How do you stop fraud? Not by adding more layers of checking.
A couple of days ago, I came across this Independent article on twitter (tweeted in a few times, not sure if it was Dorothy Bishop’s tweet where I first saw it). It talks about some of the recent scandals in the UK on data-faking (pleased it is not social psychology this time. Horrified it was actually about medication). A suggested fix for this is, wait for it, more layers of checks. Kind of like the IRB, clogging up the research stream.
Dorothy Bishop had an excellent response in her blog (with ensuing discussion). She identifies, like most of us are doing over and over and over again, the misplaced incentives, some of the publication practices, and also the training of scientists.
I agree that the root of the problem is the incentive structure inherent in most of science today. However, obviously fixing this is a far bigger and more challenging task than many of the relatively minor patches that people have been suggesting a lot these days. Although I sincerely believe that the incentives must change, unfortunately I have yet to come across a concrete proposal for how to accomplish this that seemed feasible. Do you know of any good ideas for fixing the Big problem?