Slashdot Mirror

A a class of yes or no questions is in P if there is an algorithm which in time which is bounded by a polynomial of the length of the problem instance can answer the question. Easy examples of classes in P are "Is the number k have an even number of digits?", and "Does a divide b with remainder c?".

A class of problems is in NP if when it is a set of yes or no questions where when the answer is yes, there is an algorithm to which one can supply a proof that can be checked to be valid in polynomial time. An example of a problem class in NP is "Is the number n composite?" since if the answer is yes, the proof given is to just give the divisor and someone can check that. Note that this problem turns out to actually be in P, but this is a deep result https://en.wikipedia.org/wiki/AKS_primality_test. Also, note that any problem class in P is trivially in NP because the proof one then uses is simply the empty string, and the algorithm is simply one's P-algorithm.

It turns out that certain problem classes are NP-complete. A class of problems is NP-complete essentially if the problem class is in NP, and if one had a magic device which could solve the problem instantly then one could use it to solve all NP problems. It turns out that many, many problems are NP-complete, to the point where many computer science journals will no longer accept papers proving something is NP complete unless there's something noteworthy about it (such as proving it for a historically significant problem like in this case, or proving it for a problem that people have previously not succeeded at proving to be NP complete).

The Journal of Computer Graphics Techniques and The Journal of AI Research are both respectable CS journals that are free for authors and readers. Why can't all science publishing be free? Most of the work is already funded by government grants, so why should anyone have to pay to access it?

Not all areas have pay-to-play open-access journals. In my area, JAIR and JMLR are probably the two most prestigious journals (not just most prestigious open-access journals, but most prestigious period), and they're both entirely open access and entirely free of publication fees.

The expensive-journal commercial publishers don't have much of a competitive moat: anyone can publish PDFs on the internet with the word "Journal" attached to groups of them, and you've got a journal. If that anyone is well-respected in the field and the PDFs are hosted by a well-known university that also prints off some paper copies for archival, you've got yourself a new journal.

In my area this revolt against the commercial publishers has been quite rapid and successful. The entire board of editors left the journal Machine Learning in 2000, setting up the non-profit, open-access JMLR instead, which is now at least as prestigious (possibly moreso). In more general AI, the open-access, non-profit JAIR now has a much higher impact factor than the old Elsevier journal in the area, "Artificial Intelligence".

The main open-access journals in my field don't require author payment at all. They're run through a combination of volunteer labor, frugality, and institutional or professional-society sponsorship. And they're among the top journals in their areas, maybe even the top journals now: JAIR and JMLR. There are a lot of top-tier open-access, no-publication-fee statistics journals as well. Seems to vary by area.

What is more, some first-class journals and conference proceedings in CS are completely free and persistently available on-line. Some of these have been this way for more than a decade now!

Which is exactly why these free and open but reviewed and edited online sources are much better than arXiv.

Yep, that is right, nobody pays the editorial boards, just as nobody pays the reviewers. We do it because we are saints. Also, you are right, slashdot is one big groupthink, which is why nobody ever disagrees with anybody here.

Of course, back in realityland, reviewers do get paid. They don't get paid by publishers, they get paid by their universities. No, not per review or anything. However, at a typical research university, each prof is expected to spend something like 50% of his or her time doing research, 50% of his or her time teaching, and 10% of his or her time doing service. (Does this add up to more than 100%? Well, there is a reason we work more than 40 hours per week.) Guess what? Reviewing papers can be counted as service. So can editing (or helping to edit) a journal.

So, the people doing the work are covered. The websites? These are pretty small and low traffic and the costs can be easily covered by Universities for the prestige of it. The only cost that isn't covered at this point is for the printed copies to go into University archives and these can be farmed out to publishing companies that will sell them to libraries for a couple hundred dollars per year. If people become convinced they don't need these, that cost will vanish.

The only reason more journals and conferences haven't gone this route is momentum. To start up a new journal or conference, or to switch an existing one to open, online access takes time and effort. Did I mention we already work long hours? You don't need massive government warehouses of information. You just need to pay a few faculty a couple of semesters per journal you want created/switched over and we'll take it from there. We want our research to be available to everyone.

Dean

I don't know in your field, biology, but in other areas this certainly is not the case any more. For example, the best journal in Artificial Intelligence is fully available online for free.

I believe that the scientific Journal has outlived its usefulness, and will be replaced by ... Slashdot!

I seriously doubt that. I think the more likely future is online-only Journals. IEEE conferences are certainly moving in this direction, though they require paid access to their site. IMO the best open journal at the moment is the Journal of Artificial Intelligence Research. They have free access to all papers online and keep their costs low.

But seriously, reviewers are biased and sloppy, as are the editors.

I'm sorry, are you describing Slashdot or journals?

The fact that reviews are blind means that they are also unaccountable, which fosters even more bias.

If your research can't stand up to what people might say behind your back, then maybe you ought to support it better.

Journals take months or years to respond to a submision, and often as not they respond with a rejection so the submitter has to give up or start the whole process over with another journal.

Timliness is a problem, but reviewers are human; It takes a while to find the time to do a journal review. There are many other options than outright rejection; In fact at least within computer science, "revise and resubmit" is a popular option. You have to fix what's wrong, and it will get reviewed again. I think that's pretty reasonable.

Now, if your work is completely out of left field, it will get rejected. If you didn't take the time to think something through clearly AND understand what others have already done in your area, then you are simply wasting the reviewers time.

There are so many scandals that one could quote.

Interestingly, you don't give any examples or a reference. Maybe there really aren't that many scandals. Also, there are plently of journals to choose from, so find one that hasn't had a major scandal.

The whole process seems more designed to support the status quo than to promote knowledge.

It's run by a community, and pretty much everything run by a community works that way. You seem to be confused about the roles in this process. It is your job as author to convince the reviewers that you are right, it's not their job to automatically recognize your genius when you don't make enough of an effort to present it.

I have discussed this with many people in academia and they react not with logic, but with horror that I would dare to question a system that they view almost mystical reverence.

Well, maybe that's because you advocate complete removal of a system that failed to serve you, but serves a lot of researchers just fine. Instead of thowing out the baby with the bathwater, why not help journals like JAIR which fix only what's broken in the system, and try to keep the good parts. If you submit to such journals, and reivew in fields what you are well versed in, these journals will quickly rise to prominence. There's no monopoly here; The system can be fixed with competition.

In the field of AI we have at least two highly respected journals which do not have paper editions (even though libraries can buy bound collections of papers on a yearly basis) and which make their content available for free to everyone:
Journal of Artificial Intelligence Research and
Journal of Machine Learning Research

This works because an academic journal does not really have any expenses for peer-review. Academics review for free as part of their job - it gives status to review for a prestigious journal. If you don't have any costs for editing and printing a paper edition, you suddenly have almost no operating expenses at all. Cost of bandwith is negligible. A typical research paper in pdf format is a 100k download, so any one of us could operate one of those servers from our home. Furthermore, the cost of bandwith is continually decreasing.

In sum, I don't understand what the IEEE is whining about. Let those who want a journal on paper pay for the paper, and let the rest of us have it for free!

Well, there's much good to be said about dead trees. :) On one hand, paper journals are great for archival purposes - you can go to your local library, and dig up publications from a hundred years ago. At the same time, the internet is entirely too impermanent - what if Springer Verlag publishes a journal, and then they go bankrupt in 10 years? The chances of the publications disappearing or becoming unavailable are pretty high. But endangering the access to all the accumulated knowledge simply because of economic accidents is not an acceptable risk in the scientific community.

So a joint paper/electronic model seems like the right balance. Most journals do that already - libraries subscribe to dead tree versions, and individuals can access the papers online, usually through a school-related discount subscription. Seems to work quite well although, paradoxically, it increases the cost per unit (because now you're printing far fewer issues).

But there's simply no incentive for publishing houses to make the online content completely free. Professional organizations can do it themselves (e.g. the AI Access Foundation), where they publish online papers themselves, and contract with a publisher to print each entire volume as a book. Non-profits like these will probably be the harbingers of new method of distribution for scientific findings...

There are other free journals out there as well - the one I'm most familiar with is the Journal of Artificial Intelligence Research, which is probably one of the most respected AI research volumes and has been published online since 1993.

I agree that open review should happen, in fact, I think its inevitable. The Internet IS bringing more power to the individual. No longer should we expect to see so many top-down pre-established hierachies of power, we should get used to the bottom-up emergent organization fostered by the massive connectivity of the Net.

Examples of publishing include the already mentioned physics archive and JAIR which is published online. It is still reviewed in a traditional manner but has plans for an open review process.

Also see the Interactive Paper Project for some technology that already allows open review (I think its a better approach than slashdot for papers) Another option would be that company that allows user to "post" messages to websites.

My point is that the only real barrier is the established publish-or-perish publishing-house culture and, like any culture, it is just a matter of time before it evolves to match the available technology.

--Books used to be only for monks, then came the printing press--

Of course, I have no idea how long it will take. Soon, I hope.

Jose