Michael Huemer has made my life easier. One of my tasks at the Mises Institute is to teach praxeology to students, and doing so involves explaining a priori knowledge (i.e., what we can know just by thinking about it), a notion which many students find difficult to grasp. The task becomes even harder when you add that the a priori knowledge in question is “synthetic” knowledge that isn’t analytically true but that we can still know to be true just by thinking about it.
In order to accept synthetic a priori knowledge, must we embrace Immanuel Kant’s notoriously difficult theory of knowledge? If we decline to do so, Murray Rothbard has an alternative way to justify synthetic a priori knowledge; he appeals to Aristotelian essences, or natures. I follow him in this, but, once again, this isn’t the easiest view to understand.
Michael Huemer, a brilliant philosopher who teaches at the University of Colorado Boulder, rejects a common assumption many philosophers have about synthetic a priori knowledge; namely, that it is highly puzzling. In his superb new book, Understanding Knowledge, he maintains that it is quite easy to justify the synthetic a priori. To be clear, Huemer is not attempting to support praxeology, and my impression is that he wouldn’t accept it; he is an exponent of Chicago school price theory along the lines of his friend David Friedman. Indeed, one of Huemer’s comments about knowledge poses a problem for praxeology, and after discussing Huemer on the synthetic a priori, I’ll try to respond to it.
Huemer makes a point about a priori knowledge that I have found again and again students miss. If you claim that people have such knowledge, you are making a claim about propositions, not about concepts. If you think that concepts are abstractions from experience—Huemer doesn’t, by the way—you can consistently accept a priori knowledge. Huemer says:
Now here’s another thing I have to clarify because I know some of you readers are already making this mistake: The distinction between “empirical” and “a priori” knowledge is not about how you acquire concepts. So don’t say that some item of knowledge is empirical because you acquired the concepts through observation. For example, don’t say that “all grandsons are male” is empirical because we acquire the concepts “grandson” and “male” through experience. . . . “All grandsons are male” is considered to be known a priori because you don’t have to justify it by citing observations that you’ve made of grandsons (or of anything else) (You just have to understand the meaning of “grandson.”) (emphasis in original)
This should be enough to satisfy Objectivists that they have nothing to worry about, at least on this score, when claims to a priori knowledge are advanced, but I’m betting it won’t be.
With that out of the way, we can turn to Huemer’s biggest contribution to understanding the synthetic a priori. Many people have no trouble with analytic a priori knowledge. This is a matter of definitions, parts of definitions, or tautologies like “It’s either raining or not raining.” But according to the logical positivists, this sort of knowledge doesn’t tell us anything about the empirical world. We have such knowledge just because we use words in a certain way. If we want to know what the empirical world is like, don’t we have to go beyond appeal to the meanings of our words or the concepts our words express? Aren’t truth claims about the world synthetic (i.e., nonanalytic) propositions? And if they are, how can we know they are true just by thinking about them?
If you are doubtful about synthetic a priori truth, you don’t have to give up praxeology. You can appeal to the notion of tautology. Some people (wrongly) believe that mathematics consists of tautologies, but, even in their view, we can certainly learn something by proving a theorem; we aren’t just idly reiterating what we already know. Ludwig von Mises adopts this approach in Human Action, claiming that the truths of praxeology are analytic but by no means useless, but by the time of The Ultimate Foundation of Economic Science, he seems receptive to the charms of the synthetic a priori, pointing out, as does Huemer, that the assertion “There are no synthetic a priori truths” appears to be itself a synthetic a priori proposition and thus, if true, self-refuting.
But how can we explain how synthetic a priori truth is possible? Huemer’s response will surprise you if you aren’t already familiar with his “bottom-up” approach to epistemology. He points out examples in which we know things by thinking about them but what we know does not consist of analytic propositions:
The main reason people are rationalists [i.e., accept synthetic a priori knowledge] is not that they thought about it and saw intuitively that synthetic, a priori knowledge was possible. The main reason is that they looked for examples of knowledge, and they saw some that appear to be synthetic, a priori. . . . For instance, you know in advance that you’re never going to find anything completely red and completely blue. No matter what an object looks like, you won’t say that it’s both red and blue all over, so there is no need to look.
This claim about color isn’t an analytic truth because it isn’t part of the concept of a color that a surface can have only one color all over.
The vital core of Huemer’s case for the synthetic a priori is that we don’t have to come up with a theoretical account of how it is possible: we know there is such a thing because we have examples of it. He offers his own account, but he disarmingly says of it, “Typically, we rationalists just say some vague stuff about ‘grasping the natures of abstract objects’ and such like. I’m no exception—that’s just what I’m about to do here.” (Huemer does better than this suggests.)
And this brings up the difficulty for praxeology that I mentioned earlier. Huemer suggests that our knowledge is not typically top-down or deductive. But doesn’t praxeology proceed deductively? Does this pose a problem for us? Huemer says:
Most human knowledge is bottom up. That is, one starts from a large number of cognitions about specific cases or particular individuals. When one has enough cases, one can start to see patterns and general rules. One then starts to formulate abstract principles based upon the cases. . . . If one tries to start from the abstract principles, there is almost a 100% chance of going wrong, often disastrously so.
One way to answer Huemer would be to say that praxeology is an exception to his generalization, but you can accept what he says without ditching praxeology. The deductive structure of praxeology need not be taken as an account of how Carl Menger and his followers first discovered praxeological insights. Rather, they thought about particular cases and realized certain truths about them. The deductive structure came later, as a way to systematize these insights. Further, even within the deductive structure, praxeological reasoning is not top-down in the strictest sense.
If a theorem occasionally leads to results that strike us as intuitively implausible, we can reinterpret the theorem so that it avoids these consequences. To use the term emphasized by William J. Talbott in his important book Learning from Our Mistakes, our reasoning is not monotonic, in the sense that we can only go forward from premise to conclusion. We can rethink our premises in the light of how they turn out. (See my review of Talbott’s book in the Philosophical Quarterly.)
I highly recommend Huemer’s book, which among many other good features, is very funny. For example, a blurb on the back “quotes” Kant: “Thoughts without content are empty; intuitions without concepts are blind, and students without Mike Huemer’s books are dumb.”