Saturday, January 8, 2011

Limits to Craziness

Should some academic papers not be published just because they seem too crazy? There is an interesting debate in the New York Times about whether a psychology paper claiming experimental evidence for people being able to predict supposedly random future events should have been published or not. I am in favor of papers that use solid methods being published, however, crazy the hypothesis that they investigate. The more extreme the idea though perhaps the smaller the claims they should make about the results. So I think the real question about this paper is whether the methods are sound or not. And a lot of the debate is around that.

This guy though thinks that papers that try to test too crazy ideas shouldn't be published. Then he gives an example of an imaginary study that would test whether more accidents happen on the 13th of each month. Clearly, he doesn't think like an economist (and I'm shocked to see that he is a cognitive scientist). If people think a day is unlucky that might change their behavior resulting in more accidents. Or less. But it is certainly plausible. So that is not a good example.

Another question is whether journals should only publish papers that they think are "interesting". Top journals in economics only publish papers that they think are interesting in addition to being correct. By contrast PLoS ONE claims to publish anything with sound methodology regardless of "interest". Of course, as an open access online only journal they have a financial incentive to publish more papers. I am surprised that if this is really the case that they have a pretty decent citation impact factor. The problem with only publishing interesting papers is that it can lead to bias in the scientific literature.

No comments:

Post a Comment