Monday, March 13, 2017

The OA interviews: Philip Cohen, founder of SocArXiv

Fifteen years after the launch of the Budapest Open Access Initiative (BOAI) the OA revolution has yet to achieve its objectives. It does not help that legacy publishers are busy appropriating open access, and diluting it in ways that benefit them more than the research community. As things stand we could end up with a half revolution.

But could a new development help recover the situation? More specifically, can the newly reinvigorated preprint movement gain sufficient traction, impetus, and focus to push the revolution the OA movement began in a more desirable direction?

This was the dominant question in my mind after doing the Q&A below with Philip Cohen, founder of the new social sciences preprint server SocArXiv.

Preprint servers are by no means a new phenomenon. The highly-successful physics preprint server arXiv (formally referred to as an e-print service) was founded way back in 1991, and today it hosts 1.2 million e-prints in physics, mathematics, computer science, quantitative biology, quantitative finance and statistics. Currently around 9,000-10,000 new papers each month are submitted to arXiv.

Yet arXiv has tended to complement – rather than compete with – the legacy publishing system, with the vast majority of deposited papers subsequently being published in legacy journals. As such, it has not disrupted the status quo in ways that are necessary if the OA movement is to achieve its objectives – a point that has (somewhat bizarrely) at times been celebrated by open access advocates.

In any case, subsequent attempts to propagate the arXiv model have generally proved elusive. In 2000, for instance, Elsevier launched a chemistry preprint server called ChemWeb, but closed it in 2003. In 2007, Nature launched Nature Precedings, but closed it in 2012.

Hope springs eternal

Fortunately, hope springs eternal in academia, and new attempts to build on the success of arXiv are regularly made. Notably, in 2013 Cold Spring Harbor Laboratory (CSHL) launched a preprint server for the biological sciences called bioRxiv. To the joy of preprint enthusiasts, it looks as if this may prove a long-term success. As of March 8th 2017, some 8,850 papers had been posted, and the number of monthly submissions has grown to around 620.

Buoyed up by bioRxiv’s success, and convinced that the widespread posting of preprints on the open Web has great potential for improving scholarly communication, last year life scientists launched the ASAPbio initiative. The initial meeting was deemed so successful that the normally acerbic PLOS co-founder Michael Eisen penned an uncharacteristically upbeat blog post about it (here).  

Has something significant changed since Elsevier and Nature unsuccessfully sought to monetise the arXiv model. If so, what? Perhaps the key word here is “monetise”. We can see rising anger at the way in which legacy publishers have come to dominate and control open access (see here, here, and here for instance), anger that has been amplified by a dawning realisation that the entire scholarly communication infrastructure is now in danger of being – in the words of  Geoffrey Bilderenclosed by private interests, both by commercial publishers like Elsevier, and by for-profit upstarts like ResearchGate and Academia.ede (see here, here and here for instance).

CSHL/bioRxiv and arXiv are, by contrast, non-profit initiatives whose primary focus is on research, and facilitating research, not the pursuit of profit. Many feel that this is a more worthy and appropriate mission, and so should be supported. Perhaps, therefore, what has changed is that there is a new awareness that while legacy publishers contribute very little to the scholarly communication process, they neverthelss profit from it, and excessively at that. And for this reason they are a barrier to achieving the objectives of the OA movement.

Reproducibility crisis

But what is the case for making preprints freely available online? After all, the research community has always insisted that it is far preferable (and safer) for scholars to rely on papers that have been through the peer-review process, and published in respectable scholarly journals, in order to stay up to date in their field, not on self-deposited early versions of papers that might or might not go on to be published.

Advocates for open access, however, now argue that making preprints widely available enables research to be shared with colleagues much more quickly. Moreover, they say, it enables papers to potentially be scrutinised by a much greater number of eyeballs than with the traditional peer review system. As such, they add, the published version of a paper is likely to be of higher quality if it has first been made available as a preprint. In addition, they say, posting preprints allows researchers to establish priority in their discoveries and ideas that much earlier. Finally, they argue, the widespread sharing of preprints would benefit the world at large, since it would speed up the entire research process and maximise the use of taxpayer money (which funds the research process).

Many had assumed that OA would provide these kind of benefits. In addition to making papers freely available, it was assumed that open access would introduce a quicker time-to-publish process. This has not proved the case. For instance, while the peer review “lite” model pioneered by PLOS ONE did initially lead to faster publication times, these have subsequently begun to lengthen again.

Above all, open access has failed to address the so-called reproducibility crisis (also referred to as the replication crisis). By utilising a more transparent publishing process (sometimes including open peer review) it was assumed that open access would increase the quality of published research. Unfortunately, the introduction of pay-to-publish gold OA has undermined this, not least because it has encouraged the emergence of so-called predatory OA publishers (or article brokers), who gull researchers into paying (or sometimes researchers willingly pay) to have their papers published in journals that wave papers past any review process.

The reproducibility crisis is by no means confined to open access publishing (the problem is far bigger), but it could hold out the greatest hope for the budding preprint movement.

Why do I say this? And what is the reproducibility crisis? Stanford Professor of Medicine John Ioannidis neatly summarised the reproducibility crisis in 2005, when he called his seminal paper on the topic “Why most published research findings are false”. In this and subsequent papers Ioannidis has consistently argued that the findings of many published papers are simply wrong.

Shocked at Ioannidis’ findings, other researchers set about trying to size the problem and to develop solutions. In 2011, for instance, social psychologist Brian Nosek launched the Reproducibility Project, whose first assignment consisted of a collaboration of 270 contributing authors who sought to repeat 100 published experimental and correlational psychological studies. Their conclusion: only 36.1% of the studies could be replicated, and where they did replicate their effects were smaller than the initial studies effects, seemingly confirming Ioannidis’ findings.

The Reproducibility Project has subsequently moved on to examine the situation in cancer biology (with similar initial results). Meanwhile, a survey undertaken by Nature last year would appear to confirm that there is a serious problem.

Whatever the cause and extent of the reproducibility crisis, Nosek’s work soon attracted the attention of John Arnold, a former Enron trader who has committed a large chunk of his personal fortune to funding those working to – as Wired puts it – “fix science”. In 2013, Arnold awarded Nosek a $5.25 million grant to allow him and colleague Jeffrey Spies to found the Center for Open Science (COS).

COS is a non-profit organisation based in Charlottesville, Virginia. Its mission is to “increase openness, integrity, and reproducibility of scientific research”. To this end, it has developed a set of tools that enable researchers to make their work open and transparent throughout the research cycle. So they can register their initial hypotheses, maintain a public log of all the experiments they run, and the methods and workflows they use, and then post their data online. And the whole process can be made open for all to review.