Wednesday, December 2, 2009

A response to Tyler Cowen

I'm having trouble posting at Marginal Revolution for some reason, so I'll move my teensy portion of the debate over here. Tyler Cowen made a fairly intelligent argument that there are two possible reactions to "Climategate" and that one of them is being overlooked. The first is to say that the scientists acted unethically, and hence we should disregard their beliefs about climate change or any other subject we care to name. The second is to say that the climate scientist must feel that the issue is really, really important or else s/he would never have acted dishonestly, and so we should revise our estimates of climate change up a little.

A lot of people have given him flack over this.

The very reasonable part of the argument is based on a more refined model of humanity. Are people either all good or all bad, or do each of us dwell somewhere in the middle, performing good actions under certain circumstances and bad actions under others? If you assume that people are good or evil, then one evil action gives you a lot of information about the person and you ought to discard anything else they tell you. If, one the other hand, people are basically good (will perform good actions unless there are strong incentives to do otherwise), then their negative action gives you information about how important they believe the subject is. I think he's largely defending himself by saying this was his real point.

What this does not do is tell you how correct their beliefs are, and this is where I think Tyler's argument falls apart. How is it different from ideologues on both sides of the political spectrum committing the same action? Suppose that a right-wing economist futzed with some data to show that the multiplier on government spending was less than 1 because he really strongly believes it is. Should the left revise their estimates downward? Suppose that at the same time, a left-wing economist futzed with some other data to show that the multiplier is larger than it is because he in turn really strongly believes it is. Should the right revise their estimate upwards? If both types of people futz with data, what is our poor Bayesian to do?

Answer: conclude that the issue must be important to those people. Who knows? Maybe investing a little time in reading both sides of the argument rather than just the sound bites would be time well spent. In this particular case, it suggests that the notion of "scientific consensus" is not as universal as most of us have been told and that people would be well served by actively seeking out the writings of someone with the other opinion. ... Preferably someone who hasn't futzed with their data. I think the wrong conclusion though is to throw up our hands, say they're all a bunch of crooks, and therefore we can safely ignore the issue. And that brings us back to the first point.

No comments:

Post a Comment