So let's do some straight-up Analytic philosophy. Only of science, because I still have no patience for metaphysics.
In his (apparently) famous piece `The republic of science' (1962), Michael Polanyi gives an argument for the autonomy of science. In particular, he's arguing that scientists ought to be allowed and encouraged, by society at large, to pursue whatever research projects they like, rather than the research projects that are felt to be likely to be useful to society at large. This is vague, and that reflects the fact that the conclusion of the argument is vague.
The particular argument I'm interested comes from page 61, for those following along at home. The terminology I use below is my gloss on Polanyi. (In other words, this is definitely a paraphrase.)
(1) Academic credit is an accurate indication of scientifically promising research programmes.
(2) Society ought to distribute grant money (and other material resources necessary for research) according to the scientific promise of research programmes.
(3) Therefore, society ought to distribute grant money according to academic credit.
This conclusion is to be contrasted with something like
(4) Society ought to distribute grant money according to projected social utility.
Polanyi's conclusion holds science to be autonomous (at least normatively) in the sense that the property determining the distribution of grant money is internal to the institutions of science (albeit not the sort of evidence-and-theory model of science of the positivists). It is not, in particular, the property of being expected to produce something of great value to society as a whole (eg, new technologies that will allow us to grow ten times as much corn).
One quick paragraph on `academic credit'. This term just refers to the level of esteem a research programme (or researcher) enjoys among the community of scientists. If everyone thinks nonlinear dynamics is going to be the Next Big Thing, then nonlinear dynamics enjoys a large amount of academic credit. On the other hand, if everyone thinks string theory is a complete dead end, then string theory enjoys very little academic credit. Polanyi is arguing that, in this situation, nonlinear dynamics should get lots of grant money (and grad students and lab space), while string theory should be basically cut off.
Another quick paragraph on `scientific promise'. This is supposed to be the promise (or, better, the potential or likelihood) of producting some new and valuable piece of pure knowledge. That `pure' is extremely important. Polanyi is operating with an implicit but very sharp division between science and technology (episteme and techne). Technology is a valuable product of scientific research, but is only an accidental product. The true product of scientific research is pure knowledge, and it is in this way that scientific research is valued `in itself'.
Now, I have a number of problems with this argument. First, and most trivially, we're not really given any reason to think (1) is true. But it's plausible, and there doesn't really seem any better way of measuring scientific promise around.
The blue whale that is the second problem is that (2) is begging the question, or coming so close to begging the question as makes no difference. With Polanyi's implicit distinction between science and technology, we can rewrite (4) as
(4') Society ought to distribute grant money according to technological promise.
If (4') is the anti-conclusion Polanyi is trying to reject, then (2) can't serve as a premiss. It's the conclusion he wants.
That is, if Polanyi's opponent is valuing science for its technological achievements, and not for its ability to produce pure knowledge, then of course the opponent is going to reject the premiss that grant money ought to be distributed according to the potential to produce pure knowledge. That, fundamentally, is just exactly where the disagreement lies.
Perhaps Polanyi intends (2) to be the real conclusion of the argument, and I'm just spectacularly misunderstanding him. But I don't see anything else that might serve as a premiss to get from (1) to (1).