Saturday, November 20, 2010

Scientific realism

Despite having a pretty good Pittsburgh education in the philosophy of science, I never before read Ernan McMullin's "A Case for Scientific Realism". I was especially struck by one thing that I had never noticed before, which Fr. McMullin briefly notes in one context: things are different, realism-wise, in regard to fundamental physics and other areas of science. The rest of this post is me, not McMullin.

Observe that the pessimistic meta-induction works a lot better for fundamental physics than for the special sciences. The meta-induction says that past theories have tended to be eventually refuted, and hence so will the present ones be. (It's really hard to make the statement precise, but nevermind that for now.) But it is false that the special sciences' theories have tended to be eventually refuted. Some, like the geocentric and heliocentric theories in astronomy and the phlogiston theory of combustion, have indeed been refuted. But many theories have stood for millenia. Here is a sample of these theories: (a) there are seasons that come in a cycle, and the cycle is correlated with various botanical phenomena; (b) tigers eat humans and deer; deer eat neither tigers nor humans; (c) rain comes from clouds; (d) herbivores run from apparent danger; (e) much of the earth's energy comes from the sun. And so on. We do not think of these as scientific theories any more because they are so venerable and well-confirmed. This means that we sometimes mistakenly assent to the inductive premise of the meta-induction because those venerable scientific theories that have not been refuted have often become common-sense and hence we exclude them from the sample.

Nonetheless, the pessimistic meta-induction seems to have some force in regard to fundamental physics: there, the change is much more rapid, and very little remains of past theories. We do sometimes get results like the "classical limit" theorems for Quantum Mechanics where we can show that the earlier theory's predictions approximated the predictions of the newer theory, but this approximation in prediction does not typically yield the approximate truth of the earlier theory. The one kind of exception we sometimes get is that sometimes a part of what used to be a fundamental theory survives, but no longer as fundamental—atoms, for instance.

Non-fundamental concepts—such as cell or season—can survive significant shifts in fundamental theories, but obviously fundamental concepts like force or particle find it much more difficult to do so. There is a kind of multiple realizability in the concepts of the special sciences (not along the metaphysical but the conceptual dimension of a two-dimensional modal semantics) which makes them more resilient.

Van Fraassen proposes we be realists about the observable claims of science and non-realists about the unobservable. This is, I think, really implausible. Van Fraassen would have us believe in ova but not in sperm, just because the ovum is large enough to be seen with the naked eye while a sperm is not. But I think there is a view in the vicinity that is worth taking seriously: that we should be realists about non-fundamental science and at least somewhat skeptical of fundamental science.


David Parker said...

You reminded me of a paper bookmarked on my science wars reading list . It's called "Defense of a Modest Scientific Realism" by Alan Sokal.

Kenny said...

How often do well-confirmed fundamental theories really turn out to be wrong, as opposed to merely turning out to be non-fundamental? Wouldn't it be better to just refrain from believing that our (so-called) 'fundamental' theories are really fundamental?

Alexander R Pruss said...


I think both the atomic and homogeneous matter theories of the Newtonian era have turned out to be wrong. Matter is not homogeneous and is not made out of hard atoms. It is tempting to say: "Well, they got the atom part right, and the hardness part wrong." But both were relevant explanatory parts of the theory.

It is not yet clear whether the final physics will yield anything that answers to the Newtonian concept of force. At least, there really doesn't appear to be anything like gravitational force.

Going further back, none of the pre-Socratic theories about the constitution of things have turned out right--not everything is made of fire, earth, water and air; not everything is made of hard atoms with hooks and the like; etc.

Now, if all one wants for the theory to count as right is that it generate approximately correct empirical results in its proper area of application, then pretty much by definition well-confirmed fundamental theories are going to be right.

Kenny said...

I'm not sure, on this view, what it would mean to say that special science theories are right. It seems like what happens is that we learn that our law is a ceteris paribus law, and then we figure out what the conditions of its application are. So, for instance, in the thermodynamic derivations of the ideal gas laws, we think of things in terms of perfectly hard atoms bouncing off one another. So it seems like what we thought was part of a fundamental theory turned out to be only part of a special science theory.

All or nearly all of the special sciences, I should think, involve models with entities that aren't part of the models involved in our fundamental theories.

Alexander R Pruss said...

"It seems like what happens is that we learn that our law is a ceteris paribus law, and then we figure out what the conditions of its application are."

Maybe. I am worried about triviality now. It's like saying that Thales' Law is correct: ceteris paribus, an object is made of water. But sometimes things aren't equal, and there are impurities. :-)

Kenny said...

I agree that that's a worry, but I'm not sure how to solve it without adopting some form of anti-realism. I mean, the difference, it seems to me, is that there are circumstances in which, e.g., classical mechanics is the most useful theory for some purpose. But it seems like, for any given purpose, we always have a theory that is more useful than Thales'. But if you are going to hold on to realism about special science laws, then I'm not sure how you will solve this problem, and I'm also not sure why we shouldn't think that rejected fundamental theories very frequently become special science. It certainly seems like this is what has happened with classical mechanics.

David Parker said...

Forgive an amateur for harping in. My intuition keeps insisting that if scientific measurements are finitely accurate, then no possible fundamental system can ever be "reached." Classical mechanics is an approximation but quantum mechanics is a better approximation...but ad infinitum? It seems like a potential infinity, and thus science only approaches a fundamental theory.

Do you either of you know if has this kind of thinking has been explored or rebutted? Thanks.

Alexander R Pruss said...


I am inclined to think that when a fundamental theory becomes a special science, it becomes significantly changed. Certainly, the claim "This is fundamental" gets removed. I think there may be another (perhaps consequent) serious modification of the ontology: the ontology becomes more multiply realizable than it was before. When fluid dynamics has become a special science, it became possible to understand "There is a fluid in the jar" as "The jar contains atoms arranged thus-and-so" or as "The XYZ field has such-and-such properties localized to this jar" or in some other way. If more fundamental science ended up with a theory on which the universe is all filled with stuff except where there is fluid, so that what we call "fluid" is a hole in the stuff of the universe, that, too, would be compatible with fluid dynamics once it became a special science. But when fluid dynamics was a fundamental science, its claims were not compatible with the hole- and XYZ-field-theories of fluids (I don't know about the atomic case), since on the hole- and XYZ-field-theories, the ontology really doesn't include fluids (there really aren't any holes!). Moreover, it may be that new "ceteris paribus" claims get tacked on.

Mr Parker:

This is a good point: no scientific theory can provide a fully detailed, perfectly accurate description of the world. But a fundamental science attempts to provide a description at a certain level of generality. Thus, while we cannot measure the numbers that enter into the equations with full precision, the equations might, nonetheless, be exactly true.

enigMan said...

Biology talks of species, but they are only relatively distinct categories. Meteorology talks of clouds, and divides winds and such into a few categories, e.g. a force 6 wind. Should we be realists about force 6 winds? There are force 6 winds, but they are not a distinct ontological category, they are subjectively defined and vaguely defined. Truths about force 6 winds are not crisp truths.

Conversely electrons are pretty well-defined. We should be realists about them, and should probably believe they exist. Even strings we don't believe exist we should perhaps be realists about. But what about the entities in geography or economics, or even astronomy? Mountains exist of course, but they are vague subjectively defined things. In short, the special sciences are less likely to be wrong because they are ontologically vaguer.

But isn't realism about not being ontologically vague? I guess this comment is really a question about what you all mean by "realism" here.

enigMan said...

I think that there are scientific facts that are well demonstrated, and then there are scientific speculations that are less well demonstrated, and that it's a continuum. So I would not draw the line at fundamental physics, but at what we are most familiar with.

The problems that we find in funamental physics we also find at any cutting-edge, e.g. astronomy, psychology and economics. And subjects that are not even that scientific yet (e.g. sociology, ontology) may be in an even worse position than fundamental physics. After all, there are parts of fundamental physics that we should be realists about and believe in (e.g. QED). It is the speculative physics that is more douhtful (e.g. QID).

And in mathematics, there are doubts about anything following from doubtful foundations. So if there are doubts about fundamental mathematics (e.g. set theory) then they spread throughout the rest of mathematics.

And of course, we don't know where the next cutting-edge will be. We may be justifiably confident that a science is finished, and then discover that it is not, and in such a way that the objects of the science all change, for all that our confidence was justified. (That isn't meta-induction from physics, it's a counter-example from physics.)

Alexander R Pruss said...

Maybe the way to put it is this: fundamental science seems to be always cutting edge. On this reading, fundamental things cease to be cutting edge at about the same time as they are transformed into special science.

But I don't think fundamental science is always cutting edge. General relativity isn't. But I do think we shouldn't think that general relativity is true. So it's not just about the cutting edge.

As for the question of what I mean by realism, I mean here a weak realism (something opposed to the strong anti-realism in McMullin's essay). In this sense, you're a realist about my claim that there are holes on Smythe Street if you, in the ordinary unreflective way, would endorse my claim that there are holes on Smythe Street. I don't require for this realism that your ontology should include holes (I think nobody's ontology should include holes).

I agree that the flexibility is what makes the special sciences do better. A point I've heard made is that while the special science are less precise and more certain. I don't think we want to endorse an across-the-board inverse relationship between precision and certainty. Mathematics is much more precise than fundamental physics, but also much more certain; psychology is less precise than molecular biology, but not more certain.

Chris said...

Alex, I realize that this is an old post, but I stumbled across it this morning and thought it was very good.

I'm wondering whether our levels of justified certainty might vary similarly in theological matters. That atonement results from the crucifiction is more certain than any particular theory of the atonement.

Philosophical Figments

Alexander R Pruss said...

That's an interesting and plausible application.

Glow88H said...

I think it is marvelous that even though science is imperfect, we can build progress on models that utilize only a small portion of what can be known about any given thing in the universe.

Reality is the elephant, and we are the blind men describing it. It is not wrong to say the elephant's tail is like a rope, or that its ears are like huge leaves. It is simply incomplete. If you need a rope that will bridge a river, an elephant's tail will never do. But if you need a rope-like movable cover for the elephant's delicate rear, its tail does the job perfectly.

Here's my brief defense of why we should support usable science even though it might be superseded:

"Science Is An Imperfect Science"

Gloria Merle Huffman