I don’t know much about global climate science. I’ve read a few papers and listened to a few lectures, but it’s not something I’ve ever studied in enough depth to have any business commenting on.

I do know a bit about computational modeling, though, and I also have some experience with argumentative overreaching (having spent 20 years as a litigator).

It’s fairly easy to fool oneself with computational models. Models have parameters that have to be “tuned” to fit the data, and given enough parameters (it doesn’t take very many) you can make a model fit just about anything.

In fact, I would go so far as to offer the following conjecture:

Conjecture: For any model of a complex system where a set of parameters determines the fit of the model to training data and the model generates a binary prediction A, there exists an alternative set of equivalently plausible parameter values that produces no worse fit to the training data and generates the opposite prediction (i.e. ‘not A’).

We could call it the “model – antimodel conjecture” — basically, every model has an antimodel.

I haven’t tried to prove it (I also haven’t looked to see if anyone else has, but I’d be surprised if someone hasn’t explored this territory). I suspect that there may be cases in simple models with few parameters where it would not hold strictly true. But I would bet that it is provably true within fairly undemanding bounds, which would include most or all of the models that are complex enough to be interesting as representations of real systems with a significant number of degrees of freedom. And it’s obviously trivial to prove for all models if you allow adding a single additional parameter, which is exactly what modelers do when we can’t get the model to fit the data with the parameters that we already have.

Of course, the “plausible” part is important. My antimodel conjecture doesn’t apply to a model where all the parameters are narrowly constrained by clearly established physical principles — a billiard ball colliding with another billiard ball, for example. But, pretty clearly, that isn’t the case with any of the climate models. And one has to be careful, because some models can be arbitrarily sensitive even to tiny changes in a very small parameter set. (In theory, you’re supposed to check for that, but I’ve seen plenty of published modeling studies where it wasn’t done.)

As for argumentative overreaching, here’s another, slightly less mathematical conjecture:

Conjecture: Given an agenda and sufficient motivation to advance it, a computational model can be found that is plausible and predicts the desired outcome.

That one, I think, is easy to prove: all you need is a form of model that can fit a wide range of data (easy: just keep adding parameters), generate a large number of instances, and choose the one that gives you the results that you want. No intentional dishonesty is required — all you’re doing is discarding permutations that didn’t work.

In general, as an aside, it seems to me that it is a misuse of a computational model to cite its output as evidence supporting a prediction of an inherently complex behavior that can’t be reproduced and measured under rigorously determinable conditions. The proper use of a model is to try to gain understanding of such things, not to predict them.

Anyway here’s my take on climate change:

  1. For all of the foregoing reasons, I regard the computational models as neither evidence for nor against the anthropogenic climate change hypothesis.
  2. The suggestion that we must believe climate predictions because no reputable scientists believe otherwise is unworthy of serious credence. Science is not a democracy, and the claim of unanimous agreement is in any case clearly false. More than 30,000 scientists, 9,000 with PhD’s, signed the Oregon Petition endorsing the statement “There is no convincing scientific evidence that human release of . . . greenhouse gases is causing . . . disruption of the Earth’s climate.” One may dispute the credentials and motives of the 30,000 signers (there’s a word for that in the argument biz, it’s called “ad hominem”), but clearly there are a significant number of dissenters, and you can’t dismiss them all unless you’re willing to redefine “reputable scientist” as “scientist who believes what I think he should believe”.
  3. That doesn’t mean that it’s wise to subject the environment to massive perturbations whose effects we don’t adequately understand. As Nassim Taleb has persuasively argued in a somewhat different context, when dealing with potentially existential risks, one ought to err considerably on the side of caution. The precautionary principle makes good sense, as long as it isn’t misused. I, personally, would be all in favor of reducing carbon emissions, and a lot of other kinds of chemical pollution, if there were a practical way to do it that was not a transparent political con job.
  4. However, the precautionary principle is meaningless unless you specify what precautionary measure you propose to take. “Eliminate excess CO2 emissions” is not a valid proposal because it isn’t remotely possible to accomplish. In fact, there is zero chance that any of the currently proposed “precautions” would reduce greenhouse emissions by any amount that could have any conceivable effect. When you’re dealing with existential risks, you either solve the problem or you don’t — gestures, halfway measures, and ‘steps in the right direction’ are, if anything, counterproductive because they produce an illusion of improvement when the risk is in fact unchanged. And needless to say, forcibly transferring large sums of money from first world taxpayers to third world kleptocrats is not a credible “precaution”.
  5. It’s also necessary to consider costs, vulgar though that may seem. Whenever you hear someone argue that you “can’t put a price” on whatever it is they’re trying to sell you, you know you’re dealing with a charlatan. I don’t know exactly how costs should be taken into account in the climate change equation, but they aren’t irrelevant. If the choice is between subjecting your children to a life of poverty today vs. run an indeterminate risk that the planet might be three degrees hotter 100 years from now, I know what I would choose — and for a not insignificant number of people on the planet, some of the proposed carbon reduction schemes do present something approximating that kind of choice.

So what’s the bottom line?

This is a technological problem. We wouldn’t have a carbon emission problem if not for a century or two of technological advances that enable large populations to live at a level of comfort and possibilities far exceeding anything previously dreamed of.

If there’s one thing I’m very, very, sure of, it’s that you can’t solve technological problems with politics. But you can solve them with technology. And that’s why I don’t lose too much sleep over carbon emissions. I don’t know what the technological solution will be, but as a technology optimist, I’m confident that there are some, and we’ll find them. In fact, if you left it up to engineers instead of politicians, we could probably solve the problem fairly easily.

The sky might be falling, but we’ll figure out a way to prop it up — we always do.

Leave a Reply

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