Justifying theories II: Popper and falsification
The problem of induction arose because we supposed that scientific generalizations were supported by their instances. But Karl Popper, who, like many of the twentieth-century philosophers I have mentioned, was associated (though rather antagonistically!) with the Vienna Circle, argued that this was a mistake.
Hume, Popper argued, was absolutely right. Laws are not supported by their instances. What happens in the sciences is that people such as Mendel creatively invent hypotheses. They then set out to examine their instances, not because the instances support the laws but because they know that if the instances don't support the laws, the hypotheses are false. Science, in Popper's view, does not proceed by induction and the verification of true theories. Rather, we go on with the hypotheses we make until they are falsified, until, that is, experience shows that they are not true.Popper relies here on a simple logical fact, a fact about predicate logic. The problem of induction arises, in his view, because for the law that “All A's are B's” to be true, there must not be one single A that is not a B. It follows that until we have examined every single A, we cannot be sure that the law is true. But, by the same token, we only have to find one A that is not a B in order to show that a law is false. So, while we can never be sure that a law is true, we can, apparently, be sure that a law is false.
Popper, then, doesn't solve the problem of induction, but, as he says, he dissolves it by showing there never was such a problem. There is no problem of induction in science because scientists do not proceed by induction. Rather, they proceed by conjecture— that is, imaginatively inventing new theories—and then make observations and do experiments that may lead, in the end, to refutation.
Then they try out new theories, and another cycle of conjecture and refutation begins. class=a2 style='text-indent:18.0pt'>Popper's rejection of inductivism is radical. He denies that we are ever justified in believing that scientific theories are true. Science does not produce knowledge because it does not produce justification; and so we shouldn't really believe scientific theories. We may accept them until they are falsified; but accepting a theory, for Popper, is not the same as believing it to be true. To accept a theory is to keep using it provisionally in the knowledge that at any moment observation or experiment may force us to give it up. One way of putting Popper's view is to say that he takes fallibilism very seriously.Because Popper places such emphasis on the fact that scientists give up theories that are false, rather than insisting, as classical empiricism did, on trying to find theories that are true, his position is called “falsificationism.” Indeed, Popper's answer to the demarcation problem is that what makes a statement scientific is just that it is possible to falsify it.
Popper's position has won a good deal of support among scientists, who have the experience all the time of having to give up theories because experiments show them wrong. They probably also find flattering the fact that Popper insists on the importance of the creative process of conjecture! More important, the fact that scientific theory making is, indeed, not a simple matter of generalizing from examples you have collected fits well with Popper's view. No amount of hard work collecting instances will lead to a new theory, in Popper's view, without the original creative act of the human mind. Popper's claim is, in essence, that we are justified in using theories not because we have evidence that they are true, but until we have evidence that they are false.
Despite its popularity among scientists, there are certainly problems with Popper's view.
To begin with, the simple logical point I made just now is really not so simple as it seems. It is true that whenever we have evidence that one A is not a B, we have evidence that it is false that all A's are B's. But in order to find out that one A is not a B, we always have to rely on other generalizations. (This fact, which was pointed out by the French philosopher-physicist Pierre Duhem and built on by the American philosopher W. V. O. Quine, is sometimes called the “Duhem-Quine problem.”) Thus, to find a homozygous purple pea that does not produce purple offspring when crossed with a homozygous white pea, I have to rely on such generalizations as the (rather elementary) law that homozygous WW peas look white. If I am not entitled to assume that this law is true, then I am not entitled to believe that I have found a white offspring of such a cross.Of course, this particular law is one that we are rather sure of. But in many crucial experiments we rely on a whole lot of highly theoretical laws in order to show that an old theory was wrong. Many of the experiments that showed that Mendel's laws of segregation and independent assortment were wrong involved theoretical assumptions about what was going on in particular crosses.
Moreover, Popper's theory makes it difficult to understand why science seems to progress. On the DN theory of explanation, old theories are often reduced to new ones, so that we show that the old theory is a special case of the new one. But on Popper's view, all that we are entitled to keep from the old theory are the instances where it succeeded and not any of the laws. Once the old conjecture is falsified, we are free to make any new conjecture that is consistent with the existing data. The claim that this is how science actually proceeds—throwing out the old theories and starting again from scratch—is hardly consistent with the historical evidence.
A final difficulty with Popper's view is that it is highly counterintuitive to say that we never have any reason to think that theories are true.
For the Popperian, the relevance of experimental evidence is not that it confirms the truth of our theories. Indeed, Popper explicitly rejects all inductivist talk of scientists confirming theories. Rather, evidence is relevant because theories that have survived rigorous testing are what Popper calls better “corroborated” than those that have not. But if corroboration provides no reason for thinking a theory is true, why is it a reason for accepting it at all?This question is especially urgent because for any well-corroborated theory—any theory, that is, that has survived rigorous test- ing—there are infinite numbers of different and incompatible theories that have not been tested but which are consistent with all the existing evidence. Of course, no one has even thought of most of them, and many of them are likely to seem just silly. But the point is that so far as Popper is concerned, they have just the same chance of being true as the well-corroborated theory. If the evidence of experiments does not give us reason to think that our theories are true, why should we prefer theories that have survived experimental testing to other as-yet-unfalsified theories that have not?
This question is a very serious challenge to Popper's philosophy of science. Nevertheless, without a solution to the problem of induction, Popper's theory at least provides a way of explaining what we do in science that does not depend on a form of argument, induction, that seems to be unjustifiable.
Popper's theory and inductivism each offer an answer to the demarcation problem. Inductivists say that theories are scientific if they are based on inductive evidence. This means that the criterion of demarcation belongs to the context of discovery. It has to do with how we come to believe the theory. An inductivist would say that the astrological beliefs I mentioned at the start are unscientific because they were not properly derived from and supported by inductive evidence.
But Popper's view is that how we came to believe our laws has nothing to do with what makes them scientific. Rather, what makes them scientific is that they are always open to falsification. For Popper, astrologers are unscientific because their theories are so vaguely formulated and so hedged with qualifications that they could never be shown to be false. So Popper's demarcation criterion belongs to the context of justification.
4.12