Samizdata

Samizdata.net wheels out another of our ‘mercenary independent scholars’ in the Interblog Popper Wars. Alan Forrester!

Karl Popper‘s epistemology is about how to solve problems and find better theories, and as such observations do not have the grossly overrated importance given to them by inductivism. The whole notion of probability in this context is irrelevant, since a theory is either true or false and no number of confirming instances allow us to distinguish between them. On Thursday, Will Wilkinson wrote:

It is daunting indeed to debate a man named “Rafe Champion”, a name that evokes race car-driving secret agents, or dangerous, seething, family-wrecking hunks from a “daytime drama”.

Damn straight he is. He does all that, eats broken glass for breakfast and, most excitingly of all – he’s a critical rationalist!

First, I am keen to know what knowledge is, if not a kind of belief. If I know that water is H²O (a scientific proposition), don’t I also believe it?

Not in the sense you mean, i.e.- the sense that it is definitely or probably correct. We tentatively accept that a theory is better than its rival on the basis of things such as whether they provide good explanations and whether the other theories have been refuted by observations.

Next, I find that I’m able to decisively justify all sorts of beliefs on the basis of experience. For instance, that there is a mug on my desk. I see the mug on my desk, and I thereby know that it is there. Science is rather more complex than looking at mugs on desks, but one surely can derive certain beliefs from the evidence of the senses. It’s not clear to me what bind Popper is getting us out of.

So it’s totally impossible that you are hallucinating the mug? Also, why are you so obsessed about whether there is a mug on your desk or not? It’s not a very interesting question. I would sooner debate about something substantial like a meaty scientific or philosophical problem, which brings us back to the main point. Leaving quibbles like that aside, there are actually two quite different issues here. One is whether one can derive theories from observations, the other is if not what role do observations play?

It is impossible to derive theories for observations. To take but one counterexample, up until the late 19th century every observation was compatible with Newton’s theory of gravity. All these observations are also compatible with Einstein’s General Theory of Relativity. Two quite different theories were compatible with the same set of observations, therefore one cannot derive theories from observations. Next we have to ask why you made those particular observations, rather than observations of, say, the exact weight of all the dust under your bed. Before the theories come along that differ in their predictions about the motion of the planets or whatever, there is no particular reason to watch their motion so theories can hardly be derived from observations. The problem is much worse than this though. Even if one could confirm, say, an equation of motion on the basis of observations, one could not derive the explanation provided by the theory from observations. For example, one could explain the motion of the planets by saying that they are pushed round the Sun by invisible pixies that just happen to obey Einstein’s equations. This explanation is rubbish on a truly epic scale compared to the explanation in terms of curved spacetime, but as the observations do not allow us to distinguish between them, the pixie theory of planetary motion must be criticised on other grounds. The role of observations is to distinguish between rival theories. Each prediction of a theory is either true or false, and each theory is either true or false, no number of confirming instances can change that, and hence they cannot prove it to be true, but if we see a refuting instance, then the theory is false.

Last, I said nothing about limiting science to collecting confirming instances. All I was saying is that Popper is wrong that positive instances don’t raise the probability of a hypothesis. According to Popper and Champion, the probability of Newton’s theory being true, even after all its success, was the same as the probability of cats giving birth to elephants.

Newton’s theory was false, so as it turns out the probability was indeed the same :-). However, all this talk of probability ignores the real issue of why, before it was refuted, it was reasonable to hold Newton’s theory, but not the theory that cats give birth to elephants, or why it is reasonable to prefer General Relativity or whatever to such a theory now. It is a good idea to prefer General Relativity to its rival because its rivals are poorer at solving problems.

It is reasonable to prefer the theory that elephants give birth to elephants rather than cats because the former solves problems and the latter does not, indeed it raises new ones. To be a bit more explicit, the theory of evolution, which is better than its rivals (although I don’t think there are any really serious rivals at the moment), gives us reason to think that elephants give birth to more elephants as a way of spreading elephant genes. The theory that cats give birth to elephants not only fails to solve any problems, it ruins theories that do solve problems. It makes absolutely no sense from an evolutionary perspective why would cat genes want to propagate elephants genes? Where did the cat get the elephant genes in the first place? Furthermore, no matter how you try to solve these problems, you just raise more and worse problems, so we can reject the theory that cats give birth to elephants out of hand. To summarize, good theories solve problems better than their rivals, but raise problems themselves, which will be solved by their successors and so we don’t consider that observations and so on confirm a theory.

Alan Forrester