Scientific Knowledge: Shouldn't GPOS Get Out of the Game?

Eric’s recent post about epistemology illustrates the identity crisis that General Philosophy of Science (GPOS) is currently suffering. According to Eric, observation “plays no foundational role whatsoever” in many (or even most) sciences, a conclusion he draws from the fact that measurement can be delegated to machines–the human senses, and hence “observations”, play no essential role. Eric argues that “Gettierized Epistemology” (I prefer to call it General Epistemology, but we can agree on the acronym, GE), preoccupied as it is with “ordinary cognition, perception, and locution,” has no way of coping with machine measurement, and hence not with science as it is today, and has been for going on two hundred years. GPOS asks the transcendental question: How is scientific knowledge possible? “So,” Eric writes, “GPOS is a in bad position to answer the skeptic, except to remark that within the scientific machinery one is not much perturbed by it.”

Now, the question of machine measurement ought really to be put to one side. It may very well be true, for instance, that cognitive neuroscience depends heavily on fMRI readings, and that these are highly automated. It may be that as far as cognitive neuroscience goes, these machine-generated data entirely supersede human observation. But it does not follow that GPOS should take the same position. The GPOSer wil point out that the epistemic authority of fMRI data cannot be taken for granted, and will try to provide a framework for within which we can better understand the inferential process that seems (wrongly) to start from these data alone. This kind of analysis is sadly neglected in philosophy of science, but it exists. My bottom line: resist parochial scientific judgements about the fundamentality of data provided by machines.

The broader question is: how does GPOS deal with scientific knowledge? How does it deal with sceptical challenges to science?

Well, how does GE address scepticism? First, what is scepticism? In a recent article, "How to Be Sure", I offer a structural characterization. Start with this: proposition q is a “doubt condition” for proposition p, if q falsifies p, and I am not certain that q is false. q is a sceptical doubt condition for p if it is a doubt condition that is not specific to p. Here is an example. I am looking at a ball in my left hand. It looks pink, and I think it is so. However, if the illumination is reddish, the ball might actually be white despite its pink appearance. Suppose that I can't rule out that the illumination is reddish. Then this is a doubt condition for the ball being pink.

Note, however, this doubt condition is specific to the colour of this ball now. It does not affect my belief that the object that I hold in my right hand is round. Consider, however, the proposition that I am a brain in a vat. This is a doubt condition not only for the colour of this object, but also its shape, and, for that matter, everything else I observe. The illumination doubt condition is empirical; it is specific. The brain-in-a-vat doubt condition is sceptical; it is completely domain-general.

GE does have various strategies against scepticism. But one relevant move that it can make (and this is the move I make in "How To Be Sure") is to point out that there are other sorts of doubt, more specific kinds of doubts, and that these are addressed differently than sceptical doubt. In my paper, I am concerned with doubt regarding certain perceptual judgements. That thing looks round; that thing looks pink: putting aside sceptical doubts, how should I check whether I am perceptually in error? I can check whether I am in error about the colour of the ball by taking it to the window and letting the sun shine on it. I can check its shape by grasping it in my hands and turning it over. These are methods of verification that will not dispel scepticism but will undermine more specific doubt conditions. In a similar vein, one might ask about testimony. How can one verify it? There are answers to this question too that fall short of undermining sceptical doubt–cross-examination, corroboration, etc.

GPOS has extensively discussed these more specific ways of verifying scientific data, whether these data originate from people or machines. But much of GPOS is devoted to the discussion of sceptical doubt. Regrettably, the latter dominates the subject–think of Popper, think of Feyerabend–and it is this vein of inquiry that gives rise to conventionalism, pragmatism, and other anti-realisms in philosophy of science. These positions are not specific to science, however, for scepticism itself is not specific to science. Inasmuch as general philosophy of science is concerned with science, and not with “ordinary cognition”, it should not deal with scepticism. Hume's sceptical doubt regarding induction affects day-to-day life as much as it affects scientific method, and it should be assigned to GE.

But this leads us to an uncomfortable dilemma about GPOS’s area of concern. As far as scientific knowledge is concerned, it could deal with such general topics as induction, abduction, observation, and experimentation. But these topics too are relevant outside science. In science, they work in the same way as in ordinary cognition. So these inquiries properly belong to GE. On the other hand, philosophers could deal with more domain-restricted questions such as the one mentioned earlier concerning fMRI data. Once again, however, this is not the concern of GPOS; rather, it belongs to the philosophy of cognitive neuroscience. So what question of scientific knowledge is left to GPOS?

The crisis of identity that GPOS faces has a lot to do with the difficulty of answering this question.