...beliefs must stay fixed until an impressive enough theorem or data analysis comes along that beliefs should change out of respect for that new display. It also won’t do to keep beliefs pretty much the same when each new study hardly adds much evidence – that wouldn’t offer enough respect to the new display.
If we believe that the sciences are systematically irrational, then isn't this the rational thing to do? To wait for convincing, irrefutable evidence, and after a certain point treat confirmatory evidence as adding nothing?
If scientists are herd-followers and affiliating (both socially and due to institutional pressures), then after X studies showing a link between HIV and AIDS, say, study X+1 adds nothing because the conductors know exactly what they're supposed to get and have no incentive to show the opposite unless they have irrefutably strong HIV!=AIDS evidence.
For perfect Bayesians, even murky or weak evidence is something that shifts one's beliefs; but in the real world, murky or weak evidence of something against common wisdom just makes you look ideologically driven or a young turk who wants publicity (any publicity at all). Knowing this, scientists will avoid weak evidence which is unpopular, which means that only those who are ideologically driven or attention-seekers will publish, which reinforces why other scientists will avoid weak unpopular evidence, in a feedback loop. So only very strong evidence will break through the noise of irrationality.
This is the standard "herding" hypothesis, that public behavior ignore private signals once public signals have become lopsided enough.
Bryan Caplan writes:
Robin Hanson offers an explanation:
I wonder, what does this look like in the cross section? In other words, relative to other academic disciplines, which have the strongest tendency to celebrate difficult work but ignore sound-yet-unimpressive work? My hunch is that economics, along with most other social sciences, would be the worst offenders, while the fields closer to engineering will be on the other end of the spectrum. Engineers should be more concerned with truth since whatever they build has to, you know, work. What say you? More importantly, anyone have any evidence?