Statistics

Why Falsifiability, Though Flawed, Is Alluring: Part II Climate Model Example

Al Gore, inventor

Three-card Monte Pseudo FalsificationRead Part I

Take the various theories of the doom which await us when global warming finally strikes. Like Sharknadoes. Or, better, take the theory of global warming itself; just that theory which says what will happen to the climate, and not the stuff in the climate (like us and sharks). Fifteen years ago this theory existed as a complex set of propositions. The specification of these is not of direct interest to us, but for ease call them GW1998, a theory. A prediction derived from this theory was that it was highly probable the world would have by 2013 burnt to a crisp, or whatever.

We did not roast, but GW1998 did not say we certainly would, therefore the theory is not falsified. Falsified, as we must ever remind ourselves, means to prove with 100% absolute certainty that something is false. The predictions, while awful, did not attain sufficient decrepitude to warrant falsification. Holders of GW1998 could continue to claim that it (the theory) might be true without running into logical difficulties. But you would hope, merely in the spirit of decorum, statements of support would now be couched with humility. That is, while GW1998 might still be true, the confidence of its creators should have been badly shaken. It wasn’t.

A month ago the IPCC, creators, or at least governors, of climate theories, announced that their confidence in global warming as a theory had increased and not decreased. To the world, this appeared as if somebody had discovered the keys to the liquor cabinet while the parents were away. But as clueless as the IPCC was, there were accusations from skeptics that GW1998 was falsified and should be abandoned. We now know the theory wasn’t falsified, but what civilians don’t realize is that it was abandoned.

Nobody uses GW1998 any more. It has been replaced by a succession of theories, most recently (let’s call it) GW2013. It is that theory which the IPCC cherishes, not GW1998. How they came to love it is their own business, but it was not and could not have been because they have verified skillful predictions made conditioned on the new theory (this is one way to ascertain the value of any empirical theory). There hasn’t been time to accumulate these predictions. They love it intrinsically, because of what it is and represents.

The switch to the new model was made behind the curtain, but it is rational to suspect that since most of the same people who created GW1998 also had a hand in GW2013, and that because GW1998 didn’t pan out so well, that GW2013 will be no great shakes. But that is neither here nor there for the principle of falsifiability. We can’t falsify a theory in advance unless we can show that its propositions necessarily lead to contradictions. This has not been done with GW2013.

How different GW1998 is from GW2013 is also not the business of falsifiability. But there is the sense, and there is some evidence in the form of the IPCC’s new document, that the differences are small and insubstantial. That is, the fundamental set of propositions that mankind is nefarious in his dealings with the atmosphere built into GW1998 remain current in GW2013. If this is so, then given the past performance of GW1998 it would seem “scientific” to keep quiet about GW2013 and discover something else to study with government funds until we reach some GW20XX where skillful predictions can be made. Don’t, as they say, hold your breath.

The trick that has been played is what accounts for both sides sputtering in wonderment that the other side “Just doesn’t get it.” Don’t get caught up in this. Learn instead to recognized when this is happening. Point out to your interlocutor that he is discussing theories and not a theory, and that the new theory is not yet proven.

Now, even if GW1998 were falsified—and it was not: it may have been proved of little value to civilians, and of great value to climatologists, but it was not falsified—I say, even if it were falsified, the trickster will claim that it was only a small part of the theory that was at fault, and that this was, or very soon will be, expunged. If this is so, the modified theory is a new theory; it is comprised of a different set of propositions. It must be judged anew.

The more complex the theory, the easier it is to make this maneuver; the easier it is to assume that if pieces (propositions) go missing or undergo surgery, the theory is still “the theory.” It usually happens when there is a beloved meta-theory beneath the formal theory. In this example, GW1998 is the formal theory and the meta-theory is that mankind is up to no good. But often the meta-theory is true, or likely true, as many examples in the intellectual history of physics will show.

It is the meta-theory that defenders of the formal theory often have in their minds when the speak of the formal theory, not recognizing the distinction, the formal theory almost a distraction, something to be gotten out of the way, like income tax forms.

It is hard labor to get people to focus on just what is and what isn’t being claimed and what observational evidence means to the formal or meta-theory, to tease apart the differences. And to get people to understand just why even poor theories are not often falsified.

Part III: The other examples.


Categories: Statistics

18 replies »

  1. Yes, all very fine although somewhat banal, but the real question is “How often does someone have to be wrong before you stop believing him?” or in Latin “falsus in uno, falsus in omnibus” or as Willis says “falsified for all practical purposes”. I don’t dispute your logic, Briggs, I just don’t see a debate here.

  2. Off topic I know…

    Fascinating pic of Gore, oozing sincerity and warmth, he’d have made a great snake oil salesman. Oh, wait a minute…

  3. Hi Matt
    Thanks, interesting perspectives. A complex theory that can’t be falsified, cannot be proven either, I suppose? And to be a proponent or opponent of such a theory is a matter of belief, I suppose?

    Cheers
    Marcel

  4. Marcel,

    Exactly so. If it can’t be falsified, it can’t be proven as definitely true.

    Probability is always a matter of belief conditional on evidence; change the evidence, change the belief.

  5. Briggs…

    The issue in science is that there is nothing that can be proven true, absolutely. The only thing you can do is whittle away at what isn’t. The uncomfortable part of falsifiability is that you can never stop testing. The most you can do is refine your falsifiable proposition more to more precisely find the edge of what is by removing what is not. Our successes in creating falsifiable propositions is in constantly redefining the proposition to get positive failures or our proposition.

    Your discourse above resonates with me because it is something I keep trying to inadequately express. We aren’t pointing at something, we are pointing at the not something around the something.

    Whenever I start talking like that people say “Whatever” and pick up their beer and head over to the other friends house with a game on.

  6. Well dang,put wide enough error bounds on your Representation Concentration Pathways and any outcome from Baked Alaska to Ice bound Acapulco are “In Broad Conformance with the theory”. It’s just that the theory isn’t actually useful for anything…. Begs the question as to whether anything that sloppy can actually be called “Science”.

    When people talk about “Falsifiability” I think the assumed context is that a theory should be drawn tightly enough to make testable predictions about measurable phenomena. Having to redraw the theory of the atom, because emission spectra frequency didn’t rise with temperature in the expected manner. Einstein predicting that scars could be seen when they should be obscured by the disc of Mercury. That sort of thing.

  7. As an engineer rather than a scientist, this definition of “falsify” seems too strict. As a practical matter, theories can be falsified – which is to say, shown with a reasonable degree of certainty to be wrong. As an extreme example, gravitational theories are pretty strongly “falsified” by orbital behavior later, pretty apparently, explained by relativity.

    And unless we are doing theoretical mathematics, “a practical matter” is good enough.

  8. So, “For me, when a theory’s predictions are wrong 99% of the time it is falsified for all practical purposes.” Isn’t that what we work with most of the time? Though I doubt most people recognise that they’re making a choice.

  9. John Moore, Rich, All,

    This controversy is my fault. I didn’t explain things well enough.

    Moore says “unless we are doing theoretical mathematics…” Of course, in this case, that is exactly what we’re doing. We are making precise distinctions. Now (also of course) this may seem beside the point in the present example, but it is not always. It is like I told Willis in the comments to Part I: the probability of a thing (given some evidence) is not equivalent to the decision one makes conditional on this probability. The two aren’t even in the same book.

    The decisions you make as an engineer are different than I as a logician will make, and these are both different than the decisions a politician or climatologist makes, all on the same probability. The number and range of decisions approach infinity. Therefore we have to nail the probability (and the evidence from which we deduce this probability) as exactly as we can. A probability that is absurdly low to us may be uncomfortably high for somebody else.

    For example, I cannot call something “false” which is not false, especially if my decision/goal is to explain the nature of probability. But if I’m going to award grants, I’d cut back on the money flowing to Penn State. Etc.

  10. Sorry Michael but I can’t leave this: It is stars seen from behind the Sun, not Mercury, due to the gravitational bending of light. This is a very small effect for the Sun (it would be unmeasurable for Mercury) but large for some galaxies at a great distance and is also called gravitational lensing.

    John, when does the impractical become practical? Most of the time general relativity can be ignored but it has a tendency to pop up where least expected. For example, satellite GPS would be totally useless without general relativistic corrections and the incredible precision of atomic clocks. Most people would think that knowing the time to a second over a time period measured in millions of years is totally impractical but take hand held GPS devices for granted.

    Briggs, I think that you explained it well enough – too well in fact. There would be less debate if readers took more time to absorb the content of your articles before commenting, but it would also be less fun. By the way I have been highly entertained by your responses lately, but then again I’m not a normal person.

  11. scotian, you are correct, my apologies. The lesson is don’t try and remember your general relativity late at night, after an 18 hour day, while using an i-Pad on a train…..

  12. @Scotian – I cannot give you a precise definition of when impractical becomes practical. Fortunately for technological advance, engineers don’t have to answer that question in a precise, mathematical or philosophical matter. Neither do politicians. Climatologists shouldn’t even be dealing with the issue – in their role as scientists, they should trying to advance towards truth, not advocating policy. In the actual social workings of science, the question of when something is “falsified” is also fuzzy, as it should be.

    I’m not sure what your GPS example is reaching for.

    I think most or all of the confusion here is that @Briggs is being philosophically correct, while discussing an issue that normally is not, and cannot *practically* be that precise. I don’t take issue with this – @Briggs is worlds ahead of me in his understanding and exposition of philosophical (or pure mathematical) issues of probability and, in this case, philosophy.

  13. John, you state “I’m not sure what your GPS example is reaching for”. I could go all Dirac and point out that this is not a question but instead I will observe that I am not Briggs and rarely admit to lack of clarity. I just ask people to read more carefully. 🙂

  14. @Scotian – it isn’t a question, but it is a request for explanation. Pardon this poor engineer, but to me, the concept of “practicality” is, practically 🙂 pretty simple.

  15. Briggs — Of course, Darwin’s Evolution is another example of a theory that keeps getting modified into newer and newer theories while its chief proponents claim that THE Theory of Evolution has been proven repeatedly and must not be challenged in any way upon pain of expulsion from the Ranks of the Right Thinking.

  16. There are two points to consider: 1)the purpose for which hypotheses are proposed, and 2)how to assess the probability of an unfalsifiable proposition. I would assert that the primary utility of hypotheses is for prediction. If one can improve on the toss of a coin, than the hypothesis has utility; the better the performance, the greater the utility. For the case of the null-hypothesis that the “Abominable Snowman does not exist,” it is theoretically possible to falsify it by capturing a creature. However, in the absence of a specimen, I would suggest that after a long-period of honest, concerted effort, a ‘prudent man’ would conclude there is increasing probability the proposition is true.

Leave a Reply

Your email address will not be published. Required fields are marked *