Can't we have a system where a paper is not necessarily reviewed at first, but as time goes by, more and more reviewers endorse the paper?

The review process gives authors feedback which they use to improve their paper before it's published, and authors usually appreciate that.

It's also a chance to stop major errors being published, and again authors usually appreciate things like that being caught before the paper becomes public.

I mean, why not though? We do it with wikipedia. Maybe a publication could be submitted for review, but not made public (like, its up on the wiki, but not 'publicly view-able yet, only available to editors).

This is what happens, it's the "peer review" thing being described. The reviews are issued so that corrections can be made before publication.

Wikipedia/ Stackoverflow style review process. Also, ffs can we get hyperlinked citations? Its fung ridiculous that I can't just click the link for your citation and go to that paper.

This is hard for new authors as you don’t know if the reviewer is competent. So new authors would have to pay. So a problem.

I don’t literally mean you do n reviews to get your review - I mean members of the community generally spend some time reviewing. If you're new you will get some of your papers reviewed before you start to contribute back but it’s not a problem. It’s how it works now in CS.

> reciprocal peer review so you pay other people to review your papers by agreeing to review others’ papers, and then publishing on arXiv.

This is fraught with ethical problems.

It doesn’t need to be directly reciprocal. You could have a system where having your paper peer reviewed costs X tokens, and you earn tokens by peer reviewing other papers that don’t have citation links to yours. Or you buy tokens with cash which sustains a fund for external reviewers. Or other people can transfer their tokens to you.

Would that still have problems?

Sure, but the identical exchange of value occurs in the current system, it just goes unsaid/unspoken. You review my journal and maybe ignore some of the problems, poor assumptions, whatever crap I used to make the sausage, just let me get it published; I do the same for you. You may not 'know' who is reviewing your journal, but at least in the natural sciences, there are so few technically qualified individuals to review some very domain specific publications, so its almost inevitable that although you may not know for sure who is reviewing your paper, you can figure it out without too much work. Since both reviewers and authors benefit from getting their own work published, there is a silent consensus for letting bad work slip through.

No obviously I don't mean you review a person's paper and they review yours - use your common sense, mate.

I mean if you publish a paper and get four reviews, you then review another four papers at some point - perhaps at an entirely different conference later that year. You just make sure you review at least 4n papers for each n papers you publish.

> Computer science basically does this but replaces arXiv with an industry body the ACM.

There's a big difference: arXiv papers are open-access (everyone can download them), but ACM papers are closed-access (you need a subscription to read them). I wish that computer science research were published on arXiv, but we're not there yet...

But:

1) the ACM lets each author link to their papers on the ACM library with free access.

https://www.acm.org/publications/authors/acm-author-izer-ser...

2) in practice you could always find pre-print copies of any CS paper on their author's website already and everyone knew and nobody ever cared

3) the ACM also let you negotiate an agreement so that you retain copyright for the paper so you can publish it yourself - my company does this

I disagree with these points.

1) Essentially no one knows about the author-izer system or understands how it works, especially outside academia. Readers in companies, poorer countries, etc., can't be expected to guess that the way to read an article is to go to the author's webpage and follow author-izer links (assuming they have been set up). What these potential readers will do is: search something on Google (or follow a link from somewhere else), hit the paywalled ACM DL page, give up. This convoluted system of "open-access from one place, closed-access from another" makes no sense.

2) For authors who actually post preprints of their work, yes, you can read it this way. But then you end up with multiple versions of the same work, that are often subtly different: does the author's preprint integrate reviewer feedback? does it fix some bugs that were found after the camera-ready version was submitted? And anyways, preprints posted on authors' websites usually disappear when they change institutions or retire, so it's not a good solution.

3) Yes, you can retain copyright on papers published with ACM, but then you need to give them an exclusive license to publish, so this still limits what you can do with your work (besides some narrowly worded exceptions). There is also a 3rd option of making the work open-access with no exclusive transfer, but this costs at least $700 per article, which is obviously excessive compared to the actual costs of hosting a 12-page PDF.

So I don't think it's fair to compare publishing with ACM and publishing on arXiv, because ACM is not open-access and publishing with them requires you to pay excessive fees or sign agreements restricting how you can publish your work, i.e., the opposite of what's in the interest of science.

Indeed, these and other problems exist with author-posted versions, also putting researchers at disadvantage by being less read and cited: https://gitlab.com/publishing-reform/discussion/issues/17

Re 2), there were instances I seem to recall when academic publishers went after authors that had preprints of their own papers on their own website.