In a comment on my post about drafting Krugman, Don Geddis wrote some remarks I'd like to respond to here in detail because what he says gives me the opportunity to make several points I've been meaning to make.

On the Use of Models

... you seem to be looking for mathematical proof. —Don Geddis

I'm not looking for mathematics to prove the process is right; I'm looking for proof that someone is using mathematics to search for what's right.† I want to see competent mathematics and computer science to describe solutions, to model them, and to compare them. No math or science can prove these systems right, but the use of math and science in the approach would allow repeatable experiments that can be discussed because even though the live system cannot be frozen or backed up, a model can be.

If there is something wrong with the model, such as a missing parameter, the model should be refined not until it's correct (which is an almost certainly unachievable goal in practice and probably also in philosophy) but until it represents, in our best judgment, what we know has an effect. It may be that a single mathematical model with a lot of parameters works; it may be that multiple models are required because it's not the parameter value but the actual functionality that people debate. But we should know these things. And I don't have the sense we do.

It should be possible, for example, for there to be a program that can be run that shows some degree of modeling of a recovery and allows competent people (and possibly even those who aren't very competent, if they wrote the UI right) to twist knobs and know how fragile or robust that path to recovery is.

And I daresay we have a bunch of unemployed computer programmers, mathematicians, and scientists just now who have a lot of interest in this problem and could be usefully employed in looking at it. I don't think we're in an economy where having a college education is proof you're not on the street looking for work. If we're going to literally print money in the trillions let's send a tiny bit of it to these people and ask them to help us.

On the Testability of Models

I think the unfortunate situation is that all the macroeconomic models are incomplete. —Don Geddis

This is a would-be criticism of science, too, if it weren't the very definition of what makes science good. Darwin's Theory is often criticized by radical religious factions as “only a theory” (usually in contrast with the word of God, which is frequently offered, almost in the same breath, as simple truth). It's surprising how many people think science is somehow weakened by building itself on theories.

(I'm not suggesting Don has this this misconception—I know him too well to think that—but there must be people swayed by this bit of spin doctoring or I don't think we'd hear it over and over as much as we do. Success seems to breed repetition.)

The point of theories in science is that they either stand or fall on the strength of what they say. Our reliance on them doesn't come from the act of utterance, but from the testable claims they make, and their willingness to stand those tests. A theory that offers no test is no theory at all, just dogma to be accepted or not blindly. So the strength of a theory comes when people take the tests and try it themselves and achieve comfort in its ability to be reproduced or to predict other behavior.

On the Refinement of Models

... our current financial situation wasn't predicted by any mainstream model. —Don Geddis

Excellent. This is how science moves forward. It means we must adjust our model. Hopefully there is a real actual modeling system (you know, like the WOPR in the movie Wargames, except without the dial-out modem). And hopefully it runs scenarios and tests that these things play out as we want them to. But I don't hear a lot about that. Then one can test known (past) situations to see if the model is right and can tune the model for those before relying on it for unknown situations.

This kind of process, called variously “iterative refinement” or “expert system methodology,” is what cranks out better solutions. The methodology is pretty straightforward: It suggests that if you ask experts to explain themselves, they won't, but that if you instead assert you have a better solution than they could come up with, they will tell you why your system is wrong. You then assert the new system is better than theirs and they tell you another thing that is wrong. After enough of this, they start to run out of objections and your system converges on expert behavior in a way that is codified as a programming system.

In Climate Change discussion, I hear a lot of rhetoric about people using such models, which is very encouraging. I see few actual descriptions of such models, but that may be because I read mostly mainstream sources and not many direct research papers. Still, I think it's a good trend. I wish I saw more of it in the discussion of bailout strategies. It would be good to know which parameters are well-agreed upon, which are uncertain, as well as which are uncertain due to abstract structural reasons vs the whim of political control or public perceptions. The fact that these quantities are not individually available for discussion is troubling to me.

On the Auditing of Models

Let's say that everybody agrees that some of the banks are insolvent, and that nationalization (for those banks) is the only fix. —Don Geddis

But do they believe this out of some handwaving groupthink or because some articulated and inspectable theory leads them there? How would I know?

On the first point, whether something is insolvent, that may be definitional. A definition is a kind of model that is chosen in a social or political way as the only valid model by attaching a social or political constraint that cannot be satisfied except by the model. But if there is no such social or political constraint, then perhaps there are other ways to view that situation. I suspect they will all point to the same direction, that many of these institutions are insolvent and must be repaired or eliminated.

On the second point, about whether nationalization is the only fix, I would like to see the other fixes lined up and run and their parameters explained. I am sick to death of hearing political descriptions instead of mathematical and scientific descriptions of these problems. If applying too much helium pressure to a baloon will make it pop and too little will keep it from rising into the air, the way to explain these effects is not to talk in terms of liberal or conservative uses of helium, it's to talk about the parameters of the task. There may be intermediate solutions or there may not. Have we graphed that?

Why the Pot Issue Really Matters

I'd like to allude also to the whole marijuana question. I'm not a pot smoker. Never had any interest even in just trying it. I have no agenda to encourage anyone to use it. However, I think marijuana should be decriminalized. Not based on some ideology but because I believe those who talk about the market dynamics that would result when they say that, whatever you think about the morality, federal tax revenue would go up, dollars spent needlessly on tracking so-called criminals and incarcerating them if convicted would go down, money that goes to people who trade in more serious drugs would be reduced, and people wanting access to the drug would not have to get it through someone shady offering a gateway strategy to other drugs. Those are all testable claims. I have not tested them, but I would more believe Obama's position on this if I thought he had really run the numbers and could tell us how they come out. Instead, I feel he is saying “it's obvious” and this gives me strong distrust in his entire strategy for the bank bailout because I think he thinks it's ok to leap to political conclusions instead of doing the thing I thought he would do, which is to employ experts who really know how to do science.

Trust, but Verify

What should be done right now? The Obama guys are doing the best they can, with the tools they realistically have available. Meanwhile, Krugman is writing (from the outside) that -- absent politics -- a different approach would be more effective. Can't they both be right? Is there necessarily a real conflict here? One is talking about ideals, the other about the practical realities of getting things done in the real world. —Don Geddis

It's theoretically possible that anything is right until proven otherwise. But what settles that is not whether I'm a forgiving kind of guy who understands political compromise, but whether both of these lead to solutions that work.

The question seems to presuppose that there is some way to know something works other than by some form of advance testing. Like maybe you can know that it creates magic bubbles or winged fairies in your heart when you think about it. If they're both right, they will be right for reasons other than that one is “idealistic” and one is “pragmatic.” It will be because you can show demonstrably, in at least some test modeling theory, that the right outcome will happen and then you can talk about whether the conditions of that model are likely to ever happen.

And allowing a word like “politics” to become an impenetrable shield that is allowed to keep us from looking deeper is not ok with me. There's a probably unintended but very definitely present sense of ad hominem attack there that paints the potential usability of Krugman's ideas negatively and of Obama's ideas positively even before they are even compared. I'd like them compared. I don't care whose comes out a winner after a decent comparison; I am not arguing that Krugman should win. I'm saying I like the vocabulary he talks in terms of because it has that testable feel.

The Thoughtful Planning Phase

I'd frankly like to be able to download the Java app they're using to make their projections. I'll be thrilled if you tell me there is such a program but it's something more ambitious, and runnable only on a farm of supercomputers, not as a toy on anyone's desktop. But even if that were true, I'll bet the world-wide web has a larger and more reliable set of processors working on SETI than the government has working on this problem.

As so often happens, I think the real reason this is being played close to the vest is not national security, it's fear of being judged critically. It's the opposite of whatever I expected in terms of transparency.

This may seem like a petty concern, for me as a single citizen to be whining that I don't have access. Perhaps your point is that individual citizens are too small for anyone to care about satisfying. But I don't need to be satisfied, I need to be enabled. And not just me, everyone. Because there needs to be a dialog about testing whether the answers we're getting are good ones before we are committed to them.

We're past the point where we have to do something now or the world will fail. We're now in the thoughtful phase where we pick something we can live with, for all practical purposes, forever. Certainly for many people there will never be another political world than under what Obama will decide. That's pretty heavy to say just “trust us.” We trust him to do the looking around, to lay things out, to show us the alternatives, to explain to us, and to tell us what he plans. But the price of that is that he (and all his people) must stand the scrutiny that comes with having taken on that decision.

It's even ok if they can't make bailout@home into a cool app I can download onto an iPhone and run myself. The scrutiny doesn't have to go to that level. But if I can't inspect it, I have stronger requirements on what I want to know they're doing in private. Because one way or another, I want to audit the process.

To be told these are matters that are none of our concern is not sitting well with me. Yes, we need people in there doing higher-bandwidth discussion, and we cannot all be directly involved. But there needs to be more than just a final journal published saying “this is what we did—live with it” after it's a done deal.

And this is all just a dress rehearsal for the great Climate Change debate we have yet to have. That will have all the same issues and much higher stakes.

If you got value from this post, please "rate" it.

† Sounds a bit like a “Charlie the Tuna” commercial!