Slashdot Mirror

Around the time of the tragic suicide of Aaron Swartz, who lobbied tirelessly for free access to academic articles (in his sometimes grey-hat manner, which ultimately got him in trouble), I admitted to some friends that I didn't understand how this became a problem. Why aren't all journal articles free, all the time?

I don't mean that I didn't know why the journal publishers charged exorbitant fees for their subscriptions. If academic researchers have to have access to journal articles in order to do their jobs, then you can expect the journals to gouge academic libraries on the prices. What I didn't understand was: Why do academics even publish in journals that demand exclusive publishing rights for their work, and then charge readers huge fees to read it?

Well actually, we know the answer to that too: academics want the prestige of publishing in big-name journals that have established reputations, and as a result, those well-known journals are in a position to dictate the terms of the contract. A professor might genuinely want to publish their paper in a journal where it can be read for free by all, but they can hardly be blamed for thinking of their own career path first.

Here's the question I really wanted answered: If "prestige" only exists in the minds of other academics within a field, then why don't the academics within a given field just agree to confer "prestige" on papers published in open-access journals, if they can see for themselves that the quality is equivalent to what would be published in the old-guard journals that charge an arm and a leg? And then make hiring, promotion, and tenure decisions accordingly?

I don't mean that the papers published in an open-access journal would bypass the peer-review process, and that everyone in the field would have to judge the papers for themselves without any prior certification of their quality. One of the points that Peter Suber makes repeatedly in his book Open Access is that open access is not about skipping peer review and dumping papers directly onto the web. Rather, the process would work similarly to peer review for a traditional journal:

1. Author submits a paper to journal XYZ.

2. Journal XYZ selects one or more peer reviewers from among their list of people they consider qualified to review the paper. The peer reviewers send back their usual suggestions and some consensus is reached as to whether or not to publish.

3. If Journal XYZ publishes the paper, then they have certified that the paper passed the quality controls in step #2, and the author can now legitimately claim that they had a paper published in Journal XYZ.

4. If people in the field know that Journal XYZ is not skimping on the quality controls in step #2 — that Journal XYZ is sending the papers to the same academics who would do peer review for one of the old-guard journals, and who are holding the papers to the same standard — then they should respect the paper just as much as if it were published in a traditional journal. If a person has never heard of Journal XYZ, then it should only take a minute to explain to them how it works (and crucially, that Journal XYZ is just as strict about quality as the old-guard journals that everybody has heard of).

Each step in this process should cost the journal virtually nothing. The "hard cost," the part that consumes the time of people with unique skills, is the peer review step, but peer reviewers are usually paid by universities and consider peer review for academic journals to be part of their job description. At a minimum, all the editors really have to do is maintain the list of people they consider qualified to do the peer review, and send the submitted papers off to them.

Moreover, the entire process should be fast. Again, the "hard cost" in time is the peer review, but there's no reason that the delays between submission and publication should be in the range of months or years.

(I'm assuming that the article authors would want their writings to be widely read, or at least would not be opposed to it. That may not be the case if, for example, the authors were commissioned by a pharmaceutical company for a study that cast their drug in a favorable light, but the authors realize that their research methods contained errors and want to minimize the number of eyes on their paper, to reduce the chances of their chicanery being caught. Ben Goldacre's Bad Pharma documents these types of problems very thoroughly, but I'm sidestepping that issue for now.)

So, with that in mind as the ideal, I asked my friends, including many current and former academics, why this essentially wasn't the model that was used. Several mentioned the Public Library of Science, which publishes all articles in its journals under a Creative Commons Attribution License (free for anyone to read and reproduce in full, as long as the original author is cited), and finances its operations through publication fees. These fees are in the $2,000-$3,000 range, heavily discounted for low-income countries and authors, and in any case most academic authors pay the fees out of their research grants and not out of their own pockets. That sounded much better than the traditional model, I thought, but I still didn't understand why the costs weren't even closer to zero. Another friend pointed out that PLOS costs cover the expenses for many of their other activities — which are all noble goals, to be sure, but at the same time, why isn't anybody operating a more bare-bones model which minimizes all expenses, and charges almost nothing for publication or subscription?

This, it turns out, appears to be the approach of the PeerJ project, which aims to let authors pay a one-time fee of $99 at article submission time for the right to publish one article per year — or, if you prefer to pay only if your article is accepted for publication, you can pay $129 "on acceptance" (explained here). And the author of the Techdirt piece mentions that he submitted a paper which was published in the inaugural edition of one of PeerJ's journals, 10 weeks after the submission date. This is cheap and fast enough that I'd call it a validation of the theoretical model which predicts the whole process should be able to be done for almost no cost in almost no time. In other words, I think PeerJ will succeed, but even if it does fail, it will only be because of some anomalous business snafu, not because the hard costs of the service they're providing are greater than the dirt-cheap price they're charging for it. If for any reason PeerJ doesn't happen to get it right the first time, they or some other company should keep trying until someone makes it work.

The basic algorithm at work here — taking a piece of content, submitting it to one or more suitably qualified reviewers, and then certifying the content based on the feedback of the reviewers — is something I've advocated in many contexts over the years, for many different types of problems. In one article I argued that we could make success in the music industry into much more of a meritocracy, with far less arbitrariness in determining who succeeds and fails, if a suitably popular site like Pandora simply took new submissions from artists, had the content "rated" by a random sample of listeners interested in that type of music, and if enough of them liked it, push the content out to all of the fans of that genre. In "Crowdsourcing the Censors" I suggested that Facebook's complaint review process should use the same principle: If a given page received enough complaints, have the page contents reviewed by a random subset of Facebook users who had signed up to be "abusive content" reviewers, and then only flag the page for removal if a high enough percentage of those users voted that the page had indeed violated Facebook's guidelines. This year I argued that "We The People", the White House's online petition-drive-organizing website, should rate ideas based on what a random subset of users think of each idea, rather than allowing users to organize mobs of their friends and followers to vote their own ideas to the top of the pile (which, in case you missed it, is how 4chan gave us this). Or, if you think the general public is not qualified to rate ideas according to how they should be prioritized by the White House (and I'd be inclined to agree), you could have the ideas rated by a random subset of, say, the nation's economics professors.

Of course, I haven't heard of any plans to implement this algorithm in any of those contexts. Not that I expected the key power players to be reading my articles, but it's a little surprising that none of them ever came up with this idea independently, either. (To this day, the only website I'm aware of that ever implemented random-sample voting correctly, was HotOrNot.com, where users could rate members' pictures by attractiveness — but each picture's rating was determined by showing it to a random subset of the site's visitors. That system is gone, since the site has made itself over into a date-finding service.)

But academia in general, and science specifically, is different from other arenas in a number of key ways which could help this algorithm succeed:

• Academia, uniquely, is comprised of many professionals whose love of knowledge and intellectual inquiry, is greater than their desire for money. That's not to say that I don't think the same algorithm could work just as well in a business like the music industry, where most of the stakeholders are in it for the money. But even if Pandora did successfully implement the algorithm, it would meet a lot of resistance from entrenched interests in the music industry, who make their money by finding and promoting and managing talent and would not be happy about a new system that threatened to make them irrelevant. In academia, by contrast, it's quite plausible that even the "entrenched interests" — the people who had become superstars under the old system — would see the new system's great potential for disseminating free knowledge, and would welcome it even if it gave scrappy new upstart academics a chance to dethrone them. Not everybody in academia loves knowledge more than they love their own prestige, but I know more people like that in academia than anywhere else.

• In academia, even among people who do care primarily about their own prestige, many of them have tenure and guaranteed job security, a situation that does not exist in most other industries. This gives them the freedom to experiment with new models, such as submitting papers to upstart PeerJ journals. But more importantly for our purposes, it means they can announce that in their department's hiring and promotion decisions, they will count PeerJ-published papers as legitimate professional accomplishments, for the benefit of non-tenured faculty members who do have to worry about their resume.

• Academics, particularly in maths and sciences, are more prone to the kind of thinking that would lead a person naturally in the direction of the kind of system that PeerJ embodies. First, think of a theoretical model (like the kind I described near the beginning of the article). This model predicts that, ideally, it should be possible to publish papers at very low cost with quick turnaround times, without sacrificing peer-review quality assurance. Now, try to approximate that model as closely as possible in the real world. (In most other industries that I've worked in, there's much more inertia around the existing way of doing things, and far less willingness to entertain any discussion about whether a theoretical model can show how we could accomplish the same thing with vastly less overhead.)

And that, in the end, is the real reason journal articles should be free. Not because the U.S. government is making it a condition for taxpayer-funded research, although that is a welcome development. But because there's no part of the process that should cost very much to begin with, if article authors and peer reviewers are already being paid by their employers. The last piece of the puzzle is that enough academics and faculty departments have to agree to confer "prestige" on articles published in open-access journals, equivalent to the level of prestige that they would accord for an article published in a traditional journal of the same quality. If they won't do that, then the old-guard journals will maintain their monopoly on conferring "prestige", and don't be surprised if journal prices keep growing to the point where even Harvard can't pay for them.