Constructive Empiricism and the Role of Social Values in Science

Sherri Roush

(Draft: Please do not cite or quote without permission.)

To accept a theory is to make a commitment, a commitment to the further confrontation of new phenomena within the framework of that theory, . . . , and a wager that all relevant phenomena can be accounted for without giving up that theory. Commitments are not true or false; they are vindicated or not vindicated in the course of human history. Bas C. van Fraassen, The Scientific Image

One of the most common complaints one hears against the idea of granting a legitimate role for social values in theory choice in science is that it just doesn't make sense to regard social preferences as relevant to the truth, or to the way things are. "What is at issue," writes Susan Haack, is " . . . whether it is possible to derive an 'is' from an 'ought.'" One can see that this is not possible, she concludes, "as soon as one expresses it plainly: that propositions about what states of affairs are desirable or deplorable could be evidence that things are, or are not, so." (Haack, Fall 1993, 35, emphasis in original) Haack (Fall 1993, 1993) does not provide an argument for the view that it is impossible to derive an 'is' from an 'ought,' but the intuition she expresses is strong and widespread. The purpose of this paper is not to determine whether this view is correct, but rather to show that even if we grant it (which I do), we may still consistently believe that social values have a legitimate role in theory choice in science.
I will defend this conclusion by outlining a view about social values and theory choice that is based on the Constructive Empiricism (CE) of Bas van Fraassen. Some questions about what role social values may legitimately play in science look different, I contend, depending on whether they are viewed from a realist perspective according to which the aim of science is literally true description of reality, or from the point of view of a Constructive Empiricist anti-realism, according to which the aim of science is empirical adequacy, the fit of a theory to the observables. In Helen Longino's (1990) account of the role of social values in science she expressed what appears to be a realist view when articulating one of the goals she took science to have:
My concern in this study is with a scientific practice perceived as having true or representative accounts of its subject matter as a primary goal or good. When we are troubled about the role of contextual values or value-laden assumptions in science, it is because we are thinking of scientific inquiry as an activity whose intended outcome is an accurate understanding of whatever structures and processes are being investigated. (Longino 1990, 36)

Consequently, when Longino argued that social values had to play a role in deciding between theories because one had to choose with insufficient evidence which auxiliary assumptions to adopt in order to have any view about which evidence was relevant to a theory, she appeared to step just where Haack insists we must not. She appeared to commit herself to the view that social preferences could act as reasons to believe an assumption was true. As I develop CE in what I take to be the most natural way towards a view of the role of social values in theory choice, it will become clear that the view I defend is different from and probably not consistent with Longino's 1990 view, in that we locate differently the entry of social values into theory choice, and the relation of social values to epistemic reasons for theory choice. Nevertheless, the conclusion I draw, that social values indeed may have a legitimate role in theory choice, is obviously a defense of part of Longino's overall claim.

Limitations of the Objection
The force of Haack's objection can be maintained only if it is stated pretty much just as she has put it: preferences that a thing be so cannot be evidence that it is so, or a reason to think it is so. For there are claims that sound similar but are found on inspection to be indefensible. For example, we might have thought Haack said that the way we want things to be cannot be relevant to the way things are (as I stated the matter in my opening sentence). But this is manifestly false if the 'things' referred to are themselves social things; our preferences are the reasons many social things are as they are, though not alone reasons to think they are as they are. And the way we want things to be is clearly to some extent relevant to the way they will be in the future, even with non-social nature, for example the environment.
We might have thought Haack's objection was the same as saying that social values cannot be epistemically relevant. But if it were, then it would be wrong, since everyone knows that social values can introduce bias and that bias is relevant to whether we have good epistemic reasons. Finally, it must surely be acknowledged that the social structure of a community is relevant to how successful its members will be in finding correct answers to their questions. To take an extreme case, if the social structure of a community was such that everyone believed on authority the views of a certain individual rather than investigating any claims on her own, then it would be less likely that this community would find good answers to questions about the world than it is for a community of independent and interacting investigators to find good answers. Social structures can embody social values which are at the same time epistemic values, though the way that they function as epistemic values is in governing practices, not directly as reasons to believe a certain theory is true. Because they have an epistemic role to play, and epistemic questions are constitutive of science on anyone's view, that these sorts of social values have a function in science does not need to be defended by the sort of view I am developing. Some of the most discussed ways in which social values might be relevant to science do not fall prey to Haack's objection; I will focus on those that do.

Social Values in Theory Choice
According to the Constructive Empiricist, accepting a theory need not and should not involve believing that it is true, believing, that is, that its claims about all things are correct-unobservable as well as observable things. Accepting a theory should involve believing only that it is empirically adequate, that is, believing that it fits all observable phenomena, those we have actually observed, and those we have not. The source of the flexibility about theory choice that I find in CE lies in the following two aspects of the view. 1) For any given domain there is only one true theory, whereas there are in general many empirically adequate ones (whether we can imagine them or not). This follows roughly from the meanings of the terms. 2) Virtues of a theory that go beyond consistency, empirical adequacy, and empirical strength do not concern the relation between the theory and the world: "they provide reasons to prefer the theory independently of questions of truth." (van Fraassen 1980, 88)
Famously, the explanatoriness of a theory, which many had taken to provide an additional reason beyond empirical fit to think the theory was true, neither does that, on van Fraassen's view, nor does it provide a reason to think the theory is empirically adequate. (van Fraassen 1980, 89) What makes a theory explanatory, beyond its being empirically adequate, is something entirely pragmatic, related to the concerns of the user. On this view there isn't any kind of evidence we could get that would make it more rational to think a theory was true than to think it was empirically adequate. So, since when we might be wrong there is no reason to make ourselves more wrong by believing stronger claims, we are better off restricting our beliefs about theories to claims of empirical adequacy.
CE is not as it stands a methodological view that would tell us how to decide which theories are empirically adequate or how to choose between candidate theories. Nevertheless, it has some framing implications for these questions, because it involves a view about what we ought to be doing when we choose a theory: believing that it is empirically adequate on epistemic grounds, and prefering it to any rival that we also believe to be empirically adequate on pragmatic grounds that have nothing to do with correspondence between the theory and the world. It follows, crucially, that if we were faced with rival theories which according to our present evidence both appeared to be empirically adequate, choosing between them would not be a choice of which theory we should believe to be true. Since we would not be choosing which one to believe true, a fortiori we would not be choosing which one to believe true on the basis of social values. Thus, if social values were among the pragmatic grounds we appealed to in choosing among theories we believed to be empirically adequate, we would not be commiting the fallacy Haack inveighs against. And on this way of placing social values in the activity of theory choice the possibility that social values (one sort of pragmatic factor) would be a reason to think a theory empirically adequate does not arise. It is assumed that the criteria for judging empirical adequacy are thoroughly epistemic, as they should be-taking the lesson of Haack's objection-since empirical adequacy is just truth for observables.
It is illuminating to compare this view with Longino's (1990) picture of why social values are ineliminable from theory choice. There social values enter into our judgment not after all current evidence is tallied in the column of the appropriate hypotheses, but before this tallying and as a precondition of it. Following the bootstrap model of confirmation, Longino concludes that it is only by means of auxiliary hypotheses that the relevance of evidence to hypotheses can be judged. But to require those auxiliaries to be directly confirmed would be, according to Longino, to place unreasonable constraints on science that would disqualify much of the science we admire. We should require only that auxiliaries be indirectly confirmed, she thinks, an option which allows the influence of interests and values to enter into theory choice through the choice of auxiliary assumptions.
It seems to me that if these claims can be defended in case the aim of science is truth, then they can also be defended in case that aim is empirical adequacy. This is because one who claims empirical adequacy for a theory claims much more than that the theory fits the so far observed phenomena-it is claimed to fit all observable phenomena-and the person faces questions about evidence for this claim, and the incompleteness of her evidence, just as one who wants to claim truth for a theory does. More to Longino's point, if the true-theory seeker must appeal to auxiliary assumptions to show that evidence is relevant to a theory, then so must the seeker after theories that are empirically adequate, since the latter is just as surely obligated to link her theories to observables. However, since empirical adequacy is truth for observables, I cannot accept a claim that social values can be reasons for believing in the empirical adequacy of a theory-on pain of the fallacy I began with-and so I cannot accept that they can be reasons for believing auxiliary assumptions (are empirically adequate or wholly true) either. Longino's argument applies to empirical adequacy just as much as it does to truth, but its conclusion is unacceptable to (my version of) CE if that conclusion is stated as I have just done. In this sense the account I am sketching is at odds with Longino's (1990) view.
This tension between my view and what appears to be Longino's view arises from the impression her (1990) disscussion creates that social values are being substituted for evidence-being allowed to play the role epistemic reasons would play if we had more evidence-and the consequent impression that we are relaxing an epistemic standard when we allow social values to play a role in theory choice. However, something of a rapprochement is possible, partly because, though I think the reading I've given Longino's discussion is the natural one, she does not explicitly commit herself to the view that social values are to be reasons to believe auxiliary assumptions. Moreover, there is a Constructive Empiricist way of fleshing out some of what she does say that escapes that formulation.
Notice that an auxiliary assumption is a claim about the world, just as a theory is, and so we would aim for any such assumption we accept to be empirically adequate. If two such assumptions had equal amounts of (incomplete) evidence for the claim that they were empirically adequate, and neither was falsified, then a choice in favor of either could legitimately be based on social values. Crucially, that choice would not be an indication that one believed either assumption to be true, or even more empirically adequate than the other. It would be a choice to work with one rather than the other for pragmatic reasons. And it would be a choice that should be revised in light of new evidence that showed the other was in fact more likely to be empirically adequate. Because empirical adequacy requires only fit with observables, more than one theory or auxiliary can be empirically adequate for a domain. When our evidence is incomplete more than one theory or assumption may appear to be empirically adequate. In such a circumstance epistemic considerations would leave us indifferent, in a tie that pragmatic considerations like social values can legitimately break, in a decision that is nevertheless not about which theory or assumption to believe. Thus the tie-breaker use of social preferences works just the same for auxiliaries as it does for theories themselves.

Objections (4)
The conception I have described may sound fine in abstraction, but we may wonder whether it fits the way we actually see social values at work in science. Given how difficult it is to find even one empirically adequate theory for a domain, how often are we faced with two such theories between which pragmatic factors will have the opportunity to be tie-breakers? First, the need for a tie-breaker does not depend on two theories actually being empirically adequate, but on our having some evidence that they are and (in the best case) no evidence that they are not. Second, my view does not limit the situations in which we may appeal to pragmatic factors to cases of tie-breaking, though I will not be able to discuss this here.
For now, note that apparent ties do happen in just the domains that have been discussed as most susceptible to the influence of social values. One of the most discussed cases of the intrusion of social values into theory choice involves the 'man-the-hunter' and 'woman-the-gatherer' hypotheses of how the most distinctively human traits evolved in our species. To the question both perspectives regard as pivotal about how the use of tools developed in the species, the hypotheses each have a plausible answer. On one account, male hunting provided the conditions under which having tools and cooperation gave an evolutionary advantage, and the advantage afforded by spears provided the reason why the size of the canines could decrease at a certain point in time, and allow humans to take advantage of diets requiring more effective molars. On the gynecentric account the development of tool use was a response to the nutritional stress women faced as abundant forests were replaced by grasslands in which food was further afield, and as the conditions of reproduction changed to include longer human infancy and dependency. The nutritional stress of females was greater than that of males because females fed their young through pregnancy, lactation and beyond. Tool use on this account developed much earlier than the stone implements used in hunting, as women fashioned organic materials into objects for digging, carrying, and preparing foods. As for the changes in human teeth, female sexual choice of more sociable partners can explain the decline in the number of males with the most aggressive-looking canines. Thus the ingenuity and flexibility of intelligence began with the women of the species. (Longino 1990, 106-108)
Both of these hypotheses speculate beyond the data we have, but I see no obstacle to our understanding both of them as possible accounts of the data we have. In other words, I do not have to be a partisan of the man-the-hunter view in order to see that that hypothesis along with its auxiliaries will predict much of the evidence we have found. And I do not have to be a partisan of the woman-the-gatherer view to admit that taken with the auxiliaries that sustain its relation to evidence this hypothesis fits the so far observed observables at least as well as (and probably better than) its rival. In other words, what we have in this case, so far, is two hypotheses that the evidence so far suggests are both empirically adequate. Thus, on my account I may legitimately throw my commitment behind one or the other of these two accounts on the pragmatic grounds that one or the other is in line with my social values. I don't thereby believe that the hypothesis I prefer is true, and this last step does not provide the grounds I have for believing the account to be empirically adequate. But I may make a legitimate choice to devote energy to one or the other all the same.
I will consider the next two objections together since though they come from opposite sides my answers to the two are related. First, I can imagine someone worrying that while I have incorporated social values into theory choice, the fact that the choices I allow to be based on social values come after and not as integral to, epistemic judgments, and have nothing to do with truth whether about unobservables or about observables, trivializes the role of social values in theory choice. The second objection, which has been made to Longino's account by Haack, says that the situation I consider, where two rival hypotheses are running neck and neck as far as empirical adequacy judgments are concerned, is one where we have no right to make a choice at all because the evidence is insufficient to distinguish the rivals. (Haack Fall1993, 35)
The first thing to say about the first objection is that choices made on the basis of social values with the purpose of furthering social ends are trivial only if the social ends are trivial. A theory need not correspond with reality for its ideas and stories to have social effects, effects that we may or may not think it is good to promote. Moreover, people may regard the pictures and stories that a theory contains about unobservable items as true when some part of the scientific community accepts the theory, even if (or on my account though) that is not the enlightened attitude to take toward an accepted theory, and even if it is not the attitude scientists actually take. It is legitimate for us to weigh the consequences of such takings to be true when we make a pragmatic decision about whether to accept a theory that passes epistemic muster as far as we know.
The first thing to say about Haack's objection-that we have no right to make a choice at all in case of ties-is that while it would be well-suited to address someone whose claim entailed that social values could be a reason to think that one of these tied theories was true, it can gain no traction against the position I am describing. I grant that choices about whether one theory is closer to the truth or even more empirically adequate than another must be based on evidence, and that we have no right to make a choice about such an epistemic matter when the evidence is equal for two rivals (to speak in an idealized way). However, the choice that I propose we legitimately make after the evidence is in, between two epistemically tied rivals, is not a choice about which theory is closer to the truth, or more empirically adequate, or has more evidence in its favor. It is a choice about which theory serves better our practical goals, including social goals. And I see no reason to think that two theories whose evidence makes them equally compelling will necessarily be such that which theory we decide to work with makes no difference to social goals, or such that no one could find social reasons to prefer the one to the other. Whether a theory makes much difference to any social goals does seem to depend on its subject matter, but surely not in general on how it stands epistemically when compared with its rivals.
It may be that Haack's intention was to submit that when two theories are tied epistemically we have no legitimate grounds to do anything-even protecting social goals-with either one. However, this claim would be far harder to defend than the claim that social preferences cannot be grounds for epistemic choices, and is not a claim she even tries to defend. Surely if the positive evidence in favor of the empirical adequacy of the theories is sufficiently great then we have a right to use either theory on observables to serve our practical ends, where what is 'sufficiently great' will depend on the level of reliability the practical goals demand. But it seems to me further that, even leaving social goals aside we would have to reject the claim that we have no right to do anything with epistemically tied theories, because of its long-term consequences for achievement of the epistemic goals of science, a concern that Longino's call for a more realistic epistemology has taken very seriously. Two theories that are tied according to our present evidence may not be empirically equivalent, even if our evidence so far suggests that both are empirically adequate. This is because, obviously, the two theories may coincide in what they say about the things we have in fact observed, without coinciding in what they say about everything that is observable. It can also be that we simply do not know whether the two theories coincide in what they predict. A proof of the empirical equivalence of two theories is often possible when the theories are axiomatized and mathematical, but is not as readily available or tight with the sorts of hypotheses that tend to have the most relevance to social values. In such a case, only subsequent development of the theories would tell us whether they are empirically distinguishable.
It is not only whether two hypotheses are empirically distinguishable, but also exactly how each can be linked to observables that further research may be the only route to finding out: Kant thought the question whether the universe was finite or infinite in space and time was an unresolvable antinomy of reason, but now we understand how these questions are linked to claims that are empirically testable. Some seem ready to believe that nothing that could be observed could distinguish the man-the-hunter from the woman-the-gatherer view of human evolution (Haack 1993, 35), but this is something we do not know, and should be wary of assuming. I doubt that most of us thought we would be able to distinguish summer and winter seasonal temperature variations for periods 34 million years ago, that is, to find observational evidence that could be linked to a cooling trend in the winters that was not present in the summers. But recently a method has been found to make such distinctions via variations in the ear stones found in a certain species of Gulf Coast fish that survived the mass extinction of that era. (Ivany et al. 2000) Human behavior of the past is a particularly difficult matter to link tightly to observational evidence, but cannibalism on the part of some prehistoric native Americans has recently been claimed on the basis of chemical analysis of their preserved feces. (Wilford, 2000)
In any of the sorts of cases I have just described, research on both of the rival theories will be required if we are to determine what further empirical consequences each has, and how to test those predictions. If Haack's claim is that during this period of research everyone should remain agnostic about which, if any, of the two theories is true, then I agree (though that is because I think they should always remain agnostic in that sense). However, to demand that no scientist work with either theory would make advance on either of them impossible. And to demand that no scientist work on one theory to the exclusion of the other seems an unwarranted restriction on the division of labor. So, neither of these restrictions can be hard rules. Further, though some scientists will be able to work on the development of both theories, keeping an open heart, others will not find this psychologically sustainable. This tendency should not be viewed as a weakness either, since advocacy and competition can bring discoveries in ways that we should not want to impede. (And to repeat what I have already said, it cannot be wrong to advocate a theory on grounds of social values if there is no known theory epistemically superior, as long as we understand that the advocacy is a matter of social values in cases of epistemic tie.) I conclude that it cannot be impermissible to adopt an attitude of advocacy, or what van Fraassen calls 'commitment' or 'acceptance' for one theory over another on the basis of social values in cases of epistemic tie.
If we are ever to know whether two theories apparently equally supported for the status of empirically adequate are indeed both empirically adequate or rather empirically distinct, and even what their empirical import is, then the scientific community usually has to do more research, and it will be to the positive good of that research if some people make a commitment to one theory while others accept its rival, more to the good than if no one makes any such commitment. If the commitment is made on grounds of social preference, so be it. The fact that commitments to one or other of the epistemically tied theories can actually serve the epistemic interests of science in the long run also addresses the former objection, which worried that my view had trivialized the theory choice that is allowed to be made on the basis of social values. Here we see that not only can such a choice have social consequences, but also making such a commitment at all to the one or the other theory can serve the epistemic ends of science. If social values help one make such a commitment, so much the better.
The view I am sketching has the advantages that we can acknowledge a legitimate role for social values in theory choice while not admitting that social preferences can be reasons to believe a theory true (or empirically adequate), and that we can do this without rejecting the distinction between facts and values, or between the contexts of discovery and justification. One may wonder whether the price of these advantages must be as extreme as anti-realism. In particular, one may suspect that the epistemic tie situation in which I find a tie-breaker role for social values is a situation that the realist could find just as easily in any case where insufficient evidence is (roughly) equal for two rival theories. Let us even allow, in case it helps, that the evidence is evidence for the truth of the two theories, as distinct from their empirical adequacy.
There remains a salient distinction between this and the anti-realist tie-breaker situation. The anti-realist's tied theories can both be believed to have met all the anti-realist's epistemic goals, namely empirical adequacy (and consistency), since it is possible for more than one theory to do that. In contrast, we can know a priori that it is not possible for the two theories to have met all of the realist's epistemic goals, since no more than one theory can be true. For the realist an evidentiary tie cannot be an epistemic tie for any theory that goes beyond the observables, and there is no reason to believe that we should be epistemically indifferent between two such theories. Thus, if we appealed to pragmatic factors as a tie-breaker in such a case we would be illegitimately substituting pragmatic factors for missing evidence. Because the anti-realist's epistemic goals are more modest, it is possible for thorough epistemic indifference to arise between two rival theories, and so for appeal to other kinds of reasons for theory choice to be permissible.

Haack, Susan (Fall 1993), "Epistemological Reflections of an Old Feminist," Reason Papers 18, 31-44.
-------- (1993), "Knowledge and Propaganda: Reflections of an Old Feminist," Partisan Review LX, 556-563.
Ivany, Linda C. et al. (October 19, 2000), "Cooler winters as a possible cause of mass extinctions at the Eocene/Oligocene boundary," Nature 407: 887-890.
Longino, Helen E. (1990), Science as Social Knowledge: Values and Objectivity in Scientific Inquiry. Princeton: Princeton University Press.
-------- (1994), "The Fate of Knowledge in Social Theories of Science," in F. Schmitt (ed.), Socializing Epistemology: The Social Dimensions of Knowledge. Lanham: Rowman and Littlefield Publishers, Inc., 135-157.
van Fraassen, Bas C. (1980), The Scientific Image. Oxford: Clarendon Press.
Wilford, John Noble (2000), "New Data Suggests Some Cannibalism by Ancient Indians," New York Times, September 7, 2000.