Friday, September 14, 2012

Useful Reading About Social Science And Humanities Research

For those of you who are interested in research ethics and similar questions, this Guardian article is worth a read.  I also recommend the 2010 piece on the Decline Effect.  My intention is not to cast a negative light on researchers, most of whom are honest in their work.

But it's important to reduce the publish-or-perish imperative, especially in the form of (only)  positive findings, and it's important for the professions to police themselves more carefully.  Because the reputation effect of bad research, when it is widely believed in and discussed, will taint all research.

Laypeople often cannot tell the difference between good and bad research (and neither seem most popularizers).  It's a bit like the effect tainted medications have on the markets for all pharmaceuticals.  Your firm may not produce tainted medications but the consumers cannot tell that this is so.  The only solution is to regulate the whole market, to ensure high quality.

In the context of research this means publishing the data the results are based on and also placing more emphasis on publishing no-difference or negative findings.  The idea that only results which comply with one's own theory are worth publishing is dangerous.  It's obviously so in the field of medicine where the finding that a new drug does not work is of great value, for instance.

But it's also important in other fields, as I have written before.  My pet example has to do with the science of gender differences.  To have a field focused on finding differences means that not finding them will make a study less publishable.  Thus, we get more popularizations on how men and women might differ than on how they do not differ.  The overall impression  this will leave is of large gender differences, like gulfs which cannot be bridged.  Yet measures of gender differences in various cognitive fields are usually much smaller than similar measures of gender differences in adult heights or shoe sizes.