I agree with the target you’re shooting for here but I think it’s not nearly that simple to get there.
For example, you’d just have labs doing extremely expensive replications of extremely simple or well-established experiments. Why not?
IMO maybe an alternative solution: a government-run journal that 1) pays researchers handsome sums upon publication, 2) only publishes high-quality replications, and 3) only publishes replications that increase marginal confidence on a defensible important question.
That way this journal could pick up “negative results” and other “not-blockbuster” results that might come about as a byproduct of pursuing other research goals, and there’s no incentive to replicate meaningless work, and there’s no central planning required of the scientific endeavor (of which there’s already too much due to the funding mechanisms).
> For example, you’d just have labs doing extremely expensive replications of extremely simple or well-established experiments. Why not?
You're making grants sound like a cost-plus contract. I don't think grants currently work like that. Why would they be changed?
As to why that wouldn't happen under the current system. Because that lab would go under and before they published the replications (these things do get reviewed).
And after a certain number of replications, you would (presumably) stop funding those replications. Maybe re-visit it once every decade or so in the softer sciences. You never know what's a product of the current moment in something like sociology or psychology.
> IMO maybe an alternative solution: a government-run journal
I'd worry about inertia in a government run program. You can just point funding at a different institution when you're only talking about government funding, but when something is government run, it tends to be harder to kill than just re-directing funding.
I think you'd have to set up an entire ecosystem of government run journals, but that still makes it harder to pivot if one of the non-government run journals starts doing way better work.
For example, you’d just have labs doing extremely expensive replications of extremely simple or well-established experiments. Why not?
IMO maybe an alternative solution: a government-run journal that 1) pays researchers handsome sums upon publication, 2) only publishes high-quality replications, and 3) only publishes replications that increase marginal confidence on a defensible important question.
That way this journal could pick up “negative results” and other “not-blockbuster” results that might come about as a byproduct of pursuing other research goals, and there’s no incentive to replicate meaningless work, and there’s no central planning required of the scientific endeavor (of which there’s already too much due to the funding mechanisms).