s “Moral Explanations” for a rebuttal of Harman’s position.

            So, one answer to the “Why care?” question is that you should care about the relationship between explanation and truth because it figures prominently in discussions about the objectivity of various kinds of discourse. To that end, it would be good to know:

(1)  Whether IBE provides a sound justification for this position.

(2)  If IBE can’t provide such a justification, whether it’s more defensible to:

a.     Reject realism (i.e. become an antirealist), or

b.     Find a different justification for realism.

Integrating my teaching and my academic blogging

So my blogging has fizzled miserably since its hot start a year ago. To revive it, I'll be weaving it into my senior seminar, Concepts of Explanation (PHIL0425). My students will help me to use this blog more frequently. To get the ball rolling, here's the first entry for the seminar.

--------

In this seminar, I want you to see how professional philosophy papers are made. In particular, you’ll be looking over my shoulder as I develop two related papers:

(1)  I’ll be weighing in on a debate about whether understanding entails true beliefs, and if so, which true beliefs such understanding entails.

(2)  I’ll be revising a theory of explanation that I’ve already put to print, largely in light of:

a.     Flaws that I see with my initial formulation; and

b.     Some views about explanation and understanding that I’ve developed since publishing that paper.

I want to stress that this is not a linear process; there will be missteps. Furthermore, these two projects have very different structures. I haven’t written a lick of the first, and haven’t even done the relevant background reading—this will be part of our task for the seminar. For the second, I’ve basically devoted my whole (though still relatively young) career to doing background reading. Yet, oddly, I feel much more prepared to write the first! Ask me about this…

To appreciate what I’m doing with these two papers, I’ll need to bring you up to speed on the following:

·      What are the dominant views about understanding in the philosophical literature?

·      What are the dominant views about the relationship between understanding and true belief in the literature?

·      What are the dominant views about explanation in the literature?

Lest the first two questions look unrelated to the third, observe that we frequently expect a good explanation to provide understanding. This leads to two more questions:

·      What does it mean for an explanation to “provide” understanding?

·      Why do philosophers care about the relationships between understanding, explanation, and true belief?

One of our textbooks, Peter Lipton's Inference to the Best Explanation (2004) provides a nice touchstone for addressing many of these questions. By supplementing several of Lipton’s chapters with additional readings, I’ll get you up to speed on these questions.

Weeks 1 through 3: Why care?

Week 1: Preliminaries

For the most part, we’ll be focusing on the epistemological dimensions of explanation. In other words, how do we know that we have a correct explanation, and how do correct explanations advance our knowledge? Since many (all?) of you aren’t familiar with epistemology, we’ll read Sturgeon’s quick rundown of its major ideas. Thereafter, we’ll delve into the concept that weaves the ideas of explanation, understanding, and truth together: Inference to the Best Explanation (IBE). Its general form is:

            p

            q best explains p.

            So, (probably) q.

A more precise version of both IBE and its relationship to induction comes from Lycan (2002), Thagard (1978), and subsequent chapters in Lipton’s book. We'll read these as well. Here are some things that would be good for you to think about:

1. Think of two potential explanations for the same phenomenon: one of these should be very plausible and the other really outlandish.
2. Do the "theoretical virtues" that Lycan and Thagard mention favor the plausible one?
3. Can you think of cases where IBE would lead us astray, i.e. where our best explanations aren't likely to be true? What are the more general concerns ("structural defects") with IBE this raises?

Longino on the Theoretical Virtues

So I’ve just read Helen Longino’s work on the theoretical virtues:

Longino, H. E. (1994), "In search of feminist epistemology", Monist 77 (4):472-485.

——— (1995), "Gender, politics, and the theoretical virtues", Synthese 104 (3):383-397.

——— (1997), "Feminist Epistemology as a Local Epistemology", Aristotelian Society Supplementary Volume 71 (1):19-36.

(For those who are interested, I don’t think you lose much if you just read the 1997 article, as there’s quite a bit of redundancy, and, in my estimate, Longino’s argument gets clearer over time.)

In these articles, Longino argues:

(1)  There are traditional theoretical virtues that work in certain contexts, e.g. empirical adequacy, conservatism (external consistency), simplicity, unification, scope, fruitfulness, and refutability.

(2)  There are also feminist theoretical virtues that work in other contexts, e.g. empirical adequacy, novelty, ontological heterogeneity, complexity/mutuality of interactions, applicability to human needs, and decentralization/universalization of power.

(3)  Neither of these sets of virtues is intrinsically epistemic, i.e. it doesn’t appear that the virtues have a very obvious link with truth.

(4)  Rather, the epistemic significance of these virtues is always relative to the cognitive goals of the communities that use them.

a)     In feminist inquiry, this is the goal of “revealing gender,” e.g. identifying gender biases that have been operant in scientific inquiry.

b)    It’s unclear what cognitive goals Longino takes the traditional virtues to serve, but both feminist critics and advocates of these virtues (e.g. Hugh Lacey) have taken control and manipulation (e.g. as found in experimental settings) to be a plausible candidate.

(5)  Longino then situates these ideas within her four social-epistemological criteria:

a)     Provision of venues for the articulation of criticism;

b)    Uptake (rather than mere toleration) of criticism;

c)     Public standards to which discursive interactions are referenced;

d)    Equality of intellectual authority for all (qualified) members of the community.

(6)  “Within this scheme the traditional and alternative virtues constitute partially overlapping, but distinctive sets of public community standards [i.e. item (5.c)].” (1997, p. 29)

Longino’s argument, particularly premise (3), leaves much to be desired. But I won’t quibble with it in the current project, since that would amount to an external critique, and my goal is to provide an immanent one. I want to suggest that, unlike the feminist virtues, the traditional virtues play a special role in item (5.b), i.e. uptake of criticism. Without playing my hand just yet, the virtues need not be guides to truth, but only guides to warranted acceptance (not belief!) of a theory.

Fine-Tuning Our Distinctions

Thus far, I’ve framed the S&V literature in terms of "doing good science," viz.

1. A value is constitutive of science if and only if it is necessary for doing good science.
2. A value is contextual in science if and only if it is sometimes (but not always) useful for doing good science.
3. A value is invariant in science if and only if it is always useful for doing good science.

I think this is too loosely formulated. Since I’ll be focusing primarily on the theoretical virtues, I’m interested in a much narrower set of values, namely those that are predicated of theories as criteria of acceptance. It's my sense that the following fairly glosses what most have meant by the constitutive-contextual distinction:

1. A value V is constitutive of theory acceptance if and only if, for all theories T, correctly identifying V as a property of T is necessary for being justified in the acceptance of T. 
2. A value V is contextual in theory acceptance if and only if, for some (but not all) theories T, correctly identifying V as a property of T provides some justification for the acceptance of T.

If that's right, then we have two important dimensions by which to catalog values in science. The first is the scope of theories to which the value applies (which I've italicized), and the second is the value's justificatory role (in boldface). Minimally, invariant and constitutive values have the same scope: they apply to all theories. 

However, I want to suggest that we consider three justificatory roles: necessity (as illustrated by our new definition of constitutive values), contributing (as illustrated by our new definition of contextual values), and sufficiency (which provides a natural foil to necessity). This would then suggest a stronger and a weaker brand of invariantism about values:

3.  A value V is invariantly strong in theory acceptance if and only if, for all theories T, correctly identifying V as a property of T is sufficient for being justified in the acceptance of T. 

4. A value V is invariantly contributes to theory acceptance if and only if, for all theories T, correctly identifying V as a property of T provides some justification for the acceptance of T.

I note in passing that we can do the same thing with contextual values. In this case, i.e. 

2 (revised). A value V contextually contributes to theory acceptance if and only if, for some (but not all) theories T, correctly identifying V as a property of T provides some justification for the acceptance of T.

5.  A value V is contextually strong in theory acceptance if and only if, for some (but not all) sufficient for being justified in the acceptance of T. theories T, correctly identifying V as a property of T is

6. A value V is contextually necessary in theory acceptance if and only if, for some (but not all) necessary for being justified in the acceptance of T. theories T, correctly identifying V as a property of T is

These won't be part of my story, but it's useful to observe that we should have a six-part distinction instead of a bipartite one, as has been the norm. As for me, I'm leaning towards claiming that the virtues are invariantly strong, though I'll have to do more work before I make a firm decision on this. Otherwise, I'll claim that they're invariant contributors.

More on the Constitutive-Contextual-Invariant Distinction

I just hunted down the exact Longino quote that baptizes the distinction between contextual and constitutive values, and wanted to compare it to yesterday's musings:

I will call the values generated from an understanding of the goals of science constitutive values to indicate that they are the source of the rules determining what constitutes acceptable scientific practice or scientific method. The personal, social, and cultural values, those group or individual preferences about what ought to be, I will call contextual values to indicate that they belong to the social and cultural environmental in which science is done.
Science as Social Knowledge, p. 4.

This is Longino’s clearest account of the distinction. Let’s try to unpack it. First, there is an assumption that constitutive values are “the source of rules.” Moreover, they appear to be the source of rules that tell us what to do, as "practices and methods" are naturally seen as actions. However, constitutive values are not the same as goals, as the latter “generate” the former. Goals naturally function in rules as follows:

            If one wants to achieve goal G, then one ought to do A.

But where are the values in this? Either G is valuable, or A is (instrumentally) valuable because it is a means to achieving G. That suggests the following:

(L1) A value is constitutive of science if and only if it is either a goal of science or it is a necessary means of achieving a goal of science.

I’ve added that the means must be necessary, because otherwise, a means works in some contexts but not others, which prima facie, sounds like a contextual value. Note that this gets us something very close to yesterday’s gloss. Now let’s tease out Longino’s definition of contextual values:

(L2) A value is contextual in science if and only if some but not all practitioners of a science take it to be valuable.

This is not quite the same as yesterday’s account of contextual values, as that made no reference to whether practitioners take something to be valuable or not. This yields an odd consequence: if some scientist does not take a constitutive value to be valuable, then it is contextual. But this doesn’t leave any room for the scientist to be just plain wrong in his evaluations. For instance, logical consistency is presumably a non-negotiable value, so a contradiction-loving scientist seems to be doing something more suspect than expressing an “individual preference,” as Longino puts it.

            Yesterday’s constitutive-contextual distinction blocks this, since it requires the value to be useful for doing good science some of the times. Something can be useful even if nobody takes it as such and somebody can value something that is useless, so scientists’ attitudes don’t determine contextual values. As a result, I’m sticking with that distinction, since it seems tidier. Moreover, I’m holding fast to distinguishing invariant and constitutive values.

Science and Values, Part 1: The Constitutive-Contextual Distinction

I’m beginning to research my next essay, which touches on that thorny issue of “science & values” [S&V]. I hope to be blogging quite a bit about this, as I have quite a bit of background reading to do.

I begin with a well-known distinction in the S&V literature, which is primarily associated with Helen Longino.

(1)  A value is constitutive of science if and only if it is necessary for doing good science.

(2)  A value is contextual in science if and only if it is sometimes (but not always) useful for doing good science.

While it marks a subtle change, I want to suggest that we offer something that falls in between (1) and (2):

(3)  A value is invariant in science if and only if it is always useful for doing good science.

What’s the difference between constitutive and invariant values? It is possible for something to be always useful and yet not necessary for doing an activity well. For instance, having an accurate three point shot is always useful, but not necessary, for being a good basketball player.

In the paper I’m working on, I’ll suggest that several “theoretical virtues,” e.g. simplicity, scope, precision, etc. are invariant scientific values. Moreover, I’ll be arguing that some of the most vociferous critics of the virtues should accept this claim, given their other commitments. I suspect that they have resisted this claim because the constitutive-contextual distinction poses a false dilemma. The true dilemma is between invariant and contextual values, with constitutive values simply being a particularly demanding form of invariant values.