2 ms·

You want to solve reproducibility? Apply forensic scrutiny to every publication? That's a great idea, but who is going to pay for it? Every scientist knows th

by vkou 3y ago

You want to solve reproducibility? Apply forensic scrutiny to every publication? That's a great idea, but who is going to pay for it? Every scientist knows that it's more than possible to publish deliberately fraudulent work. Every non-scientist thinks that detecting this fraud is easy.

_jal 3y ago
> Every non-scientist thinks that detecting this fraud is easy. Eh, this working engineer doesn't think so. Freeform fraud detection is a hard problem in general, and when you're talking about (apparent) cutting-edge research, the pool of people who might spot specific details is tiny. A low probability of detection does suggest that the penalty should be relatively high to have a deterrent effect.