Rethinking institutional repositories.
More recently, Eric Van de Velde seems more than ready to bury them. His July 24, 2016, post, "Let IR RIP," in the SciTechSociety blog calls IRs "obsolete." He urges his readers to phase them out and consider other alternatives to support the green road to open access (scitechsociety.blogspot.sg/2016/07/let-it-rip.html).
As recently as November 2016, George Macgregor, institutional repository coordinator, University of Strathclyde, wrote that such rumblings are not new and have been expressed in person at open access conferences. Yet at Strathclyde, IRs still experience a high volume of deposited full text and digital objects. Macgregor traces various criticisms of IRs and notes that changing academic behavior "takes a long time." He concludes that "IRs need to remain a principal mechanism for achieving Open Access, whether we like it or not" ("The Long Read: Why Do Institutional Repositories Remain One of the Only Viable Options for Green Open Access?"; strathoa.tumblr.com/post/152680188170/the-long-read-why-do-institutional).
Richard Poynder, in his September 2016 commentary that preceded his interview with Coalition for Network Information CEO Cliff Lynch, argues that while open access seems inevitable, he sees a "growing sense that green OA has lost its way" (richardpoynder.co.uk/Clifford_Lynch.pdf). He describes most IRs as full of entries with no full text, only bibliometric details. This is despite a flood of mandates from institutions and funders that apparently do not motivate researchers to self-deposit with institutional repositories.
As a librarian on the ground who has a keen interest in the area and who interacts with faculty and students regularly, I must say such pessimism is not entirely unfounded.
Some of the obvious reasons why researchers are reluctant to deposit their work in an IR include ignorance of the existence of the IR and of their rights to self-archive, along with a lack of motivation because the current academic structure does not provide incentives for making papers open access. As someone who has worked both in institutions with and without mandates, I can affirm that, in general, while open access mandates may help (depending on the type of mandate), they aren't a silver bullet in getting researchers to voluntarily submit their work into institutional repositories.
What I find more troubling is the rise of a group of researchers who are actually ready and willing to self-archive but choose to do it elsewhere rather than in the IR.
Increasingly, we see researchers who are motivated to self-archive their papers in both subject repositories or preprint repositories such as arXiv (arxiv.org) or SSRN (ssrn.com) or in so-called scholarly collaboration networks (SCNs)/social sharing networks such as ResearchGate (researchgate.net), Academia.edu, and perhaps Mendeley (mendeley.com). Yet these same researchers are stubbornly lukewarm toward IRs.
I believe it's instructive to consider why researchers are choosing such sites to archive their papers rather than IRs despite the obvious drawbacks of the other sites. Why not IRs? Here are some reasons.
LACK OF PERMANENT OWNERSHIP
Many, if not most, researchers tend to change institutions at least once in their careers. This impermanence of institutional affiliation leads to a lack of ownership, particularly if you include a researcher's time as a Ph.D. student. The main attraction of creating profiles or accounts at subject repositories such as SSRN or in SCNs such as ResearchGate is that researchers will always have control of that account and control of the papers they deposit in those venues, even if they change institutions. It's no surprise that researchers tend to have a sense of ownership regarding such accounts.
While ORCID (orcid.org) is posed to eventually diminish the impact of such issues by allowing researchers to own one unique author identifier throughout their careers while pushing information to various research profiles, the full text of the paper has to sit somewhere.
Researchers who put their papers in an IR will eventually lose direct control of those papers when they leave the institution, making it difficult for authors to easily edit or make changes to their papers. By putting all their papers in one central source, authors retain control.
Authors can also obtain aggregated usage statistics in one place, as compared to having usage statistics spread around in various IRs, which will be unwieldy to aggregate. This is assuming you could even aggregate statistics that are not standardized across various repositories.
The flip side of the fact that institutional repositories are often not permanent is that subject/discipline affiliations most likely are stable. Researchers may move from one institution to another, but if they are researching in a particular discipline, such as history or physics, they will probably continue in that subject area.
Subject repositories have the advantage of greater familiarity to scholars and can have systems custom-built for each researcher's community. Subject repositories and/or preprint servers have an advantage because researchers tend to think along disciplinary lines. In many disciplines, there is already a tradition of putting up preprints prior to publication.
By putting papers in subject repositories such as SSRN, researchers can benchmark their paper against their peers in the same discipline, something that is not possible in IRs.
Given the central mass of disciplinary-appropriate eyeballs already there, it's no surprise that IRs tend to lose out to subject repositories in terms of interest. As the saying goes, "Out of sight is out of mind."
INCONSISTENCIES AND POOR USER EXPERIENCES
Institutional repositories tend to offer poor user experience. It's fairly well known that, compared to SCNs, most institutional repositories lag behind in functionality and sophistication. For example, until recently, most institutional repositories did not automatically pull metadata to ease the task of entering bibliometric data, nor did they automatically do checks on SHERPA/RoMEO (sherpa.ac.uk/romeo) and send out emails to researchers to inform them that a paper they published could be self-archived.
In comparison, ResearchGate and Academia.edu are constantly innovating. Although many people find them very spammy and intrusive, I think they do at least try to use the latest known gamification and social networking techniques to encourage use. ResearchGate, for example, can tell you who viewed your record, who downloaded and read your paper (if the person was signed on while doing so), and you can even respond to such information by asking the identified readers for a review.
Not everyone considers such features to be positive, but the point here is that the features are iterating much quicker on these sites than on the average institutional repository.
The lack of new innovative features may not be the only reason for the failure of IRs. It is also the lack of consistency between institutional repositories. While most university IRs are using a relatively small set of common software--Digital Commons (digitalcommons.bepress.com), DSpace (dspace.org), or EPrints (eprints.org)--they can vary greatly depending on the customization and feature sets. For the already time-strapped researchers, learning to come to grips with a new system (with different submission formats, interface, and requirements) whenever they change institutions seems to be too much work, particularly when there are alternatives.
In fact, scholarly communication librarians have given up trying to get researchers to submit papers on their own and have gone with the mediated deposit model, in which librarians upload papers on their researchers' behalf. Many librarians also trawl the web looking for other papers archived by their researchers at other sites such as subject repositories or ResearchGate. But is that going to scale when repositories have poor interoperability and traditionally have been built to support individual researcher uploads and not bulk uploading?
THE QUESTION OF MASS
In the last 10 years, we have learned that having mass on the web is important and that network effects tend to dominate. This results in giants such as Facebook being almost impossible to dislodge, even with titanic efforts from companies such as Google. Facebook has become too entrenched due to network effects.
Will ResearchGate and its peers that aim to be the academia equivalents of Facebook succeed using the same centralized, walled garden strategy? We know that many of the social and networking aspects that ResearchGate and, to a lesser extent, subject repositories such as PubMed and SSRN bring are nearly impossible to replicate on isolated siloed IRs.
We know that IRs today are not destinations for visitors--most visitors discover papers on our repositories via discovery search engines such as Google Scholar, which link them directly to the PDF. Except for the few brave souls who submit papers, very few see the actual repository software pages. This in itself isn't an issue if the aim is just open access; however, it does prevent the social network effect from occurring, since it seems the first step of getting researchers to care about depositing papers in your site is to get them to come to your site in the first place!
One way to counter the lack of mass of individual systems is to allow aggregation of each IR. While aggregator systems such as CORE (core.ac.uk), BASE (Bielefeld Academic Search Engine; base-search.net), and SHARE (share-research.org) exist to attempt to aggregate all data into a centralized repository, the lack of standardization among repositories in terms of consistency of metadata makes the whole aggregator process a little pointless, particularly with outdated protocols such as OAI-PMH (Open Archives Initiative Protocol for Metadata Harvesting).
This is also where the earlier weaknesses of the lack of consistency and standards among repositories come into play. It's not just surface usability and features that differ between one institutional repository and another; it's also the limitation of almost no standards for metadata and content. It becomes a mishmash of formats when you try to search across them using aggregator systems.
Even something as simple as identifying whether an entry harvested via OAI-PMH has full text attached becomes a nightmare. Incidentally, this is a reason why most libraries using web scale discovery systems often do not include IR contents from outside their university. Many of the contents in IRs are, in fact, indexed in web scale discovery systems, including Summon (proquest.com/products-services/The-Summon-Service.html), Primo (ex librisgroup.com/category/ PrimoOver view), and EBSCO Discovery Service (ebscohost.com/discovery). However, most discovery system managers prefer not to turn them on because this results in many items surfacing that can't be reliably marked as full text, which leads to a lot of confusion.
The exception to this rule is the Digital Commons network by bepress (digitalcommons.bepress.com), which shows what can be achieved by ensuring repositories have a constant set of standards. By using the cloud-based digital commons repository, and assuming you keep the recommended subject scheme, you can easily compare the usage of items on your repository versus other repositories on the same network in the same discipline. It also has no issue detecting which items are full text and which aren't.
For an example of how the lack of consistency hurts, I was studying oaDOI (oadoi.org), a nifty new service that allows you to feed it a digital object identifier (DOI) to see if a postprint version exists on a repository by checking the BASE aggregator (among other sources).
While this works fine in theory, I've found it can fail for various reasons such as the repository not assigning DOIs to postprints. In other cases, it simply does not expose the DOI to the BASE harvester. A consistent standard would help greatly here.
THE RISE OF REPOSITORIES 2.0?
IRs have a natural advantage over centralized silos in that they are less easily taken over or disrupted. The recent purchase (May 2016) of SSRN by Elsevier is a good example of the vulnerability of centralized repositories. But beyond that, there are of course defenders of repositories who rally the repository crowd by mooting the idea of next-generation repositories that overcome the weaknesses I mentioned earlier. Chief among them is Kathleen Shearer, executive director of the Confederation of Open Access Repositories (COAR; coar-repositories.org). While admitting that repositories haven't been as successful as originally hoped, Shearer believes the answer is to work on the flaws of repositories to improve on them and not to give up (coar-repositories.org/news-media/more-on-the-future-of-repositories-responseto-richard-poynder).
COAR has launched various initiatives and working groups to address many of the IR-related issues, including working on guidelines for repository interoperability, standardizing Controlled Vocabularies for Repository Assets, and studying metrics such as usage. Coupled with work on new protocols to replace the aging OAI-PMH standard and discussions into value-added services that repositories could serve beyond being just repositories of content, this is what I believe constitutes the next-generation repositories.
The hope is that such next-generation repositories will be interoperable, and serve as knowledge nodes in the scholarly system that can be seamlessly aggregated to counter the mass of centralized repositories. Already we see the rise of regional repository networks such as LA Referencia (lareferencia.info) and OpenAIRE (openaire.eu) leading the way.
Will these initiatives take off? Only time will tell. But I hope they do.
by Aaron Tay
Aaron Tay (email@example.com) is library analytics manager, Singapore Management University.
Comments? Email the editor-in-chief (firstname.lastname@example.org).
Reasons Why Researchers Avoid IRs
Institutional affiliations are not permanent, leading to lack of ownership.
Subject/discipline affiliations are stable, and subject repositories might be a more natural level of aggregation.
Institutional repositories tend to offer poor user experience and are not consistent with each other.
|Printer friendly Cite/link Email Feedback|
|Date:||Mar 1, 2017|
|Previous Article:||Search engine update.|
|Next Article:||Online resources to creatively teach career research.|