In his article "Practical Inadequacy: Bas Van Fraassen's Failures of Systematicity", Curtis Forbes argues that constructive empiricism fails to provide a systematic account of science. An account of science is systematic when it is "able to account for the uses, methods, goals, and place of actual human science, as it exists within its institutional and industrial context." Forbes takes it, and I agree, that any acceptable philosophy of science must be systematic in this sense.
Forbes's case against the constructive empiricist begins with the famous conjunction argument. Scientists routinely conjoin theories; if scientists accept two theories T1 and T2, they will also be prepared to accepted the conjunction (T1 & T2). The realist can easily account for this since truth is preserved under conjunction: if both T1 and T2 are true, then the conjunction (T1 & T2) is also true. The constructive empiricist supposedly runs into trouble here because empirical adequacy is not preserved under conjunction. Indeed, one of the main reasons why conjoining theories is useful is that (T1 & T2) may entail some observable consequence E that is not entailed by either T1 or T2 separately. So the fact that T1 is empirically adequate and T2 is empirically adequate doesn't entail that (T1 & T2) is empirically adequate. So how does the constructive empiricist account for theory conjunction?
As Forbes points out, van Fraassen has a simple response to this. Scientists conjoin theories in order to test the empirical adequacy of the conjunction. They don't need to believe that the conjunction is empirically adequate. Ultimately scientists aim for empirically adequate theories, but when a theory conjunction entails a novel consequence, belief in the empirical adequacy of the conjunction can be withheld until after testing. Conjoining-to-test accounts for theory conjunction just as well as conjoining-to-believe.
Forbes argues that this response is inadequate, and leaves a failure of systematicity in constructive empiricism. The problem arises from the fact that science plays an important role in public policy, where we use science to make predictions about the future and guide our actions. Forbes cites the example of global warming research. Global warming theory is a conjunction of a variety of theories concerning the composition of the atmosphere, the properties of emissions released through fossil fuel burning, the impact of solar radiation on Earth's temperature, etc. These conjuncts have each individually been well-tested but, says Forbes, the conjunction in the form of global warming theory has not been. Indeed, global warming theory makes a variety of novel predictions concerning catastrophic climate change that no scientist intends to test. We don't want to test what will happen if we continue to burn fossil fuels at the current rates for another 50 years. Instead, we want to avoid such catastrophic outcomes.
So in this case, scientists must be conjoining-to-believe the empirical adequacy of global warming theory, not conjoining-to-test it. If scientists only ever conjoin-to-test, they would advise us to continue burning fossil fuels to check the novel predictions of global warming theory. This would be disastrous. In general then, constructive empiricism fails to account for the role science plays in informing public policy, where theory conjunctions are used to predict and then try to avoid certain outcomes.
There are three problems with Forbes's objection to constructive empiricism:
(1) The example of global warming theory depends on the claim that this theory as a whole has not been well-tested, only the conjuncts (e.g. theories about atmospheric composition) have been well-tested. This is simply false. Global warming theory has existed for decades, and many of its novel predictions have been tested and confirmed: glacial retreat and Arctic sea ice decline, accelerated sea level rise, increasing numbers of record high temperatures, etc. Of course, we haven't yet tested all the predictions of the theory. But that is never possible for any theory. Conjoining-to-test requires only that we specify some novel consequences of the theory conjunction and test them. Given the many confirmed novel predictions of global warming theory, we are by constructive empiricist lights entirely justified in treating the theory as well-tested, and applying it in policy decisions.
Suppose global warming theory did not exhibit such successes. Suppose it were a new theory, and scientists had not yet tested its novel predictions. In this case, it's not at all obvious that we should use it to inform important policies. After all, new theories are proposed all the time, many of which fail to be accepted by the scientific community. In the 1970s it was conjectured by a minority of climatologists that the emission of aerosols would increase Earth's albedo and induce global cooling and extensive glaciation. How should policy makers decide which theories to adopt to predict and respond to potential problems? Surely they should adopt just those theories that have passed a sufficient number of tests. That is the difference between global warming and global cooling.
(2) The second problem with Forbes's argument is that it is possible for a constructive empiricist to hold that scientists sometimes conjoin-to-believe, not just conjoin-to-test. As Andre Kukla points out in chapter 3 of "Studies in Scientific Realism", we can distinguish two forms of constructive empiricism. In Van Fraassen's original form, to accept some theory T involves believing only that T is empirically adequate. But we can modify this slightly to "conjunctive" constructive empiricism, where to accept T involves believing that T, in conjunction with any other accepted theories, is empirically adequate. When the conjunctive empiricist accepts a theory, she believes it to be empirically adequate and also believes that a new theory generated by conjoining T with any other accepted theory will also be empirically adequate. Obviously, conjoining-to-believe empirical adequacy is entirely compatible with conjunctive empiricism.
Note that there are independent reasons to favour this form of constructive empiricism. Scientists seem to aim for a unified view of the world; they want theories from different fields to provide a coherent world-picture. On van Fraassen's original formulation of constructive empiricism, it would be no problem for scientists to accept mutually inconsistent theories provided each theory is thought to be empirically adequate in its domain. By contrast the conjunctive empiricist will need to ensure that all accepted theories are, at minimum, consistent with each other. Aiming for conjunctive empirical adequacy seems to be more in line with scientific practice.
(3) If the conjunction argument is successful, it also refutes all plausible forms of realism. The sensible realist does not say that our best theories are true. She says that our best theories are approximately true. This qualification is necessary because nobody thinks that science has given us a complete, final, and unified theory of everything, and we all expect that even our best theories are likely to change somewhat in the future. So the realist asserts the approximate truth of our best theories, not truth period. But the conjunction of two approximately true theories need not be approximately true. In particular, the conjunction of two approximately true theories need not entail approximately true claims about observables. Even if we assume that the individual conjuncts of global warming theory are approximately true, this is no guarantee that global warming theory itself will provide reliable predictions about the future.
