Van Fraassen doesn't see the disagreement between realists and constructive empiricists as being about what we ought to believe. Instead, he sees the dispute as being about the proper aim of science. Here's one way he put the point:
Scientific realism and constructive empiricism are. as I understand them, not epistemologies but views of what science is. Both views characterize science as an activity with an aim - a point, a criterion of success - and construe (unqualified) acceptance of science as involving the belief that science meets that criterion. According to scientific realism the aim is truth (literally true theories about what things are like). Constructive empiricism sees the aim as not truth but empirical adequacy. [Analysis 58.3, 1998]
So scientific realism and constructive empiricism (as van Fraassen understands them) both need for there to be a purpose to science altogether - SCIENCE write large. As I intimate in my dcog paper, I don't think there is such a purpose. Even supposing that there is, however, it is not something that can be divined by a priori rumination. As van Fraassen admits, our account of what science is about must accommodate the actual history of science. It is a partly empirical enquiry responsible to evidence.
In this enquiry, the phenomena include historical documents and physical evidence. They probably also include the actual historical activities of scientists. Yet under no account is the aim or purpose of the activity itself among the phenomena. The aim of the activity is a posit, introduced as part of a philosophical-historical theory. Moreover, it is an unobservable posit.
Therefore, an agnostic (who declines to believe in the unobservable posits of even the most successful theories) must decline to believe in the aim of science. This follows regardless of what the aim of science is posited to be, so an agnostic must decline to be a constructive empiricist. This is a problem for van Fraassen, who thinks that agnosticism is a comfortable epistemic position for constructive empiricists. I see two possible replies.
First, he might stick to his agnostic guns. Refusing to believe in constructive empiricism, he still might accept it. That is, he could treat constructive empiricism as involving not a true theory about science but instead an empirically adequate one. This would involve some mental gymnastics, but being an agnostic already involves mental gymnastics. This meta move is only a small additional flourish.
Second, he might deny that the aim of science is a theoretical posit. Perhaps history is not a science. Perhaps discovering what what science is is not history. I don't see this line as terribly promising.
Van Fraassen has argued that we need a richer epistemology, one which allows for more than just binary beliefs or probabilistic degrees of belief. Moreover, he resists formal models of belief as direct representations of entities in the mind or brain. Yet he does seem to genuinely believe in states of opinion, "real epistemic attitudes, pointed to by traditional epistemology, which cannot be accommodated in the probabilist models we have developed so far" [ibid.].
As Sellars and Churchland convincingly argue, though, epistemic attitudes like this are not among the immediate phenomena of the world. We posit them as part of a (folk) psychological theory. An agnostic about scientific and folk scientific theories ought not to believe in beliefs.
Does van Fraassen acknowledge this anywhere? or is his psychological musing a personal matter rather than an announcement ex cathedra qua constructive empiricist?