It'd be really cool if we can add more social dynamic features to papers / research, and have people correct papers in the open, like twitter community notes, to the point that if you can really disprove it, the paper is now marked "wrong" or something like that on arxiv. Sort of inbetween collaborative science or hunting for patent prior art or bug bounty hunting. Getting paid for it would be even better, making money by sniping issues in research. One can dream.
I wonder how much money in bounties it'd take to proof-invalidate whole branches of psychology (insert your own pet peeve here) research in a methodical way, probably not that much.
I wish academic papers were more like Wikipedia articles. Currently what I'm working on is really "Building on" one pretty pivotal paper from the 90s, and there's a whole constellation of work that has spawned.
So much ink is spilled re-defining the problem, and reading any paper requires going through the system model every time because tons of arbitrary decisions may have been done different. It makes it hard to compare results, and makes almost every statement that reads "Over in this area we're not innovating on, we used the SOTA" wrong, because some other group is innovating in that corner.
If instead there was one canonical version of it with an edit history, and I could go try to just re-write one little para and argue in the talk section about it with the one-or-two other groups picking away at that, I feel like things could move faster and be done higher quality.
It'd also be a lot easier to peek at other areas. Currently if I have a question like "What's the latest in NeRFs underwater? I remember seeing a paper about that a while ago" I've basically got no idea.
> I wish academic papers were more like Wikipedia articles.
I don't think that would be helpful. Scientific development happens in branches, not linearly. The fact that a field is going in one direction does not mean that somebody won't make a breakthrough next year based on a poorly-cited paper from the 1970s, leapfrogging a whole bunch of studies that happened in the meantime.
Most of the time, there is simply no "state of the art" that covers a whole field, and even in limited sub-fields, quite often there is no consensus.
Academic publishing doesn’t make the authors money, it costs them money. To publish is a requirement for academic employment, but there’s no incentive to retract, either than revenge, or one’s academic honesty.
Without knowing the specific context: I think this really is a good example of how errors should be disclosed.
We need to acknowledge that scientists/academics are human; even very competent mathematicians make mistakes and some of these mistakes appear in published papers. What we lack in many fields is a culture and process that allows (and ideally, encourages) one to disclose: "this was wrong, here is how I fixed it, or how it's actually correct". E.g., in the communities I know in Computer Science & AI, I rarely even see errata lists on personal webpages, not to speak of journals that provide a straightforward process for updates. I would even go so far to claim that the current culture, in which honest errors cannot be straightforwardly corrected, plays into the hands of the clearly dishonest "bad apples".
Science is, obviously, not a "monotonic" process in which every single paper adds to the truth; this is practically not even the case for mathematics, which is at least monotonic on object-level (but mistakes happen all the time). As a prominent example, consider this impressive list of Feynman errata: https://www.feynmanlectures.caltech.edu/info/flp_errata.html.
> 1999: It is not clear to us in general how to avoid this sort of false proof, the problem being that the false statement seemed so natural to us that we did not think to look at it carefully.
Assuming I understand correctly, this is basically the common issue of being unable to be objective when you’ve lived and breathed a subject matter for long enough. The answer is rigorous peer review, I think.
Perhaps its because your comment seemed to equate errors with crimes, or at least malicious intent. The language seems a bit provocative for many, detracting from whatever message was intended.
I like that they are very blunt and to the point.
It'd be really cool if we can add more social dynamic features to papers / research, and have people correct papers in the open, like twitter community notes, to the point that if you can really disprove it, the paper is now marked "wrong" or something like that on arxiv. Sort of inbetween collaborative science or hunting for patent prior art or bug bounty hunting. Getting paid for it would be even better, making money by sniping issues in research. One can dream.
I wonder how much money in bounties it'd take to proof-invalidate whole branches of psychology (insert your own pet peeve here) research in a methodical way, probably not that much.
I wish academic papers were more like Wikipedia articles. Currently what I'm working on is really "Building on" one pretty pivotal paper from the 90s, and there's a whole constellation of work that has spawned.
So much ink is spilled re-defining the problem, and reading any paper requires going through the system model every time because tons of arbitrary decisions may have been done different. It makes it hard to compare results, and makes almost every statement that reads "Over in this area we're not innovating on, we used the SOTA" wrong, because some other group is innovating in that corner.
If instead there was one canonical version of it with an edit history, and I could go try to just re-write one little para and argue in the talk section about it with the one-or-two other groups picking away at that, I feel like things could move faster and be done higher quality.
It'd also be a lot easier to peek at other areas. Currently if I have a question like "What's the latest in NeRFs underwater? I remember seeing a paper about that a while ago" I've basically got no idea.
> I wish academic papers were more like Wikipedia articles.
I don't think that would be helpful. Scientific development happens in branches, not linearly. The fact that a field is going in one direction does not mean that somebody won't make a breakthrough next year based on a poorly-cited paper from the 1970s, leapfrogging a whole bunch of studies that happened in the meantime.
Most of the time, there is simply no "state of the art" that covers a whole field, and even in limited sub-fields, quite often there is no consensus.
I too, would like to have what you’re having.
Academic publishing doesn’t make the authors money, it costs them money. To publish is a requirement for academic employment, but there’s no incentive to retract, either than revenge, or one’s academic honesty.
Without knowing the specific context: I think this really is a good example of how errors should be disclosed. We need to acknowledge that scientists/academics are human; even very competent mathematicians make mistakes and some of these mistakes appear in published papers. What we lack in many fields is a culture and process that allows (and ideally, encourages) one to disclose: "this was wrong, here is how I fixed it, or how it's actually correct". E.g., in the communities I know in Computer Science & AI, I rarely even see errata lists on personal webpages, not to speak of journals that provide a straightforward process for updates. I would even go so far to claim that the current culture, in which honest errors cannot be straightforwardly corrected, plays into the hands of the clearly dishonest "bad apples".
Science is, obviously, not a "monotonic" process in which every single paper adds to the truth; this is practically not even the case for mathematics, which is at least monotonic on object-level (but mistakes happen all the time). As a prominent example, consider this impressive list of Feynman errata: https://www.feynmanlectures.caltech.edu/info/flp_errata.html.
> 1999: It is not clear to us in general how to avoid this sort of false proof, the problem being that the false statement seemed so natural to us that we did not think to look at it carefully.
Assuming I understand correctly, this is basically the common issue of being unable to be objective when you’ve lived and breathed a subject matter for long enough. The answer is rigorous peer review, I think.
> "Should the Democrats move to the left?
>
> Because of a data coding error, all of our analysis of social issues is incorrect.
Yeah, sure, that's why the analysis was incorrect, it was all about that typo...
The cover up is usually worse than the crime.
Do you think these errata being published are a good thing?
Yes of course. That's the point of my comment, which is being downvoted for some reason that eludes me.
Perhaps its because your comment seemed to equate errors with crimes, or at least malicious intent. The language seems a bit provocative for many, detracting from whatever message was intended.
Sometimes I write with "downvotes be damned" in mind and connected with the audience exactly how I intended.
But not connecting with the audience is the usual reason.
When it happens to me, I take it as feedback on my writing. Maybe I was unclear. Maybe I was wrong. Maybe it was written for a different audience.
In those cases, I just try to improve my writing.
Anyway, where can responses to your original comment go?
They could dispute your maxim and the internet gets another argument where nobody changes their mind.
Or they could agree with it and the internet gets another dog pile of cynicisms.
Generally, those are not why people come to HN...at least when the form is one-liners.
Finally, complaining about downvotes is contrary to the HN guidelines. Good luck.