Bad Incentives Make for Bad Science

A Harvard research group recently got caught committing a form of fraud. In their cancer research papers, they manipulated images in such a way as to make the images better support the conclusions of the paper. While Harvard itself says the investigation is ongoing, there are 37 affected papers. Seems a lot for honest mistakes. Now, I could make a joke about how the replication issue is supposed to be a soft sciences thing, but I think this just highlights how poorly we do science in this country. The incentives are all wrong.

We pay researchers for publishing, and there is a bias towards publishing positive results. That is, if you want to advance your career, get tenure, have a fancy company, etc. you need to prove something IS not that something IS NOT. And that is not how science works. There is a great deal of luck in finding out that something is. Much, if not most, research ends up proving that something is not. And that should be fine — proving that something doesn’t work the way we hypothesized has a lot of value.

But that is not what is rewarded by the system as it is. You must prove a hypothesis in order to move your career forward. Advancing knowledge by invalidating hypothesis is not as lucrative in any sense of the term. And so the temptation to make your own luck, to push the numbers or figures in your direction a bit. Which can, and apparently has, lead to an entire research department bending the rules in order to make their work look more positive and certain than the research can support.

And that is a deep shame. As I mentioned, proving that things are not is just as helpful as riving things that are. Good research leads where it leads. And I suspect most scientists would prefer to work in an environment where they could follow their research and honestly report their work. This is a group of people generally obsessed with learning how things operate. Funneling them toward certain types of outcomes and certain types of research (in this system, research that appears easier to generate positive responses will inevitably get more attention, whether they are more important than other lines of inquiry or not) restricts their imaginations and drive.

Nothing here is surprising nor even all that new. But the seeming increase in these kinds of frauds should stand as a reminder that making every incentive monetary does not lead to good outcomes much of the time. We should be focused on letting scientists who do solid research have stable careers even f all they do is replication and proving that things are not. We would have a much healthier scientific environment, and thus all benefit, if that were the case.

People are not economic units. Treating them like such just produces perverse outcomes, no matter the field.