The Krugman Kahan Klein debate

Brad DeLong sent me to this wonderful post by Ezra Klein. Read it.

Psychologist Dan Kahan studies motivated reasoning, confirmation bias or, to use the technical term derp. He has found similar derpitude in liberal and conservative experimental subjects. Paul Krugman is puzzled as he thinks that Conservatives in the wild are way derpier than liberals in the wild. Kahan notes that this is exactly the perception that he would predict Krugman would have and that in response to Krugman's challenge to find examples of collective liberal derp similar to global warming denial he argues that the facts have a clear liberal bias, that the pattern which he agrees is in the data and which doesn't fit his hypothesis is due to coincidence.

I comment

Psychologist cure thyself. In the Krugman, Kahan, Klein debate (note to self not a good case for an acronym) Kahan essentially bases his case on the assertion that the facts have a clear liberal bias (OK the Krugman Kahan Colbert Klein debate KKCK).

In response to Krugman's question he (as perhaps unfairly paraphrased by Klein) said (note that the really embarrassing claims are not direct quotations lf Kahan)

"Kahan solves the problem by arguing that being right is irrelevant. "It's not whether one gets the answer right or wrong but how one reasons that counts," he argues. A liberal who works backwards from conclusions but happens to believe in climate change is "to be congratulated for being lucky that a position they unreasoningly subscribe to happens to be true," but nothing more."

As Kahan said of Krugman this is perfect (for all I know perfected by unfair paraphrasing). There is nothing more characteristic of derp (the technical term for the phenomenon Kahan studies) than claiming that a pattern which doesn't fit the theory is due to pure coincidence.

Kahan also manages to (allow himself to be perhaps unfailrly presented as) the perfect strawman reductionist. The straw reductionist says that if you understand the parts you understand the whole, that the whole is the sum of the parts where the key bit is that summation is a liniear function. This is an unfair parody of true reductionists based on the assumed general linearity property of functions (also used by economists who assert that if assumptions are approximately true then their conclusions are approximately true).

But sometimes it is fair. Kahan as presented by Klein seems to think he can determine the behavior of groups of people (not predict with some confidence but know with enough certainty that he can dismiss data on the observed behavior of the groups) by studying individuals alone interacting with a computer (or piece of paper or experimenter reading a script and following a protocol -- I don't know exactly how the experiments were performed). As Klein notes, this assumption is absurd and should not be taken seriously even if it didn't lead Kahan to complain about the liberal bias of the facts or argue that a pattern (which Klein didn't quote him denying exists) is due to coincidence.

Now one methodological rule is don't assume all fields of inquiry other than your own are totally sterile wastes of time. If you are a psychologist, don't assume that sociologists, political scientists and historians have nothing to teach you. Another is Paul Krugman is usually right and if you debate with him you will probably end up looking like a fool. A third may be that the wonkospherical bubble is as impregnable as the conservabubble and that Krugman, Chait, Klein and DeLong with instinctively unite against you if you challenge any one of them.

I am interested in a further critique of Kahan. Something within pyschology as understood by myself -- a dilettante economist. I think I will go over to my blog as, stick toeven if you don't put a limit on comment length, you should.

OK here I am at my blog.

I want to stick to psychological experiments in which individuals interact with machines and such. It just isn't true that the behavior of liberals and conservatives is the same. The degree of derpitude seems to be similar, but people who self identify as Conservatives are more conservative in the ordinary sense of the word -- http://www.scientificamerican.com/article/calling-truce-political-wars/

I think there is also independent evidence that US conservatives are somewhat more likely to be authoritarians -- more likely to say it is our duty to loyally follow are leaders and less likely to say we should question authority.

I didn't manage to google up the link, but there was an experiment in which people were offered capsules with numbers printed on them. If accepted they got a prize (which could be negative so accepting a capsule was risky). Same number on outside same reward or punishment for accepting. Political liberals tried more different capsules experimenting and taking losses to learn Political conservatives more nearly stuck to the tried and true.

Derps include contrarian derps who are convinced that everything you say must be wrong. They aren't pleasant but they are socially useful. Kahan has no data on contrarian-ness vs conformism.

Here the point is that there is evidence from experiments with one person at a time which suggest ways in which liberals and conservatives beliefs evolve over time (neither rationally updated given the data but less table for novelty loving liberals than for conservative conservatives). If one never mutates one never evolves. If one mutates and there is a very weak selection of beliefs which fit the facts (too weak to be noticible in one trial as in Kahan's study) the mutators's beliefs will evolve to fit the facts. We will also have more silly crazy notions which last a long while but not forever.

If people think that dissent is disloyalty, then they will be even less open to evidence when interacting with each other as a group than when isolated. If they think debate is generally good, they will ... well be not much more irrational when together than when alone.

OK so something against liberalism and pro-psychological reductionism. I have just (reluctantly) conceded that the core liberal belief that discussion and debate lead us to find the truth together is false. Consider two bunches of people and a forecast to be made. One approach is to ask people and then average their answers (so the belief based on the whole is, by construction, the sum of the beliefs of the parts oh then divided by the number of parts). Another is to bring people together to share their knowledge, critique each other's reasoning, discuss and reason together. I think it is demonstrated that the averaging approach gives more accurate beliefs about verifiable facts (better forecasts better guesses of I don't know the surface area of lake Superior whatever). I cite Robert Shiller wrote this somewhere.

So I think liberals reason together less badly than conservatives. As evidence, I claim that I have been convinced by other liberals that the core tenet of the liberal faith is false