A primary motivation for finding a place for useful falsehoods resides in wanting to explain how one can use a false theory to acquire knowledge. Ptolemaic astronomers knew where the planets would be at fixed points in the future, and Newtonian physicists can predict the behavior of medium-sized objects close to the earth–“predict” in the sense that implies knowledge. The reason they have knowledge of this sort is that objects behave as if the theory is true even if in fact the theory is false.
This instrumentalist fact guides Klein’s theory. He requires that the useful falsehood entail a truth that justifies the same belief justified by the useful falsehood. The instrumentalist claim is entailed by the false theory, is true, and justifies the same belief justified by the useful falsehood. It’s exactly the claim he’s looking for (and cites).
It is interesting, I think, to compare the situation of ancient Ptolemaic astronomy with ancient astrology. In both cases, the theories are subject to anomalies, but were accepted because of some predictive success for the theories. So, instead of considering the situation of the behavior of a heavenly body, consider an astrological prediction of a baby becoming a well-known public figure in virtue of being born on a certain day. Further, suppose the astrological theory has been tested sufficiently for such cases that the body of information available justifies believing that the child will be famous, and assume that the prediction is correct.
Here, the instrumentalist fact threatens to give the wrong results, doesn’t it? We’re not very tempted to say that the astrologers knew the child would be famous. So what’s the difference between ancient astrology and Ptolemaic astronomy?
We can’t say, “they knew better than to rely on such a sorry theory.” I don’t believe astrology was looked on as intellectually inferior at the time in question. In addition, we can’t say, “we know better, so they didn’t know.” That’s not relevant at all, and if it were, it would affect both astrology and astronomy.
As far as I can tell, Klein’s theory will have to find a genuine defeater for both evidential paths, thus being a genuine defeater that is not merely the denial of the astrological theory in question. Presumably, there are such defeaters: for example, situations in which the astrological theory gives wrong predictions. The problem is that such defeaters exist for all but very carefully crafted versions of Ptolemaic astronomy, and we aren’t inclined to think that the less carefully crafted theories can’t be used to gain knowledge. (Not to mention the fact that, for any contrary experimental result, there will be a way to accommodate it within astrological thought as well…)
It is tempting here to abandon the defeasibility theory in favor of reliabilism, since a reliabilist only needs to explain how the astrological inferences are not reliable. Things are not quite that simple, though, since some astrological inferences are reliable, and there are ways to adjust the theory to correct for any particular unreliable inference (I assume). This is just a particular instance of the generality problem, but it’s important to note that going reliabilist here won’t somehow make the problem go away.
So, what’s the difference between ancient astronomy, which we want to allow can yield knowledge, and ancient astrology, which we don’t want to allow to yield knowledge? Or should we swallow hard and admit that the astrologers knew the child would be famous?