This would be ultimately filled with trolls, nonexperts, and people with an ulterior motive or grudge: so any such effort would be less trustworthy than the original publication.
If you want to find mistakes, you can find one in every paper. The character of the mistakes is important - is it fraud, is it incompetence, does it negate the results? Or is it good science being done by a human? An open website doesn't seem to me likely to be able to draw that out.
> any such effort would be less trustworthy than the original publication.
The idea that scrutiny would be less reliable than blind trust is absurd. The question in the OP, for example, could have been in the comments sections of these papers.
I understood that this was about scrutinising papers in academic journals - the academic journal's value is quite literally its trustworthiness. (The journal exists and is employed to do the scrutiny)
A comment from any random person (in general) holds a lower level of trustworthiness.
This post - and many other conversations we’ve had on the subject on HN - are about the lack of integrity of academic journals. More broadly, this contributes to discussions about fraud in academia, the reproducibility crisis, and the pressure to publish.
“Any random person” includes many researchers, including phd holders or just random people with time on their hands, but whose commentary could be judged on its own merits, not by some credentials or stamp of approval from journals that don’t even examine the data used by studies they publish. This does not mean that comments should go completely unmoderated.
As far as I’ve seen, no journal does a thorough examination of data referenced by studies it publishes.
Credentials, papers, citations, and studies do little to increase the levels of trustworthiness precisely because papers like these are not publicly scrutinized.
This would be ultimately filled with trolls, nonexperts, and people with an ulterior motive or grudge: so any such effort would be less trustworthy than the original publication.
If you want to find mistakes, you can find one in every paper. The character of the mistakes is important - is it fraud, is it incompetence, does it negate the results? Or is it good science being done by a human? An open website doesn't seem to me likely to be able to draw that out.