Earlier this week, Rufus Pollock published a proposal for reforming academic publishing. I do agree with some of the basic assumptions behind this piece: that academic publishing is hopelessly broken and doesn’t serve anybody well except for-profit publishers, and that technology can play a significant part in the solution. (I have in fact said these things before.) But here’s the thing: this is a social problem, and social problems do not easily lend themselves to purely technical solutions. None of the problems Pollock’s proposed solutions address are the actual problems that need solving to make academic research widely available to the public at no cost at the point of use.
There are three factors we need to consider to better understand the problems we actually need to solve: 1. the traditional model of academic publishing; 2. open access both as intended and as implemented; 3. what academics actually get CV points for (which is slightly different to what we actually care about, but that’s a separate problem).
1. I’ll stick to the model for academic papers. Books are slightly different but similarly broken. Academic journals, which is where most academic papers appear, are traditionally run by one of a handful of academic publishing houses. Sage, Elsevier, Springer, Wiley-Blackwell and Taylor & Francis between them publish 50% of all academic papers. Their costs are fairly minimal. Most journals will have a paid editor and some production staff. Print distribution does happen but with the advent of electronic journals it’s in severe decline. Importantly, there are two big-ticket items that publishers don’t pay for: content, and quality assurance. Content is provided free by academics because that’s how we get CV points. Quality assurance (also known as peer review) is also provided free by academics because that’s how we get CV points/get to snark at our colleagues/arch-nemeses.
Getting hold of a single academic paper through this model costs the end user anywhere between US$30 and US$120. Subscriptions significantly lower the cost per paper, as do subscription packages – but most people don’t want to read every single paper in a specific journal over the course of a year, so subscriptions/packages only make sense for large institutions, mainly academic libraries. There’s another part to this system that Pollock picks up on, which is indexing, i.e. making papers discoverable. I actually don’t know how the indexing business model works but if I were to guess based on how broken the rest of the industry is, journals and libraries both pay subscription fees for anything other than maybe Google Scholar. (Google is in a different business and also only partially useful for indexing academic papers.)
2. It is a perfectly reasonable position to look at the above and declare it broken. It’s broken in several ways, but most glaringly in that both the people who create the content and the people who (indirectly, through taxes) pay for the vast majority of research to be done in the first place, can’t access the content without putting money in the pockets of large private corporations who have contributed, at most, a bit of branding and a bit of admin to the entire thing. And since we have the internet, it seems entirely reasonable to go “just publish it all online for free!” And thus Open Access was born. Except you still have massively powerful corporations in the system who aren’t going to keel over so easily and lose their nearly pure profit. So the first thing they tried was to discredit the system. “Oh, but if you have Open Access then who will do quality assurance? It’ll be a free for all!” And when that didn’t work well enough (it did do some damage to the credibility of Open Access unfortunately) they lobbied until Open Access got redesigned in a way that would still let them profit from it in much the same way as before.
This is how Open Access works: there are two levels – green and gold. Some journals operate both, some only one. There might even still be a few out there that offer neither. Green level OA works on time: the journal has exclusive distribution rights for a given amount of time (normally between 6 and 24 months), after which the paper becomes Open Access and can be accessed by everyone for free (which may or may not be hosted by the journal – but most UK universities now have their own OA research repositories where stuff gets hosted). Add 24 months to the academic publishing cycle, which can take several years in the first place, and by the time the general public gets to read “cutting edge research”, it’s based on ten-year-old data. In physics, where the speed of light doesn’t change that much, this is mildly inconvenient. In computer science, which moves at the speed of Moore’s Law, this system is not even remotely fit for purpose. (Let’s not talk about the arts and humanities, that’s an entirely separate rant.)
Gold OA simply shifts which university department pays the publisher: if you want your paper to be published under Gold OA, you as the author – or your institution – have to pay the publisher for the privilege. Depending on journal, field and publisher, we’re talking US$1,000 – US$6,000 per paper. Now, the assumption here is that eventually all academic publishing will transition to full OA so libraries won’t have to pay subscription fees to access articles anymore, and therefore the money that’s currently being spent on that can be spent on on Article Processing Charges instead, and universities won’t be any worse off. There are about six million problems with this. For a start there is what is now looking like it’ll be quite a lengthy transition period. Between nuclear fusion and full OA, my money is on getting nuclear fusion first. During this transition period, universities are stuck having to pay both ways. The people this screws over most are PhD students and early career researchers, particularly the ones not funded by research councils. If you’re funded by a research council, there may be some money as part of your funding for dissemination and OA publication. If you’re not, tough luck, Gold OA is something you can only dream about. Other people this screws over are increasingly casualised researchers (often also early career) and independent scholars who don’t have an institution or funder behind them (who, unsurprisingly, are mostly from already marginalised groups). To add insult to injury, people who don’t understand the gory details of OA tend to view Gold OA as vanity publishing.
And yes, there are independent academic journals which run entirely not for profit on the Green OA model, making papers available immediately. They are few and far between, those that do exist have to fight hard to establish credibility and reputation (because of the work publishers did to discredit OA originally), and of course this model relies entirely on people wanting to contribute to their community for very little reward. Some examples of this working well are Transformative Works and Cultures which is a fan studies journal run by the Organisation for Tranformative Works, and the one where the entire academic staff of one of the biggest linguistics journals told Elsevier they could stick it and started an OA journal instead. But in the vast majority of cases we’re still stuck with the big publishers.
But what about arXiv.org I hear you ask! Which brings us neatly to 3: what do academics actually get CV points for?
3. (This bit covers the more or less gory details of research funding in the UK, but the broad principles of prestige, peer review, and funding being tied to publication applies, with minor tweaks to the exact metrics, to anyone wanting to pursue an academic career at a Western university.) Research at UK universities is funded in two main ways: by the research councils, which tend to fund specific projects, and (in England) by HEFCE which allocates generic lump sums of money every five years or so. I used to know the percentage split of how much money comes in from HEFCE and how much from research councils but I don’t anymore – either way, the HEFCE money is a significant chunk for many universities. How much HEFCE money a university gets is determined by an arcane process called the REF (Research Excellence Framework), where every five or so years universities submit their research output to a panel (by subject) and the panel decides how good the research output is, on a scale from one to four stars (four being world-leading). Every star on every paper means a certain amount of money annually for the university. A four-star paper is worth about £10k a year, which means that if you have four four-star papers in a given REF period, your salary is more or less covered by the HEFCE money that brings in. Which in turn means you might be able to get a/keep your job for the next REF cycle. (It’s more complicated than that, but that’s the gist of it.)
Theoretically the REF panel judges papers on their own merits, not on where they were published. Practically, the sheer volume of work they have to get through… Yeah. There’s a strong incentive to publish in established, high-impact-factor (don’t ask), for-profit journals. There is literally zero incentive to just chuck your paper on something like arXiv and run. The reason arXiv works for the community that uses it is that it’s a pre-publication archive: the papers, in slightly altered form, are still published in the big academic journals where they get their impact factor and REF eligibility etc. (Now, HEFCE has said that any paper published after April 1, 2016, to be eligible for the 2020 REF, needs to be in an OA repository. What this means in practice is that if your paper is still embargoed under Green OA, or is in a non-OA journal but is sitting behind a password on your institutional repository, your’re fine. Which is some great hoop-jumping for not a lot of direct benefit to the general public.)
So these are the actual problems any review of Open Access or any other solution seeking to make academic research available to the public should be looking to solve. Pollock’s proposed model doesn’t come near it. From what I can see, it eliminates peer review entirely. Now, peer review is broken in several interesting ways, but unless we fix how research is funded and academics are employed, it’s one of the less bad ways of doing quality assurance on academic research. Because all three of Pollock’s filtering/selection models eliminate the double-blind aspect of peer review, the system automatically becomes even less accessible to marginalised groups (aka the “try hanging out on Reddit as not-a-cishet-white-dude phenomenon”). Because it doesn’t take into account any of how research itself is funded and what the key stakeholders in this process (researchers, their employers and funders) actually care about, the model is completely unworkable. And because it doesn’t take into account the vested interests which shape the current system (for-profit publishers), even if it was workable there’s no clear path to implementation.
All of which is why requirements engineering is an art, and why you shouldn’t try to fix complex social problems with technical solutions.