The hierarchical power structure of academia (post-docs and students depend on the lab director for recommendations and funding to progress in their career), as well as incentive structures for promotion and tenure that emphasize publications and grants over all else do us no favors here.
I agree 100%. This is why I think we really need to change the incentive structures and norms, and make it much easier and safer for whistleblowers to come forward (jay)
So much of the research process as a team is built on trust. Questioning our colleagues' work and asking to double check the data ourselves can be socially uncomfortable. It can be interpreted as a lack of trust, which can have negative consequences in the relationship. Putting policies in place before the research even starts can alleviate that discomfort because your questions and verifications mean that you're just following protocol, not that you don't trust your colleague.
I agree with previous comments that point to the incentive system as the problem. Of course we should have robust data analysis pipelines and preferably an internal review process in the team or across the lab as a common practice to discover errors in data analysis but these will not combat the incentive to game the system. If gaming is easy and rewarding, it will persist. The real lever is making gaming less advantageous by playing acording to different rules (i.e., incentivise generalizability strengh, collaborations, implementation readiness etc instead of weak proxies of quality like num of citations and num of publications).
The hierarchical power structure of academia (post-docs and students depend on the lab director for recommendations and funding to progress in their career), as well as incentive structures for promotion and tenure that emphasize publications and grants over all else do us no favors here.
I agree 100%. This is why I think we really need to change the incentive structures and norms, and make it much easier and safer for whistleblowers to come forward (jay)
So much of the research process as a team is built on trust. Questioning our colleagues' work and asking to double check the data ourselves can be socially uncomfortable. It can be interpreted as a lack of trust, which can have negative consequences in the relationship. Putting policies in place before the research even starts can alleviate that discomfort because your questions and verifications mean that you're just following protocol, not that you don't trust your colleague.
Observation:
A fraudulent result can spread remarkably far when it confirms what people already hope to be true.
Observation:
The article shows that scientific errors are not merely technical failures, but also failures of scrutiny, incentives, and skepticism.
Conclusion:
The problem was not fraud alone.
It was the number of checkpoints the fraud managed to pass.
A false observation may deceive one researcher.
A false conclusion that deceives an entire system is a different mystery altogether.
I agree with previous comments that point to the incentive system as the problem. Of course we should have robust data analysis pipelines and preferably an internal review process in the team or across the lab as a common practice to discover errors in data analysis but these will not combat the incentive to game the system. If gaming is easy and rewarding, it will persist. The real lever is making gaming less advantageous by playing acording to different rules (i.e., incentivise generalizability strengh, collaborations, implementation readiness etc instead of weak proxies of quality like num of citations and num of publications).