Does believing a theory entail believing the entities it involves exist?
In physics lessons, secondary school children are taught about electrons. They are shown pictures of the structure of electrons, told about their involvement in charging atoms, and how when they flow they can be used to power electrical appliances. Nobody, however, has ever seen an electron, and in fact, due to their size being in the same order as that of protons, we will never see an electron, at least not the way we see things currently, even with the most advanced microscope or if our eyes evolved to millions of times the precision they currently have. So is it irresponsible to teach children in school about these unseen, unseeable, entities, as if they really exist?
Realism is the doctrine which states that statements make claims about how the world 'out there' actually is; it holds that there is a mind-independent reality, and that to say something true is to make a statement which corresponds to the way things actually are. Scientific realists, therefore, hold that talk about theoretical entities, if correct, discusses things which, though not readily apparent like tables or sheep (and the specifics of where this line is to be drawn is important), truly exist in the world. Thus discussion of a scientific theory commits a realist to a belief in the entities under consideration.
Fairly recently, however, a number of philosophers of science have debated whether realism should be taken for granted, with many of these denying that science is talking about mind independent matters-of-fact at all. The rise of Logical Positivism led to certain philosophers taking a purely empirical view and arguing that the realm of science cannot be a separate 'world out there', but that terms and theories are meaningful only in respect of how they might observably affect us. The question of whether theories commit Positivists to the existence of the entities involved is somewhat more complex - initially the answer might appear to be no, but in fact I suspect they are so committed: they deny meaning to terms for (pseudo-)things which cannot be observed, and about which no empirical data can be amassed, and demand nothing more of objects than that they make a verifiable impact upon us before granting them existence, so it seems likely that for a Positivist a meaningful theory will by definition involve existent entities.
Other philosophers, largely continuing this trend, argued in favour of Instrumentalism, possibly the most popular anti-realist position, which holds that science aims not to find out truth, but simply to enable accurate predictions. The theory holds that two supposedly competing theories which always predict the same empirical results are logically equivalent and cannot in fact be said to compete at all. Many philosophers have, however, denied that prediction is a good enough aim for science, and argue that there is still more to be said on the matter once we have established rules for predicting what might happen in the future, namely what the underlying facts governing the process are: what is actually going on in the real world. We wish to understand what is going on outside the cave, rather than simply to be able to predict the shadows projected onto the walls within. For an instrumentalist, therefore, there is no question of believing in theoretical terms' actual existence - they are simply convenient fictions used for the purpose of enabling prediction.
Presumably the most extreme anti-realists are idealists, who argue that the world is merely a creation of the mind, and does not exist in any way separate from it. Obviously this will preclude the existence of a mind-independent reality straight off, and idealists argues that the object of knowledge (and therefore science) must simply be the workings of the mind, and the concepts established therein. Such a view is fairly uncontentiously problematic: how could the world just be in our minds, and if it is, how can more than one person share experience, how can there ever be evidence for other minds, and where does this ineffable worldless consciousness spring from? If idealism were correct, it would seem to do away with science altogether; luckily the many problems make idealism a generally unloved, last-resort view.
Growing out of all these views are a number of further positions that are loosely grouped together (despite having interestingly difference nuances). Van Fraasen's Constructive Empiricism argues that empirical adequacy, and not truth is the aim of science - namely that within the framework in which we use terms, they should acceptably fit with our observations. Putnam's Internal Realism is similar, and views the framework as language, so within a linguistic domain a term has a position. Fine argues that taking the Natural Ontological Attitude we need not determine a general aim for science, and do not need to interpret theoretical discussions in a particular way, and can accept theories without making any decision as to whether they discuss truth or anything else. These views, and various others like them, tend to be agnostic about whether terms should be literally construed, although their proponents (somewhat strangely in my mind) frequently come down on the side of realistic interpretations, holding that whilst terms are used within contexts or within the general scheme of science, they do refer to actual things in the world.
The question, therefore, of whether to construe theories literally seems to boil down to whether we should be realists about science or not. Putnam's 'Miracle Argument', which, though now largely dismissed, was once considered a very good argument (though presumably never conclusive proof) for being a realist. The argument basically states that if scientific theories didn't describe how the world was, it would be miraculous that science is so successful, since this success would have to be put down simply to coincidence, rather than the theories being correct. Therefore, Putnam argues, the most rational explanation for science's success is that scientific claims are true. I think the Miracle Argument is so appealing because it maps well with the reason that realism seems intuitively correct - namely that the idea that language just describes how the world is is the simplest view, needing the fewest leaps of faith. The defeat of the Miracle Argument, therefore, may well mark not just one fewer arguments in favour of realism, but will probably mark a serious attack on the intuitive (and strongest) foundations of the doctrine. Unfortunately for the realist, there are a number of (serious) problems for the Miracle Argument.
Van Fraasen argues that we can justify a position based on the success of science because the development of our body of scientific theories is akin to a Darwinian evolutionary process, whereby we keep theories that have worked in the past and which have correctly predicted events, and throw out those which have failed to do so. Since we have preselected theories which are good at prediction, van Fraasen argues, it doesn't make sense to argue on the basis of their success at this. Lipton's response, however, is that van Fraasen is mistaking explaining the set for explaining the elements. He asks us to consider the example of a red-head convention, which would be an adequate explanation for why every person in the room has red hair, but wouldn't go any of the way to explaining why each person in the room has red hair, which is a question for a geneticist, not a conference organiser. Likewise, explaining that the set of theories are good at prediction due to an evolutionary process does not explain why each theory is good at prediction, and we certainly do ask why, for example, evolution favours having two eyes as better for survival, even if evolution explains the reason why we tend to have traits favourable to survival. I am not convinced Lipton's response to van Fraasen, however, is a problem for van Fraasen in terms of objecting to the Miracle Argument. By all means we can investigate what makes an individual theory successful, but I think van Fraasen is right that their preselected nature prohibits us from making conclusions based on the success of our current set of theories.
Another problem for claiming that abductive reasoning (inference to the best explanation) will provide us with the truth is that it is fairly unclear which explanation exactly we take to be best in many situations, and upon what basis we should decide this. This would seem to make science at best very imprecise. Do we, for example, choose the simplest explanation, the one which most coheres with our other theories, that which gives the most interesting results, or the one which, following Occam's razor, introduces the fewest new variables for the sole purpose of explanation? It might be argued that we need not (for the Miracle Argument to be fruitful) worry about how we choose explanations, since the Miracle Argument holds simply as long as they are successful, but this leads worryingly directly into the pessimistic induction: that since all the theories we have previously abandoned (the vast majority!) have been false, we have no reason (by induction) to believe our current ones will be more successful. The only way to escape the pessimistic induction, it seems to me, is to show why our current methods for choosing theories are substantially better than those we have used in the past.
Worse still, it seems to me that the Miracle Argument doesn't even support realism against various other theories. Consider Instrumentalism, which states that theoretical terms are just useful fictions for the purpose of accurate prediction. What could be less miraculous than that science is successful by definition, because that which we call science is just that which predicts well? If we are truly to take the most rational, least miraculous account of science's success to be correct, it seems likely that the one which can account for this a priori should be adopted.
There is, however, a major problem for all anti-realisms that seek to treat theoretical terms as something less real than everyday objects around us, and that is where this line is to be drawn. Maxwell argues that this line cannot be drawn anywhere, except arbitrarily, and I think I would agree with this claim. Philosophers of science frequently assert that the difference between observable and theoretical objects is that observables can be detected without any apparatus, whereas theoretical objects require addition equipment to detect, and cannot be seen in the everyday way. This formulation seems not only to ignore the possibility of our eyes evolving further (or ever having been less competent than they are now) - or simply to make the distinction entirely dependent on our current physical make-up - but it also seems to preclude the existence (in the usual sense) of a huge number of things we surely would be happy to accept. It does not seem clear, for example, why we allow observation using glasses, or through an atmosphere or window; nor is it obvious why eyes, retinas, optical nerves, et al do not count as apparatus: after all, it is far from clear that we directly observe anything (except perhaps pain or sense data). It seems to me that the 'observation' of radiation, for example, using a Geiger counter, is simply further along the continuum of observation which has tables at one end, and superstrings at the other. Unless some substantial division can be found, there seems no reason to separate terms in this way. Either we should accept electrons along with tables, or reject them together.
Overall, therefore, whilst there may be severe problems facing realism, and perhaps the argument for a retreat to an anti-realist position is persuasive, unless we are simply to resign ourselves to scepticism about all terms I am not convinced the case against a literal construal of 'theoretical' terms is proved. Whilst we might accept that science does not seek truth, it seems to be throwing out the baby with the bathwater to suggest that the things it discusses are not objective objects in the world - after all, the initial intuition that led us to begin from a realistic position seems to stand: if we do not take objects to exist, what exactly do we mean by 'existence', and what exactly would be necessary for this property to be instantiated?