Is Constructive Empiricism Less Risky than Scientific Realism

Is Constructive Empiricism Less Risky than Scientific Realism? Abstract: To put it simply, constructive empiricists like Bas van Fraassen believe whatever it is that scientific realists believe, but only the stuff about observables. Van Fraassen notes that the constructive empiricist’s set of beliefs is therefore “a great deal weaker” than the scientific realist’s, going on to claim that believing only this limited subset of what the realist believes “delivers us from metaphysics” (1980, p.69). Constructive empiricism is thus seen by its defenders as a less risky strategy for belief formation than scientific realism’s more risky “metaphysical” strategy. Furthermore, asserts van Fraassen, this is risk without any potential gain in relevant information. I argue that, while constructive empiricism is less risky in that it usually leads one to hold less false beliefs, there are important alternative senses of risk in which it comes out as a) no less risky, and b) more risky. Through a discussion of the so-called “conjunctive practice,” I argue that some information about unobservables can be very relevant and useful in lieu of an ability to conduct direct empirical testing. This gives us some epistemological and practical reasons to prefer a belief-forming strategy that aims for deciding whether theories are true (i.e. realism) over and above one that settles for deciding whether theories are empirically adequate (i.e. constructive empiricism).

1. 2. 3. 4.

Constructive Empiricism vs. Scientific Realism Van Fraassen on Truth and Risk Putting Information about Unobservable to Good Use The Value of Metaphysical Speculation

1. Introduction: Constructive Empiricism vs. Scientific Realism “Never let the fear of striking out get in your way.” Babe Ruth Bas van Fraassen has been a staunch empiricist for a long time. He is one of anti-realist empiricism's few modern defenders, and has defended it in at least two distinct forms (cf. Kukla, 1995). I here concentrate on criticizing his (1980) brand of empiricism, which he called “constructive empiricism.” My criticism is fairly simple: constructive empiricism, understood as a belief-forming strategy, fails to take the right sorts of epistemic risks because it refuses to seek out potentially relevant information. Its main philosophical opponent scientific realism does not share in this deficiency, and is therefore a preferable belief-forming strategy. I begin by explicating these two opposed positions, doing my best to stay completely in line with van Fraassen's own explication in (1980).

Constructive empiricism and scientific realism are stated as opposed theories of science. A theory of science is an attempt to account for and make sense of all the salient aspects of the activity we call “science,” including its method and empirical successes (cf. Psillos, 2000, p.69 and van Fraassen, 1994, p.191). The main questions to be answered by such a theory are a) “what is the ultimate aim of science?” and b) “what kind of attitude does a scientist takes towards a theory when she endorses it?” Constructive empiricism and scientific realism give divergent answers to these questions. With respect to the first question, realists assert that the ultimate aim of science is the production of true theories (e.g. Popper, 1983). Constructive empiricists instead assert that it is the production of empirically adequate theories, where empirically adequate just means true with respect to observables. With respect to the second question, realists claim that a scientist who endorses a theory believes that theory to be true, whereas constructive empiricists claim that a scientist who endorses a theory merely accepts that theory as empirically adequate. A theory of science must also account for the methods of science, including the regular appeal to the “virtues” of a theory when justifying an endorsement of that theory. The virtues that scientists will appeal to include things like fruitfulness, scope, depth, simplicity, consistency, etc. The realist will explain that most of these virtues are epistemic virtues, meaning that their possession by a theory is a good indication of its truth, which after all is the ultimate aim of science. The constructive empiricist instead divides these virtues into epistemic and pragmatic virtues, saying that empirical adequacy and empirical strength are the only truth-directed or epistemic virtues of scientific theories. All other virtues, such as simplicity and ease of use, are then rendered as merely pragmatic virtues. Unlike the epistemic virtues, the pragmatic virtues of a theory, says van Fraassen, only give us reasons for accepting a theory independently of questions of truth (van Fraassen, 1980, p.88). Both scientific realists and constructive empiricists believe that the attitude they attribute to scientists is the proper epistemic attitude to take. In this way, both theories of science may be rendered as belief-forming strategies: realists would recommend that we look for true theories, whereas constructive empiricists would recommend that we look for empirically adequate theories. The realist and constructive empiricist's accounts of the theoretical virtues can then be seen as a method for determining which theory is worth our endorsement, and the account of endorsement given then determines the proper attitude to take towards the theories that are worthy of being endorsed: belief or acceptance. 2. van Fraassen on Truth and Risk “There is a precipice on either side of you: a precipice of caution and a precipice of overdaring” Winston Churchill

In order to compare the relative risks of constructive empiricism and scientific realism, let us make the following tentative assumption, subject of course to revision after sufficient discussion: constructive empiricism and scientific realism are both adequate theories of science. This means that they are able to account for the salient features of a) the history of science, b) modern scientific practice, and c) the scientific method, considered abstractly. Within the scope of this assumption we can pose the questions at hand: in what sense is constructive empiricism less risky than scientific realism, and are there any senses in which it is more risky? 2.1. van Fraassen on Epistemic Goals One sense in which constructive empiricism is surely less risky than scientific realism is clear: constructive empiricism will all but assuredly result in less false beliefs than scientific realism. At worst, it will result in no more false beliefs than scientific realism, for the constructive empiricist's belief set is simply a subset of the realist's belief set. Thus, constructive empiricism, taken as a belief-forming strategy, risks less than scientific realism in that it minimizes our false beliefs, for nearly any limitation on a belief set is likely to eliminate some false beliefs. There is another important sense of risk, however, in which constructive empiricism is equally risky. As Cliff Hooker notes, the claims that the scientific community accepts about observables at any given time are just as likely to be revised as claims about unobservables, since the empirical tests that would force us to give up our belief that a theory is true are precisely the same as the tests that would force us to give up our beliefs that this theory is empirically adequate (Hooker, 1987, p.168). Thus, when it comes to the risk of being forced by the evidence to give up our beliefs, the risks of the constructive empiricist are “in fact exactly the same as are those of the realist.” (Hooker, 1987, p.169). Minimizing our false beliefs and remaining consistent with the evidence are not our only epistemic desires, however, and not the only senses of risk with which we might compare scientific realism and constructive empiricism. Drawing on William James's The Will to Believe, van Fraassen (2002, p.86-87) lists another epistemic goal that is at least as central to our epistemic aims as minimizing false beliefs: maximizing true beliefs. While it would be easy to completely achieve either one of these goals independently of the other – e.g. by believing either nothing or everything – together these goals inevitably pull us in different directions, for we can never have maximal information together with maximal security or credibility for this information. In short, the informativeness of our beliefs varies inversely with their security; there is an inevitable trade-off between information and credibility. Given that we cannot jointly maximize these two goals, van Fraassen claims that they must represent an incomplete account of our epistemic goals and desires. In addition to our desire for information and freedom from error, van Fraassen says that we desire “a properly well balanced body of opinion, with some equilibrium of information and security” (2002, p. 87). Unlike truth and falsity, which are objective features of beliefs,

the “right” balance of information and security is a matter of subjective judgment or opinion. How much falsity we are willing to risk in our quest for truth is a choice that we all must make for ourselves. Thus, “the responsibility for a[n epistemologically] crucial value judgment has landed on us ourselves” (ibid.). Van Fraassen goes on to identify a fourth epistemic goal, saying that “[a]s we are weighing the value of increasing our information content against the higher risk of falsehood in our beliefs, we'll certainly take into account what that information is about and what sort of information it is” (ibid.). In consensus over this last point, I think that everyone can agree we should never risk error for clearly useless information. So, this leaves us with four mainly, though not purely, epistemic goals, values, or desires: 1) 2) 3) 4)

a maximum of truthful information a minimum of error a good ratio of true to false beliefs never to risk error for irrelevant information

2.2. van Fraassen on the Risk of Scientific Realism Van Fraassen's (1980) critique of scientific realism can be clearly and succinctly stated in light of these four desires: realism risks more error only in order to gain irrelevant information. If, as is commonly done, we see risk simply as a function of potential benefits (relevant true beliefs) and potential losses (additional false beliefs), we can represent van Fraassen's argument in this way: 1. The potential gains in relevant information from accepting the empirical adequacy of a successful scientific theory is equal to the potential gains in relevant information from accepting its truth, for only information about observables can be relevant to our needs. 2. The potential for error in accepting the empirical adequacy of a successful scientific theory is less than the potential for error in accepting its truth. Therefore, 3. Scientific realism is a less preferable strategy for belief-formation than constructive empiricism, for it risks more error without the potential for any gain in relevant information. So, while we do need to take some risk when explaining and employing science, we can make do with the modicum of risk present in constructive empiricism. If an empirically adequate theory is just as good for all scientific purposes as a true theory, then we have no need to seek truth, believe in a theory's truth, or interpret scientists as believing a theory they endorse to be true. As James Ladyman succinctly puts it:

“[V]an Fraassen argues that if we need go no further than belief in the empirical adequacy of theories to account for the nature and practice of science, then, if we do go further, we take an unnecessary epistemic risk for no extra empirical gain … Van Fraassen rejects realism not because he thinks it irrational but because he rejects the 'inflationary metaphysics' that is an account of laws, causes, kinds and so on, which he thinks must accompany it … He thinks constructive empiricism offers an alternative view that offers a better account of scientific practice without such extravagance” (Ladyman, 2002, p.224-225, cf. van Fraassen, 1980, p.73). So, the question for the realist at this point is whether information about unobservables really is as irrelevant and extravagant as van Fraassen implies it is. If information about unobservables can prove useful in some important circumstances, then it not irrelevant, extravagant, or useless. In this case, the additional risk taken by realists when they evaluate scientific theories as true rather than only as empirically adequate might actually be worth it. If refusing to search for information about unobservables has the potential to result in negative consequences, there would be an important sense in which constructive empiricism might actually be more risky than scientific realism. 3. Putting Information about Unobservables to Good Use: The Conjunctive Practice “The policy of being too cautious is the greatest risk of all.” Jawaharial Nehru – First Prime Minister of India While it has been assumed for the sake of argument that constructive empiricism, like scientific realism, is an adequate theory of science, there is one arguably scientific practice that many realists have asserted cannot be accounted for by constructive empiricism: the so-called “conjunctive practice.” Both constructive empiricists and scientific realists admit that people will often endorse the conjunction of two already endorsed theories, even if this conjunction entails additional empirical consequences. Van Fraassen nicely articulates the difficulty for constructive empiricism in accounting for this practice when he writes: “[I]f one believes both T and T' to be true, then of course (on pain of inconsistency) one believes their conjunction to be true. But if T and T' are theories which are both empirically adequate, their conjunction need not be” (1980, p.83). 3.1 Conjoining-to-Test vs. Conjoining-to-Believe Given that we have supposed constructive empiricism to be capable of accounting for all the distinctly scientific phenomena, we must deem the conjunctive practice to be a nonscientific phenomenon, for unlike truth empirical adequacy is not closed under

conjunction. This is precisely what van Fraassen does when he ends his (1980) discussion of the conjunctive practice by saying: “There can be no phenomena of the scientific life of which this account draws a faithful picture. The reason is that, as long as we are scientific in spirit, we cannot become dogmatic about even those theories which we whole-heartedly believe to be true. Hence a scientist must always, even if tacitly, reason at least as follows in such a case: if I believe T and T' to be true, then I also believe that (T and T') is true, and that it is empirically adequate. But in this new area of application T and T' are genuinely being used in conjunction; therefore, I will have a chance to see whether (T and T') really is empirically adequate, as I believe … Thus my beliefs are about to be put to a more stringent test in this joint application than they have ever been before. What I have just described is the nearest practice can come to the simple account of putting theories together as mere conjunction. In my opinion, practice does not come that near; the preamble about believing T and T' to be true is missing, and what is about to be put to a more stringent test is the hypothesis that (T and T') is empirically adequate” (1980, p.85). Thus, on van Fraassen's account, the distinctly scientific practice is conjoining theories to test the empirical adequacy of their conjunction, not conjoining theories to believe in the empirical adequacy of their conjunction. Scientists, qua scientists, do not conjoin-tobelieve; they only conjoin-to-test. By deeming “conjoining-to-believe” a non-scientific practice, van Fraassen is able to maintain his claim that constructive empiricism is able to account for all the phenomena of science. While the adequacy of constructive empiricism as a theory of science qua science has been assumed for the sake of argument, I shall now argue that constructive empiricism is not an adequate theory of science qua social-political phenomenon or science qua producer of tools for long-term planning. My argument for this claim is fairly simple – conjoining-to-believe is a regularly observed and eminently rational practice in our society. Policy makers, for example, will regularly conjoin scientific theories in order to believe the novel predictions of such conjunctions. Whereas scientists qua scientists only need to conjoin theories to test their novel predictions, public policy makers qua policy makers often need to conjoin theories to believe their novel predictions. In fact, I argue, not conjoining-to-believe would be an unnecessary risk for the policy maker, meaning that being a constructive empiricist would be riskier than being a scientific realist. This is because constructive empiricism does not have the resources to discriminate between belief-worthy and non-belief-worthy conjunctions, leaving it with an important deficiency as a theory of modern science, which exists inextricably within a democratic, policy-driven, and highly technocratic society that demands such realist discriminations from its scientists. This explains why, in many social-political contexts, scientific theories are often debated and settled along unmistakably realist lines, and the

untested predictions entailed by the conjunction of well-tested theories are often relied upon simply because the conjuncts are already endorsed. 3.2. The Conjunctive Practice as a part of Science vs. The Conjunctive Practice as a part of Policy Making The scientific method would ideally require scientists to await the actual execution of any experiment or test rather than commit to a certain result before the test was conducted. Scientists, qua scientists, have this luxury because a) they have no preferences with respect to any test conditions, and b) they are never in a position to rely on the theories they endorse. Unlike scientists, however, the citizens of modern democratic nations are not so indifferent to many test conditions, and have the need and ability to rely on the scientific theories they endorse. Because of this, public-policy makers will often need to conjoin-to-believe, so that they can predict and rely on such conjunctions before they are ever able to actually go out and test the empirical adequacy of these conjunctions. A good real-life example of the need to conjoin-to-believe is found in contemporary debates over currently accepted climatological models. These models predict that drastic consequences will result from our current environmental and economic policies. Somewhat ironically, the definitive test conditions of these models are precisely the drastic circumstances we wish to avoid by enacting preventative policies. Because of this, we find ourselves in an unfortunate circumstance: an empirical test of these models could only be conducted once it was too late for us to act on them. As a result, the empirical adequacy of these climatological models could never be established through empirical tests in a useful way. These climatological models are very complex amalgamations of other theories and models that have been individually well-tested, e.g. theories about the properties of carbon dioxide, the paths of ocean currents, etc. But the novel predictions that we get by conjoining all these theories together have not been so well-tested; nor could they ever be well-tested, for practical reasons, at least not before the catastrophic global climate change they predict occurs or fails to occur. Thus, these climatological models could only ever be tested partially and piecemeal by testing the theories and models that conjoined to construct them. Because of this, the only way to reach a belief in the novel predictions that result from conjoining such well-tested theories and models with enough time to act is through the process of conjoining-to-believe. The situation that the policy-maker finds themselves in is very different from that of the scientist. The policy maker wants to craft policy as a preventative action, and would ideally not simply just sit back and wait to see if global climate will occur as predicted. The policy maker needs to know if these models are empirically adequate before the definite test can be conducted: only then can they formulate policy in order to avoid it. As a result, policy makers cannot make do with simply conjoining-to-test. Sometimes they will need to conjoin-to-believe: but they will have no basis for a belief about the empirical adequacy of a conjunction of empirically adequate theories unless they believe

that the conjuncts are true, not merely empirically adequate. This means that policy makers will often need to use realist methods of theory evaluation and extension. While metaphysical speculation of the realist variety regarding the truth of scientific theories may be an inessential part of the scientific method, it is an essential part of the modern political method, for we now have unprecedented ability to actually rely on the predictions of scientific theories in the kinds of policy we craft. Unlike a collective of 17th century peasants, we are in a position to legislate on the basis of untested but credible theory conjunctions, which gives us a new political responsibility to make well-informed policy. Since constructive empiricism is incapable of making responsible discriminations of the relevant sort, the public policy maker must often risk engaging in metaphysical speculation in order to make responsible discriminations between legitimate and illegitimate conjunctions of currently accepted scientific theories. We should, of course, ask ourselves if the distinction between the scientific life and the social-political life that van Fraassen has forced upon us can really remain tenable in the modern world. If we give up our initial assumption that constructive empiricism is an adequate theory of science, we can see conjoining-to-believe as a salient feature of modern science as it exists within modern society, i.e. as a scientific phenomenon that any adequate theory of science must account for. In this case, constructive empiricism fails to be an adequate theory of “science qua science”. If, on the other hand, we insist that constructive empiricism is an adequate theory of science, it will fail to be an adequate theory of “science as it is within a scientifically minded, policy-driven society.” Our choice of view regarding the proper place of the conjunctive practice is ultimately up to us philosophers of science as a group to decide. Should realists allow van Fraassen to tear science away from its inevitable social-political context in the way he has, thereby allowing constructive empiricism to retain its status as an adequate theory of science? I do not think so. But even if we allow him to do this, realists can remain confident that there is a very important sense in which constructive empiricism is more risky than scientific realism, and that (contra van Fraassen) information about unobservables can often be very useful. 4. Conclusion: The Viability of Empiricism and the Value of Metaphysics What I have argued here is that super-empirical methods of theory evaluation have the potential to provide us with information about theories that may prove relevant and useful when we are attempting to plan for the future, and are not granted the luxury of conducting the relevant empirical tests. The information we can attain through this additional risk is whether the empirical adequacy of a theory can be treated as closed under conjunction with other empirically adequate theories. In the development of robust long-term plans, information about unobservables can be very useful. Knowing deep truths about the nature of unobservable reality can even be more important than knowing which theory is the most empirically adequate. When planning for the future in a robust manner, as we do when we craft public policy, we might rightly trade-off some precise

information about observables for some deeper information about unobservables. This is because we expect policy crafted on the basis of truth to remain more effective in the long run than policy crafted on the basis of mere empirical adequacy. This why, as Cliff Hooker puts it, only the scientific realist can rationally account for why people do not always rely on the most empirically adequate theory available: despite its being more empirically adequate than some alternative, it seemed to be further from the truth. Thus, the choice to be metaphysical and accept scientific realism cannot be condemned as an unnecessary risk. While the amount of risk one is willing to take is, ultimately, a subjective choice that we all have to make, I for one am more than willing to trade-off a little more security for my beliefs in an effort to gain such extremely relevant and useful information about unobservables. While the scientific activity might be made sense of as the search for empirically adequate theories, science's place in society and public policy making cannot be made sense of in this way, for the activity of conjoining-to-believe makes no sense on this view. At the very least, empiricists should admit that, while risky, believing in the truth of theories is often warranted, necessary, and potentially rewarding, even if this is only in some “extra-scientific” contexts. In these extra-scientific contexts we will need nuanced, super-empirical, and distinctly realist methods of theory evaluation; and if the admission that we sometimes need realist methods of theory evaluation, for epistemic reasons, does not serve as a vindication of scientific realism and the metaphysical stance, I do not know what would. Works Cited