« Lileks on Soap Aroma | Main | Bush's National Guard Service: Info from the Thornburgh Report »

January 19, 2005

Why No Theory is Falsifiable. with Application to Evolution

I was just reading a criticism of Intelligent Design theory on some blog. It was making the typical Popperian argument that the big criterion for whether a theory is good is whether it is falsifiable. This idea has, I think, been long abandoned in philosophy of science. It is common among the less well educated in economics. It seems to be very common among people thinking about intelligent design. I have criticized its use as an attack on Intelligent Design before. (I also have posts on speciation and sunflowers and the usefulness of theology to science.) I am not up on the general criticisms of the falsifiability criterion, but my impression is that the first steps beyond it are the Kuhnian idea of paradigm shift (that even if a theory is falsified, people don't abandon it until a better theory comes along) and Lakatos's notion of science as process (that rather than rejecting a theory after it fails a test, we just fix up the theory, and then abandon it only if it acquires an ugly number of patches).

I just had another idea, though, one that if I'm right is a quite deadly attack on the falsifiability criterion. I'll start with the application to Intelligent Design, and then do the general case.

Suppose we have two theories of Evolution. Both are theories of Evolution, because in both, new species arise by mutation and spread by natural selection, the ones thriving that are best fitted to the environment at the time. One theory is Standard Evolution: that mutations are all the result of things such as cosmic rays and random mistakes in genetic copying. Another theory is Intelligent Design: that some (though not most) mutations are the result of some kind of intelligent intervention, by God, Martians, or whatever. Thus, one theory deals with the difficult problem of improbably complicated mutations by saying that improbable things do happen sometimes by chance (or is the result of something we haven't discovered yet), the other by saying that it isn't random after all, but designed.

Suppose someone says, "Standard Evolution is good because it is a falsifiable theory and Intelligent Design is not. Standard Evolution would be falsified if we discovered human footprints from the Jurassic Age." That is wrong. The footprints falsify Evolution, but they falsify Intelligent Design just as much as Standard Evolution. Where the two theories differ is not in saying that Evolution occurs, but in saying how it occurs. Thus, a falsifiability test of Intelligent Design or Standard Evolution per se must rely on the particular theory's special features.

Each theory could be falsified, but only improbably. Standard Evolution could be falsified if evidence came up that God guided mutations-- say, the Second Coming, with miracles and a divine explanation. Intelligent Design could be falsified if evidence came up showing that all the improbable mutations are not improbable after all, but have a simple explanation we just haven't discovered yet, so there is nothing left to be designed. (Of course, a weak form of Intelligent Design, not falsifiable, is that God is guiding even the ordinary mutations that easily explained by natural causes.) Or, Intelligent Design could be falsified by showing that no guiding intelligence exists-- a negative that is very hard to prove.

Let's focus on a particular form of Intelligent Design: that the Biblical God, Jehovah, is the intelligent designer. The problem then is that to falsify Standard Evolution we need to show that God exists, and to falsify Intelligent Design we need to show that God does not exist. Both are tasks, which, even if the premise in each case is true, are of the same order of difficulty. If it is hard to falsify Intelligent Design, it is also hard to falsify Standard Evolution, which is "Not-Intelligent-Design".

Now let us move to the general case.

Consider the theory (akin to invoking God, as X), THEORY 1. "If X is true, then Y will happen."

but suppose we cannot tell if X is true or not (though we can tell if Y happens or not). Theory 1 is nonfalsifiable.

A second theory says, THEORY 2. "If Z is true, then Y will happen."

and we can tell if Z and Y are true or not. Theory 2, you may say, is falsifiable. If we find a case where Z is true but Y did not happen, we have falsified the theory.

Theory 2, however, can also be written as THEORY 3. "If Z is true, then Y will happen, even if X is false."

Theory 3 and Theory 2 are the same aren't they? Theory 3 just makes explicit a detail which is implicit in the statement of Theory 2.

But Theory 3 is not falsifiable. To falsify it would require us to find a case where Z is true, X is false, and Y happens. But we cannot tell if X is true or not.

Thus, I conclude that Theory 2 is not really falsifiable. Since for any theory I can come up with a competing theory which, like Theory 1, is nonfalsifiable, and that competing theory's falsity is implicit in the original theory's statement this means no theory is falsifiable. Comments are very much welcome. Is this idea well known? Is there a flaw in my reasoning? Even if true, does not it not matter to the usefulness of the idea of falsifiability?

I will add as a postcript that falsifiability is actually a very useful idea. It's just that its usefulness is not so much in deciding whether to accept a theory or not as to figure out if it is coherently stated. If I ask myself,"What would falsify this theory?", I am forced to think about what the theory actually means, and frequently I find I don't know what I'm talking about-- that I hadn't really stated a theory to begin with.

Posted by erasmuse at January 19, 2005 09:42 PM

Trackback Pings

TrackBack URL for this entry: http://www.rasmusen.org/mt-new/mt-tb.cgi/392


Post a comment

Remember Me?

(you may use HTML tags for style)