Some versions of logical positivism were also committed to the "phe nomenalist" idea that all meaningful sentences can be translated into sentences that refer only to sensations. If phenomenalism is true, then when we seem to make claims about real external objects, all we are talking about are patterns in our sensations. Some more holistic empiricist views about language, of the kind associated with logical empiricism, have the same consequence. Even if translations are not possible, the nature of language prevents us from hoping to describe the structure of a world beyond our senses. Language and thought just cannot "reach" that far. I believe that a lot of twentiethcentury empiricism held onto a version of this view (though some commentators on this book have objected to this claim).
In recent years the tension between realism and empiricism has often been debated under the topic of the "underdetermination of theory by evidence." Empiricists argue that there will always be a range of alternative theories compatible with all our actual evidence, and maybe a range of alternative theories compatible with all our possible evidence. So we never have good empirical grounds for choosing one of these theories over others and regarding it as representing how the world really is. This takes us back to the discussion in the previous section about the right level of optimism we should have about our scientific theories. I expressed scientific realism in a way compatible with a fair degree of pessimism, but the problem of underdetermination is important in its own right (see also sections 15.2 and 15.3).
12.5 Metaphysical Constructivism
I use the term "metaphysical constructivism" for a family of views including those of Kuhn and Latour. These views hold that, in some sense, we have to regard the world as created or constructed by scientific theorizing. Kuhn expressed this claim by saying that when paradigms change, the world changes too. Latour expresses the view by saying that nature (the real world) is the product of the decisions made by scientists in the settlement of controversies. Nelson Goodman is another example; he argues that when we invent new languages and theories, we create new "worlds" as well (1978). For a metaphysical constructivist, it is not even possible for a scientific theory to describe the world as it exists independent of thought, because reality itself is dependent on what people say and think.
These views are always hard to interpret, because they look so strange when interpreted literally. How could we possibly make the world just by making up a new theory? Maybe Kuhn, Latour, and Goodman are just using a metaphor of some kind? Perhaps. Kuhn sometimes expressed a different view on the question, a kind of skeptical realism, and he struggled to make his position clear. But when writers such as Goodman have been asked about this, they have generally insisted that their claims are not just metaphorical (Goodman 1996, 145). They think there is something quite wrong with scientific realism of the kind I described in section 11.3. They accept that it's hard to describe a good alternative, but they think we should use the concept of "construction," or something like it, to express the relationship between theories and reality.
Some of these ideas can be seen as modified versions of the view of Immanuel Kant ([1781] 1998). Kant distinguished the "noumenal" world from the "phenomenal" world. The noumenal world is the world as it is in itself. This is a world we are bound to believe in, but which we can never know anything about. The phenomenal world is the world as it appears to us. The phenomenal world is knowable, but it is partly our creation. It does not exist independently of the structure of our minds.
This kind of picture has often seemed appealing to philosophers who want to deny scientific realism but do so in a moderate way. HoyningenHuene (1993) has argued that we should interpret Kuhn's views as similar to Kant's. In Michael Devitt's analysis of the realism debates (1997), a wide range of philosophers are seen as either deliberately or inadvertently following the Kantian pattern. According to Devitt, constructivist antirealism works by combining the Kantian picture with a kind of relativism, with the idea that different people or communities create different "phenomenal worlds" via the imposition of their different concepts on experience. This relativist idea was not part of Kant's original view; for Kant all humans apply the same basic conceptual framework and have no choice in the matter.
The Kantian picture is sometimes seen as a way of holding onto the idea that there is a real world constraining what we believe but doing so in a way that does not permit our knowing or representing this world. This move is often tempting, but the resulting views are unhelpful. Understanding our access to reality is difficult, but adding an extra layer called "the phenomenal world" in between us and the real world achieves nothing.
The term "social constructivism" is often used for roughly the same kind of view that I am calling metaphysical constructivism. But "social constructivism" is also used for more moderate ideas as well. If someone argues that we make or construct our theories, or our classifications of objects, that claim is not opposed to scientific realism. We do indeed "construct" our ideas and classifications. Nature does not hand them to us on a platter. But a scientific realist insists that beyond ideas and theories there is also the rest of reality.
In fields like sociology of science, as we saw in chapter 8, there is an unfortunate tradition of not explicitly distinguishing between the construc tion of ideas and the construction of reality. What is it about these fields that has encouraged such strange-sounding formulations of ideas? There are various reasons, but I will venture some meta-sociology here-sociology of the sociology of science. A lot of work in these fields has been organized around the desire to oppose a particular Bad View that is seen as completely wrong. The Bad View holds that reality determines thought by stamping itself on the passive mind; reality acts on scientific belief with "unmediated compulsory force" (Shapin 198z, 163). That picture is to be avoided at all costs; it is often seen as not only false but even politically harmful, because it suggests a passive, inactive view of human thought. Many traditional philosophical theories are interpreted as implicitly committed to this Bad View. This is one source for descriptions of logical positivism as reactionary, helpful to oppressors, and so on.
What results from this is a tendency for people to go as far as possible away from the Bad View. This encourages people to assert simple reversals of the Bad View's relationship between mind and world. Thus we reach the idea that theories construct reality.
Some explicitly embrace the idea of an "inversion" of the traditional picture (Woolgar 1988, 65), while others leave things more ambiguous. But there is little pressure within the field to discourage people from going too far in these statements. (Bloor 1999 is an interesting exception.) Indeed, those who express more moderate denials of the Bad View leave themselves vulnerable to criticism from within the field. The result is a literature in which one error-the view that reality stamps itself on the passive mind-is exchanged for another error, the view that thought or theory constructs reality.
12.6 Van Fraassen's View
The last form of opposition to scientific realism that I will discuss is a more moderate and careful form; this is the position of Bas van Fraassen (1980). Van Fraassen's ideas lie within the empiricist tradition, but they are not based on a linguistic or psychological theory. Instead, van Fraassen confronts realism on the proper aims of science. So his antirealism is a direct denial of the kind of scientific realism defended in this chapter. This is no accident, since my formulation of scientific realism was influenced by his.
In discussions of realism, the term "instrumentalism" is used to refer to a variety of antirealist views. Sometimes it is used for traditional empiricist positions of the kind discussed earlier. But sometimes it is used in a different way, which I think is more appropriate. According to instrumentalism in this sense, we should think of scientific theories as devices for helping us deal with experience. Rather than saying that describing the real world is impossible, an instrumentalist will urge us not to worry about whether a theory is a true description of the world, or whether electrons "really, really exist." If a
theory enables us to make good predictions, what more can we ask? If we have a theory that gives us the right answers with respect to what we can observe, we might occasionally find ourselves wondering if these right answers result from some deeper "match" between the theory and the world. But we can never expect to know the answer to this question, so what relevance does it have to science? Quite a few scientists have expressed instrumentalist views, especially in physics. The idea that we should ignore questions about the "real reality" of theoretical entities because these questions have no practical relevance is also linked to one strand of the pragmatist tradition in philosophy (Rorty 198z).
A detailed version of this kind of position has been worked out by van Fraassen (1980). Van Fraassen does not use the term "instrumentalist" to describe his view; he calls it "constructive empiricism" The term "constructive" is used by so many people that it often seems to have no meaning at all, so I have reserved it for the views discussed in section z z.5. I see van Fraassen's view as a version of the instrumentalist approach, but it does not matter much what we call it.
Van Fraassen suggests that all we should ask of theories is that they accurately describe the observable parts of the world. Theories that do this are "empirically adequate." An empirically adequate theory might also describe the hidden structure of reality, but whether or not it does so is of no interest to science. For van Fraassen, when a theory passes a lot of tests and becomes well established, the right attitude to have toward the theory is to "accept" it, in a special sense. To accept a theory is to (z) believe (provisionally) that the theory is empirically adequate, and to (z) use the concepts the theory provides when thinking about further problems and when trying to extend and refine the theory.
Regarding point i, for a theory to be empirically adequate, it must describe all the observable phenomena that come within its domain, including those we have not yet investigated. Some of the familiar problems of induction and confirmation appear here. Regarding point z, van Fraassen wants to recognize that scientists do come to "live inside" their theories; they make use of the theory's picture of the world when exploring new phenomena. Some versions of instrumentalism struggle to make sense of this fact. But van Fraassen says a scientist can "live inside" a theory while remaining agnostic about whether the theory is true.
How can we decide between van Fraassen's view and the version of scientific realism that I outlined earlier?
First we need to be sure that the two positions conflict. I said that one aim of science is to give us accurate representations of the world, including the unobservable parts. Van Fraassen says "science aims to give us theories which are empirically adequate" (1980, iz). So far, our views seem compatible. In some cases science could aim only at empirical adequacy, but in other cases it could aim at representing the hidden structure of the world as well.
And this is the right attitude for a realist to have. For various reasons and in various situations, it might make sense for a scientist to be cautious, or unconcerned, about the application of a theory to the unobserved structure of the world, even when he or she is becoming confident about empirical adequacy.
So van Fraassen has described an attitude that scientists can reasonably have toward some theories in some circumstances. But van Fraassen thinks that science should aim at no more than empirical adequacy.
As many have argued, one place where van Fraassen's view runs into trouble is the distinction between observable and unobservable parts of the world. Realists have argued that there is a continuum, rather than a sharp boundary, between the observable and the unobservable (Maxwell 196z). Some things can be observed with the naked eye, like trees. Other things, like the smallest subatomic particles, are unobservable and can only have their presence inferred from their effects on the behavior of observable things. But between the clear cases we have lots of unclear ones. Is it observation if you use a telescope? How about a light microscope? An X-ray machine? An MRI scan? An electron microscope? The realist thinks that the distinction between the observable and unobservable is vague, and not of the right kind to support general conclusions about what science aims to represent.
Van Fraassen accepts that the distinction between the observable and the unobservable is vague, and he accepts that there is nothing "unreal" about the unobservable. He also accepts that we learn about the boundary from science itself. Still, he argues, science is only concerned with empirical adequacy-making true claims about the observable part of the world. But this view cannot be defended. Van Fraassen is saying it's never reasonable for science to aim at describing the structure of the world beyond this particular boundary. Suppose we describe a slightly different boundary, based on a concept a bit broader than "observation." Let's say that something is detectable if either it is observable or its presence can be very reliably inferred from what is observable. As with van Fraassen's concept of observability, science itself tells us which things are detectable. In this sense, the chemical structures of various important molecules like sugars and DNA are detectable although not observable. So why shouldn't science aim at giving us accurate representations of the detectable features of the world as well as the observable features? Why shouldn't science aim to tell us what the molecular structure of complex sugars is like?
Perhaps our beliefs about the detectable structures are not as reliable as our beliefs about the observable structures. If so, we need to be more cautious when we take theories to be telling us what the detectable structure of the world is like. But that is no problem; we often need to be cautious.
What is so special about the "detectable"? Nothing, of course. We could define an even broader category of objects and structures, which includes the detectable things plus those that can have their presence inferred from observations with moderate reliability. Why should science stop before trying to work out what lies beyond this boundary? We might need to be even more careful with our beliefs about those features of the world, but that is no problem.
You can see how the argument is going. From the realist point of view, there is no boundary that marks the distinction between features of the world that science can reasonably aim to tell us about and features that science cannot reasonably aim to tell us about. As we learn about the world, we also learn more and more about which parts of the world we can expect to have reliable information about. And there is no reason why science should not try to describe all the aspects of the world that we can hope to gain reliable information about. As we move from one area to another, we must often adjust our level of confidence. Sometimes, especially in areas such as theoretical physics, which are fraught with strange puzzles, we might have reason to adopt something like van Fraassen's attitude, at least temporarily. But it is a mistake to think that empirical adequacy of van Fraassen's kind is the aim of science.
12.7 Representation, Models, and Truth (Optional Section)
I will finish the chapter with further discussion of an issue introduced in section 1z.3. I formulated scientific realism by saying that science tries to give us "accurate representations" of the world. Most discussion of this topic in twentiethcentury philosophy treated theories as linguistic entities, as collections of sentences. So when people tried to work out what sorts of relationships theories have with reality, they drew on concepts from the philosophy of language. In particular, the concepts of truth and reference were emphasized. A good scientific theory is a true theory; how can we determine which theories are true? Electrons exist if the word "electron" refers to them; how do we decide whether a term in a scientific theory refers to anything? A range of problems came to be addressed via the concepts of truth and reference.
This emphasis might be a bad idea. There are several issues to consider. One has to do with the "representational vehicles," or representational media, used by science. Science does express hypotheses about the world using sentences in language, either ordinary language or technical extensions of ordinary language. But in other cases, science uses representational vehicles of a
different kind. Many hypotheses in science are expressed using models. Consider the case of mathematical models. These are abstract mathematical structures that are supposed to represent key features of real systems in the world. But in thinking about how a mathematical model might succeed in representing the world, the linguistic concepts of truth, falsity, reference, and so forth do not seem to be useful. Models have a different kind of representational relationship with the world from that found in language. A good model is one that has some kind of similarity relationship, probably of an abstract kind, with the system that the model is "targeted" at (Giere 1988). It is hard to work out the details of this idea.
The role of models in science did become an important topic in latetwentieth-century philosophy (Suppe 1977). Some argued that we should use the idea of a model to give a different description of how all theories work in science. But it is a mistake to think that all of science uses the same "vehicles" to represent the world. We should not replace a language-based analysis of all science with a model-based analysis. What we find in science is a range of different representational vehicles.
Consider Darwin's Origin of Species. Darwin's book contained a set of hypotheses about the world, supported with elaborate arguments, expressed using rather ordinary language. But not all science is like this. Even the topics that Darwin was addressing are now treated differently. Recent discussions of how natural selection changes biological populations tend to be expressed in the form of mathematical models. These models are written down, of course. They are formulated using mathematical symbolism, and they have to be supplemented with a commentary telling us (for example) which phenomena in the real world are being represented by the model. But we should not expect an analysis of how mathematical models relate to the world to use the same concepts as an analysis of how hypotheses expressed in ordinary language relate to the world.
Theory and Reality Page 26