Justificationists often complain that Popper’s philosophy doesn’t enable us to pick out one scientific theory as being decisively the best at any particular time. But there is something extremely odd about this objection.
Critical rationalists advocate various restrictions on what theories we ought to prefer and those restrictions are quite hard to meet: a theory has to be experimentally testable, but not yet refuted, non-ad-hoc, explains how to solve problems and so on. So it isn’t the case that we say theories should be picked at random. We have tough standards. What criticism do justificationists have of those standards other than that they don’t justify anything? And that complaint is an empty one because there are unrefuted arguments indicating that justification is impossible.
Why should it be possible or even desirable to pick out just one theory? Different theories solve different problems and no single theory solves all of our problems. So why should we want to force ourselves to adopt just one of the inadequate theories floating around?
Let’s take an example. Quantum mechanics and general relativity can’t both be true, so we ought to eliminate one of them, right? But trying to say we shouldn’t use one of them when we don’t have a better replacement is only going to restrict our ability to solve problems, not increase it. And in any case, we can generate knowledge about the flaws in our theories and refrain from applying them to situations in which they’re useless. In other words, we can develop a preference about what we ought to do in any particular situation. Sometimes that preference will involving using quantum mechanics to do a calculation and sometimes we will want to use general relativity instead. So what problem is solved by coming up with a rule that prevents us from using quantum mechanics and general relativity in situations where we have no better option? I think quantum mechanics and general relativity will be replaced by better theories, but until then what would be the point in picking just one of them?
Also, sometimes scientists disagree with one another. Perhaps sometimes one side of a dispute is actually right and the other side is actually wrong. However, I can imagine that there are many occasions when the disagreement is just due to the different sides of the dispute paying attention to different problems. In other words, both sides are wrong and they have something to learn from one another. The idea that there should be some mechanical rule picking one side rather than the other as right and that this should be built into our epistemology seems unnecessarily restricting.
Moreover, scientific progress comes from generating lots of variations on our theories and then eliminating ones that don’t work. Eliminating variation by instituting a rule that picks just one theory would lead to stagnation.
UPDATE: A brief addition to make this post clearer. Any given theory is either true or false. We want to have true theories. We try to get true theories by solving problems: that is, by coming up with new theories that address problems with current theories. We do this by coming up with many different theories and discussing what problems they solve. Different people will have different ideas about what problems are important. This is appropriate in many cases because we learn what problems are interesting through conjecture and criticism. This means we can’t lay down hard and fast rules for what theory is the best at any particular time.
That point blows up in the face of the justificationist: on their own terms they have no way to pick one theory as decisively better than another.
In practice of course they have to resort to critical preferences like critical rationalists and everyone else. By some strange twist of illogic they try to make out that we are the oddballs!
As you said, theories are normally selected for a purpose, so you have different criteria for different purposes.
Non Popperians only need to maintain that, all other things being equal, subtracting positive support will result in less ability to sort or grade theories. Being able to exclude all but one theory is a further step, and confusion between the two is just a further variation on the confusion between justification qua infallible proof, and justification
qua positive support.
If Popperians can replace positive support with some other novel criteria, they can defuse the objection; however the claim that mainstream de facto science does not care about explanatory value or testability could be hard to defend.
The origin of the idea that there is ultimately only one really true theory is the correspondence theory of truth: that there is only one way the world is. Popper held to this: he was a realist, and it does him no favours to recast him as an instrumentalist in response to his crtics.
Pragmatic or instrumental considerations can still lead realists to juggle multiple theories that are flawed in different ways, since realism is an “in principle” theory.
Peter D wrote:
Non-Popperians have not explained what positive support is, how it solves the problem of induction, or how it deals with the tu quoque problem.
I don’t see why Popperians should have to replace something that doesn’t exist.
I think scientists do care about explanatory value and testability. I think philosophers like to mix those things up with justificationism, and this leads to bad philosophy.
Many non-Popperians continue talking about positive support as if Popper and Miller had not proved that there’s nothing ampliative about it. Simply saying “Oh I’ve heard that proof is flawed” or something along those lines isn’t good enough. They need to point out where the error is. Otherwise, any further talk of support is simply a waste of time.
Non-Popperian philosophers of science are very much engaged in explaining positive support and induction. (Some even have mathematical arguments). Have you read any?
If Popperians are ditching positive support and if science is already using every criterion Popperians recommend, the argument goes through. All other things are equal, and without positive support Popperians are less able to sift and grade theories.
I do not deny the existence of positive support per se. Insofar as criticism changes the relative standing of competing hypotheses, it must be negative for some and positive for others. To put it another way, the contention seems to concern the nature of positive support, rather than its existence in a purely logical or mathematical sense.
It seems as though conventional philosophers treat support like physicists once treated space and time. That is, an absolute frame of reference is assumed, against which the movement of objects can be reckoned — in this case the standing of competing hypotheses. We see this most clearly in the arguments of irrationalists, like the tu quoque, which purport to criticise all competing hypotheses at the same time. For critical rationalists, there is no absolute frame of reference. If a criticism does not change the relative position of hypotheses, then it is not a criticism of anything in particular*. Admittedly, this is something like irrationalism, but while irrationalists throw out truth with justification, critical rationalists do not.
If positive support is reckoned by this absolute frame, which itself is unsupported, then positive support is impossible. However, the concept is not altogether objectionable when defined in a weaker, more purely logical or mathematical, sense.
*This is why I have suggested the dictum: “an objection that can be brought against everything ought not to be brought against anything.”
By the way, when I say there is positive support, I am not referring to induction.
Consider competing hypotheses A and B. Before experiment e, A and B have equal standing. The purpose of the experiment is to support A. Since neither A nor B are implied by e, these are the possible results:
1) Both A and B are consistent with e. The relative standing of each has been unchanged. Neither has been supported or criticised.
2) Both ~A and ~B are implied by e. The relative standing of each has been unchanged. Neither has been supported, and (supposing a another possibility exists) both have been criticised.
3) Either ~A or ~B is implied by e. The relative standing of each changes, either either A or B or supported.
Suppose that A is supported by t. Notice that A is only positively supported if B is negatively supported. It is a zero-sum game. Conventional philosophers, meanwhile, seem to believe that both A and B can be supported at the same time — such as when e is consistent with both. Support is not zero-sum, but accumulative. The total amount of support we have for A^B after e may be greater than before — not so for critical rationalists.
Notice also that, because positive support is relative, a hypothesis cannot be supported in and of itself, in vacuo so to speak.
Peter D. Jones wrote:
I have read many of those arguments. They are all flawed. As a matter of fact, no finite set of events logically implies anything about the future. Any alleged positive support can be undermined by pointing out that all of those arguments make assumptions about the way the world works and that these assumptions are hypotheses.
A short list of hypotheses implicitly made by Bayesian epistemologists: it is appropriate to discuss knowledge in terms of belief, and there are degrees of belief, and they obey the calculus of probabilities, and that those probabilities are linked in certain ways to the probabilities predicted by physical theories. (All of these hypotheses are false BTW. For example, saying knowledge is belief would imply that books contain no knowledge.)
Those arguments are exactly as ‘justified’ (proven true or likely to be true) as the assumptions that go into them, which is to say they are completely unjustified. And any piece of evidence itself is hypothetical – it depends on hypotheses about how the relevant experiments work.
A usual, that is a series of arguments that tacitly assumes nothing can be shown to work unless it can be shown to work infallibly.
If a deductive justification can be given for probablistic inference, the fact
that an argument uses hypotheses does not render it invalid, since hypotheses can be given apriori probabilities — and must be even to make Popperism work.
There are major problems involved in assigning aprori probabilities, or there were ten years ago when I investigated the matter for my thesis on the Duhem problem. Until they are resolved Bayesian subjectivism has not got to first base.
I still want to know how working scientists would use the kind of probabilities (attached to theories, not events) that inductivists want to generate.
Lee,
It is not a zero sum game, since we have no reason to suppose that the set of hypotheses we have at a point in time are exhaustive. Evidence that counts against all of them impels us to seek a new one.
The implication that a solitary no theory cannot be raised in likelihood by confirmatory evidence is a major problem . It suggests that the startling and accurate novel predictions of relativity would have counted for nothing if there had been no rival theory.
Peter,
In the context of critical preference between competing hypotheses, “support” really is zero-sum. Because critical preference concerns hypotheses that are “on the table”, so to speak, the existence of other logical possibilities is irrelevant.
What about a rule in the philosophy of science that all papers should relate to some actual situation in science (past or present)? That would have two benefits, first, readers would learn something about science even if they did not learn anything about philosophy, and second, philosophers would write less papers because they would have to spend some time learning about science. If the philosophers are too busy to learn a lot of science they could collaborate with one or more scientists as co-authors of their papers.
Rafe
Most philosophers of science are well versed in science already, and already provide appropriate examples from its history. Kuhn had a PhD in physics for instance
Rafe
There are problems with all positions — these questions are open.
Popperism needs some sort of apriori probabilities too, so long as it maintains that factual evidence is conjectural.
Scientists regularly rank theories according to considerations of likelihood. Putting an exact figure on them is another matter.
Peter,
Anti-justificationism can be usefully seen as the proposition that, while we may obtain a non-arbitrary ordinal ranking of theories, there can be no such cardinal ranking. In other words, we can rank theories only in a relative sense. This doesn’t imply that the best corroborated theory is necessarily the best theory: corroboration merely demotes criticisms. Thus it may change the relative rankings among competing theories. A theory so corroborated may as a result move to the head of the pack, but it need not do so. In any event, it is highly misleading to say as you do that a criticism of one theory, though negative for the theory being criticized, must be positive for some competing theory. When you say “positive” in this context you’re merely exploiting an ambiguity. Sure, a successful criticism of one theory may clear the way for a competitor, and in that sense be positive for the competitor. But the notion of positive evidence or support is a different matter entirely — a fact which your offhand remark conveniently leaves out of account. There simply is no sense at all in which a successful criticism of one theory must constitute positive support for a competitor.
The example of Kuhn shows that you can study the history of science and still make no useful contribution to the philosophy and methodology of science. Peter, perhaps I have got Kuhn wrong, can you explain what problem he advanced, what theories he contributed that are novel, robust and helpful for working scientists?
I think some questions are not open. After Godel certain lines of investigation became closed. It looks as though the same applies to the quest for p values to be attached to theories.
Even if that line of investigation was not closed on grounds of logic there would still be the problem of how working scientists would actually use the p values. I don’t think any explanation has been provided, beyond the claim that we all have to use “some sort of apriori probabilities”. But that is subjectivism and in the critical rationalist philosophy of science we are concerned with theories that are stated in an objective or inter-subjective manner and our own subjective thoughts are not the issue.
Kenneth
That theories can only be ranked ordinally may well be the case, but it should not
be called non-justification since it is entirely compatible with positive support
Rafe
I don’t accept that science necessarily needs fixing. I also don’t accept
that philosophy of science need do anything but understand how
science works.
Inductivism attaches formal p values to inductive generalisations, not theories, which are more complex entities.