Frances: We agree on more than we disagree on yet the interesting stuff is the arguments, because disagreements settled through data analysis is where progress comes from.

Ling: Sorry, my message on westerness was a response to a different comment. I am afraid I am unfamiliar with the tradition from which your comment derives and cannot give it a meaningful response. I do not understand your point and guess that it would require a great deal of reading to understand the context from which your comment comes.

Yamamoto-san: Another apology. It took me a while to see how your comments were a response to my argument. One of the main lessons to learn from these kinds of exchanges is the difficulty of communication altogether. One must share a great deal of context in order to be able to communicate effectively.

I said that deductive logic of the sort involved in geometry is not natural to the human brain. You said people are not always irrational.

Of course, the opposite of rational is irrational and irrational means wrong. Acutally both Riker and Tsebelis have made statements indicating that if people were not rational they would be unpredictable. Yet, it is quite possible to get the right answer without deductive logic. Trial and error is not a bad system. People who do not live up to the standards of rationality as defined by rat choicers can still be quite reasonable, can learn, and are predictable. They are less irrational than non-rational.

I know of no one who believes that rat choice assumptions accurately describe human pyschology. If you ask, how often do people think rationally, the answer is virtually never. One can be trained to think rationally, but even those so trained make common human errors. Most of the effort has gone into showing why erroneous assumptions make no (or little) difference in the deductions. I am completely unconvinced. Another tack is to try to define the situations in which rat choice predictions will be accurate, no matter what thought processes go into the production of that behavior. S?? and Ferejohn argue that rat choice works best when choices are highly constrained by the situation, i.e., people make rational choices when they have no choice. Tsebelis argues that rat choice predicts well whenever the costs of behaving irrationality are high. If taken seriously, this sounds very good to me. I think it would mean that most of the complex math would prove irrelevant. My own hypothesis is that rat choice will predict well when the feedback from learning is accurate and the situaion is either permanent or repetitive, i.e., not very often.

There is an interesting parallel with psychologists who study attribution errors. They went through strenuous training in the proper techniques of statistical induction. They then came up with the obvious hypotheis that people's thought processes can be modelled as statistical induction. (Introspection is the easiest source for hypotheses about how other people think.) It turns out that people make systematic errors. They are not good scientists. The errors are, however, highly predictable.