G. Naturalism & Normativity > Stump on "Pierre Duhem's Virtue Epistemology"

David Stump’s recent paper, available in our library of papers, “Pierre Duhem’s Virtue Epistemology” is an exciting addition to the VE literature for a number of reasons. It is very interesting to see an appeal to the virtues in epistemology that pre-dates the current interest in VE by 30 years or more; and from a philosopher working on ostensibly different issues. David’s paper also resonates with many of the issues we have been discussing and may move us forward in understanding some of the difficult questions that arise within VE. Below I will (all too briefly) summarize the main arguments in David’s paper, and will then consider how Duhem’s VE bears on The Value Problem and the issue of Weak/Strong VE .

The main argument of David’s paper is nicely captured here: “Duhem’s holism provides a basis for arguing that reliabilism is not sufficient to characterize knowledge and that the virtues of epistemic agents need to be taken into account in epistemology.” (Stump, pg. 16) Most of us are familiar with Duhem’s well known argument that scientific theories meet the tribunal of experience as a whole, and thus no hypothesis can be directly refuted by recalcitrant observation. Scientists always faces the choice of which hypothesis from the mass of theory (including auxiliary hypotheses) being tested should be abandoned. Theory choice in science is underdetermined by empirical evidence. A similar point is that every scientific theory (T) has empirical equivalents, and our choice between (T) and its equivalents is, ex hypothesi, underdetermined by all available empirical evidence.

Underdetermination arguments raise plenty of interesting issues, but David’s paper is more focused on Duhem’s proposed solution. David draws our attention to the much less known point that, in the face of underdetermination, Duhem appeals to the “good sense” of scientists to make theory choice determinate. Moreover, good sense is a decidedly moral norm, as it includes virtues such as fairness and impartiality (Stump, pg. 2) The good sense practiced by scientists “allows scientists, like judges in a legal setting, to be able to weigh evidence and to be fair and impartial.” (Stump, pg. 2) Good sense is a non-rule governed set of virtues that allows scientists to be neutral arbiters of truth, and scientific theories to be objective. For Duhem, a scientific theory becomes choiceworthy by being the theory that scientists exercising good sense would choose.

Duhem thus advocates an agent centered epistemology. As David says, “The source of this objectivity is the epistemic agent – the scientist who acts as an impartial judge and makes a final decision.” (Stump, pg. 11) Surprisingly (at least to me), we see that Duhem advocates a form of Virtue Epistemology. David goes on to argue that Duhem’s VE contributes to current work in VE by showing the inadequacy of reliabilism and the necessity of basing epistemology on the virtues. While there are too many good issues here to hope for a comprehensive discussion of the paper, I will focus on a few points below that I found most interesting about Duhem’s VE.

VE & Reliabilism
I want to start by looking at the argument against reliabilism David finds in Duhem’s work. A reliabilist will claim that following methods that generally lead to truth is both necessary and sufficient for knowledge. As David notes, Duhem’s instrumentalism has him more concerned with useful theories than true ones. But, it seems clear that reliabilism alone will not allow us to adjudicate between empirically equivalent theories, or to resolve underdetermination problems, even if our concern is for true, rather than useful theories. If T and T-1 are empirically equivalent theories, the methods that produce T and T-1 will presumably be equally reliable. Reliability thus leads us to recognizing two (or more) equally justified theories, but gives us no resources for converging on a single theory. This is not a desirable position, especially when T and T-1 are incompatible theories. With Duhem’s virtue theory in hand, we can explain the rationality of theory choice in the face of underdetermination. Virtue epistemology thus has a decided advantage over reliabilism.

While the scope of this argument may be limited to scientific rationality, some version is likely to have application to the broader concerns of epistemology. Below I pursue a few points that resonate with issues we have discussed in this forum.

The first casualty of Duhem’s move to an agent centered epistemology appears to be Evidentialism. David makes this point about Inductivism, but I believe the same holds for Evidentialism. Evidence alone is never sufficient to uniquely justify a set of beliefs, or for choosing one theory over its competitors. Thus, whatever value a scientist’s good sense brings to a scientific theory, it will not be an evidential value. The virtuous scientist cannot say that a theory is deemed choiceworthy over competitors because it is better favored by the available evidence. How, then, would we explain the stronger epistemic credentials of theories favored by good sense, or of knowledge over merely true belief?

This question bears on the Value Problem. Duhem appears to agree that reliability alone cannot explain the additional value of knowledge over merely true belief. But, how does his appeal to good sense manage to do so? David emphasizes that good sense is seen as a decidedly moral excellence by Duhem.(Stump pg. 5) How does a moral excellence of a scientist confer epistemic value on his science, especially if he cannot make any appeal to increased evidential support? One could say that a theory gains evidential ly when favored by moral virtues, but this cannot be the case when we are choosing between theories that are equally supported by all available evidence.

David’s paper suggests that Duhem’s answer would be that a theory favored by the moral virtues of impartiality and fairness gain the epistemic virtue of objectivity. This strikes me as a promising approach to the Value Problem. While we cannot say that an objective theory has stronger evidential credentials, objectivity is an epistemic value of a theory conferred by moral virtues of the agent, and is not conferred by mere reliability.

I’m curious to know what others think of this line on the Value Problem.

I am also interested in whether Duhem sees the virtues of good sense as enablers of knowledge or as constitutive of what knowledge is. In the former case, Duhem’s VE may be a weak one, as virtue is not involved in defining what knowledge is. In the latter case, Duhem would be advocating a strong version of VE, as character traits of the agent are part of the definition of knowledge. If Duhem is giving us a strong VE, then, because good sense is a moral excellence, he appears to be arguing for a radical broadening of epistemology along the lines of Zagzebski. I wonder if his intentions are this radical. If the Duhem is giving us a weak VE, then the potential solution to the Value Problem above appears to be undermined. If good sense merely enables us to get to knowledge, but is no part of its definition, then we cannot claim that good sense is what explains the added value of knowledge over merely true belief.

I conclude by briefly noting how different Duhem’s solution to underdetermination is from Quine’s, his partner in crime. Quine’s naturalism favors acquiescing in the mother tongue (or home theory) when confronted with a multitude of empirically equivalent theories. This move is also favored by conservatism in science. What strikes me is that Quine seems to advocate partiality and bias to resolve underdetermination, while Duhem appeals to impartiality and neutrality. Duhem’s approach has a better ring to it for sure!

Thanks again to David for a such a nice and unexpected contribution to Virtue Epistemology. I’m interested to hear what other JanusBloggers think of the issues.
September 28, 2006 | Registered CommenterAbrol Fairweather
David and Abrol, Thanks for the very interesting paper and Post. Drawing attention to the importance of bon sens for Duhem does indeed seem pertinent to VE, and the issue of the relationship between personal intellectual virtues and “cogntive values” in science is an unexplored topic, it seems to me. The contrast of holism and reliabilism seems to have merit, and I can certainly see how Duhem’s inclusion of his moral values, and the idea of the “whole person” being responsible for theory choice, indicates connections with a specific kind of strong and agent-focused form of VE. For all these reasons the paper breaks interesting new ground.
I’m not sure my coments here will address the concerns that Abrol, in his post, is most interested in—VE and reliabilism, the value problem, and virtue as enablers vs conditions on knowing. Maybe my concerns are more basic, needing to be satisfied before I can make those connections. But I have two points of constructive criticism to raise.

1. The underdetermination of theory by data vs the underdetermination of theory choice by methodological rules. This is a distinction Laudan makes, that he thinks a lot of post-Kuhnian discussion tends to conflate. This may be a worry more about Abrol’s synopsis of the paper. The point that “every scientific theory (T) has empirical equivalents, and our choice between (T) and its equivalents is, ex hypothesi, underdetermined by all available empirical evidence,” seems really to be about the underdetermination of theory by data. This may be true enough, but isn’t the real issue, is it? It needs to be set aside, I think, in order to focus on what is the issue, which is that ‘empirical adequacy’—not always but sometimes—isn’t sufficient to indicate preferability over two or more extant explanatory theories. This is the issue of the underdetermination of theory choice by methodological rules. Here Laudan would say that sometimes it does, sometimes it doesn’t-instances of this kind of underdetermination are ocassional, while the other kind is universal. But its when we conflate the two that we get something like “Theory choice in science is underdetermined by empirical evidence,” which seems true (but trivially so) about the one kind of underdetermination, and false about the other (because there are plenty of cases where ‘empirical adequacy’ is enough to determine a preference for the other theory over the other. So, in other words, I’m just saying I agree with Laudan insofar as its important to be clear about the distinction, and that the issues you discuss should really be addressed to a certain restricted range of cases where both theories equally-well ‘save the phenomena’ and where the scientist therefore has to turn to ampliative criteria (cognitive values) in order to determine (and here in a more ‘all things considered manner) preferability between the competing theories.

2. Personal virtues vs virtues of theories (cognitive values).
The relationship between these fascinates me, but I worry that the transition from virtues of theories to personal virtues is too quick, or there’s a missing argument. Recall that Kuhn like Duhem said that theory choice isn’t a matter of ‘logical algorythm’ but of “values at work” in science. Hence science isn’t value free as the received logicist tradition had wanted to take it. Kuhn’s reasons clearly involve aspects of methodological holism and there’s a clear line from Duhem through Lakatos on the importance of the recognition of a moderate methodological holism in science. [As a sidenote, I’d personally leave Quine out of it, because the Quinean quagmire of “semantic” holism isn’t something that one needs to get drawn into here]. Now Kuhn (and Quine) lists several cogntive values as those values at work in theory choice—simplicity, comprehensiveness, unfying power, etc. But here’s the point. Kuhn said scientists can and do sometimes disagree about 1) what these terms mean, 2) how they apply to the particular choice/case, and 3) how to weigh and balance them in the particular case. But he was conservative enough to say scientists don’t really agree much about what values should be on the list. They are 1) communal values, and 2) virtues of “theories.” I worry that in the paper and the post, this gets somewhat negelcted. Kuhn would agree that the choice of theories is something that must be made all things considered, and that the ‘balancing’ is something that goes on inside the individual who makes the choice. But still, “simplicity” is a virtue of a THEORY, not a personal virtue. Ditto for “fertility,” “unifying power” and the rest. They are about what makes one theory objectively preferable to others, (in a way that I guess a reliabilist would want to say is related to reliability in a way that may comit them to viewing it as some kind of macro-value). But though I wouldn't follow the reliabilist story whole hog here, I would worry that a vital social/communal aspect of these values is getting ignored in any form of VE the tendency of which is to “replace” them with the personal intellectual virtues of the agent. It strikes me that to do so may smack of a methodological individualist approach that ignores social epistemology. I understand that you both hold that your approach supports and explains the rationality of theory-choice in science. But I worry that the "radical broadening" called for by the agent-centered Zagzebskian pure virtue theorist, where its the bon sens of the person who has it (or we accord as having it) determines what makes a theory choiceworthy, irrespective of the ampliative but communal criteria that we have in the "cogntive values." Yet it seems incredible that David would neglect this, given his deep interest in social epistemology, so I suspect I’m a bit confused on this and need some clarification. “Cognitive values” like simplicity, fertility, etc. seem to me to be 1) communal (Kuhn says scientific revolutions have changed them somewhat, but not greatly), and 2) theory-directed (fertility being about an objective feature of a theory or hypothesis, the ability to derive observational consequences from it, but simplicity being much more 'aesthetic'. for example). So the moral/intellectual judgment of the individual scientist seems filtered by such cognitive values. But Einstein says eloquently similar thing as Duhem does about the importance of the individual scientists' character. So I think that's right, but just want to say we need to distinguish the virtues of people and the virtues of theories. So I’m wondering if you could respond to this concern, too, and whether you agree that a sound theory should respect the two distinctions I pointed out. I suspect that you'd both agree considerably with Phil Olson's proposal "that an account of the intellectual virtues should be inquiry-focused," but it looks like there are plural accounts of how it might be. I'll look forward to exploring them further with you!
October 2, 2006 | Registered CommenterGuy Axtell
Thanks for the long replies to my paper. I will be brief. I think that Guy's two concerns are related. Regarding the first, Duhem surely claims both kinds of underdetermination (Kuhn would too). Regarding the second concern, it seems to me that virtues of theories are only going to be decisive if one rejects the methodological underdetermination thesis. For those who hold it, like Duhem, the virtues of a theory would never give complete justification for adopting it.

I share some of Guy's concerns about VE. I wrote the paper mostly as an insterpretation of Duhem. I do think that he has interesting points to make in the VE debates, as Abrol points out, but I am not sure what position I personally would want to defend in epistemology.

October 9, 2006 | Registered CommenterDavid Stump
David, Thanks. My first point that underdetermination of theories by methodological rules should be distinguished from underdetermination of theory by facts, comes from my own influence by Laudan. If you're saying you think the two are inseparable, that's OK, I just wanted to remind you of the literature on it, and that SOME PEOPLE think its a mistake to treat them as coming to the same thing.

Your second point is addressed to my distinction between the "personal" virtues of the scientist faced with a choice, and "cognitive values" (fertility, simplicity, fruitfulness, comprehensiveness, unifying power, etc) as "virtues of theories." The latter, you say, are "only going to be decisive if one rejects the methodological underdetermination thesis." This is especially interesting to me. I wonder if we're partly just talking at cross-purposes, since I don't deny that it is individuals who do the choosing (rather than amorphous 'communities'), nor am I agreeing with the positivists that choice is a kind of 'logical algorithym'. But the question it seems to me is whether the personal choice is mediated BY this other level that isn't addressed in the paper, the cognitive values. It seems like there are extremes to avoid--1) that the cognitive values are the whole story, and the personal, including moral values of the individual scientist are inessential to understanding theory-choice, and 2) the opposite of this, that the personal virtues are everything, and choice doesn't proceed through or get mediated by the 'virtues of theories'. I don't think that avoiding these two extreme requires us to reject the methodological underdetermination thesis. (In general, personal choice seems rarely about what theory to 'accept,' but what theory to 'pursue' for one's research projects, and so it is natural that these questions involve more personal virtues, including moral values, than the question (which perhaps rarely needs to be answered anyway (in turn because as Duhem says, we never abandon one theory until an alternative to it is extant anyway), but would involve more focus on the cogntive values) of which of two competing theory is simply better scientifically.

Underdetermination is a real and interesting problem I agree. But maybe there's another set of extremes to be avoided in approaching it as well: 3) underdetermination of theory-choice by methodological rules is ALWAYS a major issue (and challenge to the rationality of theory-choice in science), and 4) that it is NEVER a major concern, because the cognitive values are always available and sufficient to settle matters in an objective fashion even where theories are empirically equivalent.

So can we agree on needing to avoid these two sets of extremes, or do you see these as mischaracterizing the situation?
October 10, 2006 | Registered CommenterGuy Axtell
Thanks Guy. This seems right to me. I think that for Duhem, the personal virtues only come fully into play when the ordinary methodological rules don't determine the theory choice. In some cases, the evidence and the methodological virtues may be very stacked on one side of the question. Neverthelss, Duhem does argue for underdetermination, that is, in principle, methods and evidence are NEVER enough to decide between two competing theories. So I think he would hold that personal virtues have to come in to play, to allow the scientitist to make a judgement. One can interpret Duhem as a radical, or as more middle of the road. In the latter case, he would hold that common sense just dictates a clear answer in many cases of theory choice.
October 17, 2006 | Registered CommenterDavid Stump
Yes, this paints a picture of Duhem that seems helpful and enlightening. Its a picture where we don't need to be really radical about holism, or repeat some of the post-Kuhn era's excessive claims about underdetermination, in order to appreciate the force of your main points. It leaves open a kind of radicalness, but I think in a somewhat different direction than radicalism about holism.

I mean that it leaves open a radicalism about the "bon sens" more akin to Lynn Holt's "aretaic account of the role of apprehension in inquiry." I see some analogies between how you and Abrol want to use the bon sens, and what Lynn's central argument of his book in favor of "a conception of reason in which reason is primarily apprehensive, secondarily discursive and calculative," and that "contemporary virtue epistemology fails to conceive of virtue as an achievement available only to a few with the right combination of potential, education, experience and hard work" because it assimilates it "to a skill which anyone could learn or a reliable mechanism which anyone could possess" (p. 3). I can't explain all that is radical in Lynn's approach, but think you'd benefit from looking at his book. But part of what's radical in it is that he thinks out attitudes of respect towards science "are routinely coupled with a methodological conception of reason" that smacks of scientism; the alternative is to show that this routine coupling is mistaken, and that "the exercise of apprehensive reason [Aristotelian modes of reasoning] accounts for what are taken to be scientific successes [i.e., successes of its method or methods." Is this, as I suspect, close to your own view, as here the issue of reliabilism's faults, which Abrol wanted to push, seem closely engaged?
October 24, 2006 | Registered CommenterGuy Axtell
David Stump's published paper "Pierre Duhem's Virtue Epistemology"(2007) is available at

Milena Ivanova's "Pierre Duhem’s good sense as a guide to theory choice," partly a development of and response to Stump, is available at

Abstract: This paper examines Duhem’s concept of good sense as an attempt to support a non rule-governed
account of rationality in theory choice. Faced with the underdetermination of theory by evidence thesis
and the continuity thesis, Duhem tried to account for the ability of scientists to choose theories that continuously
grow to a natural classification. I will examine the concept of good sense and the problems that
stem from it. I will also present a recent attempt by David Stump to link good sense to virtue epistemology.
I will argue that even though this approach can be useful for the better comprehension of the concept
of good sense, there are some substantial differences between virtue epistemologists and Duhem. In
the light of this reconstruction of good sense, I will propose a possible way to interpret the concept of
good sense, which overcomes the noted problems and fits better with Duhem’s views on scientific
method and motivation in developing the concept of good sense.
July 22, 2010 | Registered CommenterGuy Axtell