The rise of "Access brokers" - why now and a comparison with RA21
Update Aug 2020 - The main thrust of this response is to point out RA21 type solutions do not handle the appropriate copy problem. However as of 2019, we see that the publishers have come up with the GetFTR service which is an application of RA21/Seamless Access to handle this problem.
Last week Scholarly Kitchen's Kent Anderson published The New Plugins — What Goals Are the Access Solutions Pursuing? The long piece covered "positioning moves by entities with plugins designed to exploit Green OA or address access issues in a virtual, seemingly decentralized way — Unpaywall, Kopernio, and Anywhere Access." He compared it to "STM community’s effort to improve access, RA21" - (Unsure what is RA21?)
I think it's fair to say it's a controversial piece, with some errors in facts e.g. eyebrow raising privacy charges against Kopernio (some since then retracted) but for librarians the main area of interest is the charge that librarians are against RA21 because of "anti-publisher sentiment", while giving browser plugins a free pass on privacy issues.
On the same day (Kent Anderson seems to indicate this was pure coincidence), "RA21 Position Statement on Access broker" was released. "Access Broker" is a interesting term as it isn't well defined except as "tools and services have been launched which attempt to ease the problem of accessing scholarly information resources from outside institutional networks". The following were listed as access brokers
Kopernio from Clarivate,
Anywhere Access from Digital Science
Campus Activated Subscriber Access (CASA) from Google.
I suppose others not mentioned that would fall under this definition would include
Lean Library's Library Access browser extension
Open Access Button (the recently announced GetPDF extension will work with institution copies as well as Open Access)
Some "access brokers" as of April 2018 (prior to launch of Anywhere Access)
The statement is basically negative on these tools, claiming that these tools are merely band-aids and lead to privacy issues either by storing user's institutional credentials, creating accounts and generally monitoring pages visited. The implication here is RA21 is the real solution , that will solve the fundemental problems of access issues and "reduce the need for Access Broker tools and allow simple access to scholarly resources from anywhere, on any device, at any time, with no pre-configuration or additional software required."
I'm not qualified to go into the potential privacy issues of each of these browser plugins (they all have different implementations under the hood and I'm sure by the nature of how they work will definitely involve privacy tradeoffs), but in this post I would like to discuss the definition of "Access broker" and the idea that RA21 alone will solve the problems these "access brokers" are trying to solve.
Definition of Access broker
"Tools and services have been launched which attempt to ease the problem of accessing scholarly information resources from outside institutional networks" (as stated in RA21's statement) is a very broad definition.
For instance are library openurl resolvers considered as a "Access Broker"? I notice that Google's CASA is included, so I assume library links programme in Google Scholar which incorporates the library link resolvers in the Google Scholar results is that also an access broker?
Library link resolver links in Google Scholar, is this a access broker?
In fact, it's logical to wonder if is RA21/Federated Identity technology (most clearly demostrated by OpenAthens) is a Access Broker since they do help ease the problem of accessing scholarly resources.
The declaration by RA21 doesn't explict spell it out, but I suspect they do not think so.
So obviously the definition of an access broker needs tightening.
I would say that RA21 is different from all of the services I have listed above (perhaps argubly with the exception of CASA) is that it focuses on authenication of a user when the user is already on the publisher's website. To be fair, I could imagine everyone from Google Scholar to ResearchGate could use this technology to authenicate the user but chances are it will be mostly by the publishers.
I would guess even in a world where RA21 and federated identity solutions predominate, these tools still have a role. Perhaps one way to see this is to list the reasons why such tools and services have started popping up recently and see which of them will be no longer a reason in a world where we live in a RA 21 world and IP authentication is no longer a thing.
1. Users are no longer starting from the library homepage and less likely to be in campus when doing researching
This has always been the bane of researchers and even as a young librarian ten years ago in 2008, I grasped that this was a big barrier for users.
This led me to study and experiment with bookmarklets, library custom toolbars (including Libx, Conduit toolbars, opensearch plugins, greasemonkey scripts and more to learn ways to mitigate this.
The Libx toolbar from my blog post "Adding ezproxy to the url - 5 different methods" dated Apr 2009
While this is a compelling reason, I don't quite buy that this is the major reason why "Access brokers" emerged only recently. Even 10 years ago, the majority of researchers did not start from the library homepage and faced the same issues when researching off campus.
Still, in terms of RA21's mission, it is indeed true that a lot of the pain from not starting off the library homepage, or being in campus will be solved by RA21, if they don't need to figure out complicated ways to login as in the past (typically via embedding the ezproxy string or somehow automatically redirecting to library systems).
2. Difficulty of discovery of open access (particularly Green OA from Institutional repositories and to lesser degree hybrid articles)
One way to see the rise of these browser extensions is to see them as a response to the poor discoverability of OA material , in particular Green OA (and even more specifically institutional repositories).
For a long while Google Scholar was the main and only reliable way to discover OA and most library and publisher tools and databases either ignored them or paid lip service to them at best.
Tools like Unpaywall, Open Access button started the trend of making OA papers more discoverable and many browser extensions built on these and made the next logical step by also supporting finding of and access to subscribed institutional versions. These include Anywhere Access, Kopernio, GetPDF (from OAB), Lean library library access etc
Of course these days things are changing. Librarians are adding access to open access via link resolvers (or less frequently by directly adding databases that have article level links to OA from say 1Science's 1Findr), Web Scale Discovery services like Primo and Summon are starting to include open access content from Unpaywall, BASE and CORE, while major databases like Scopus, Web of Science, EuropePMC are starting to add links to OA material.
Improve #Discovery of #OpenAccess Content. Can @ExLibrisGroup confirm that @BASEsearch content will be added to PCI and Summon? #igelu2018 #OA pic.twitter.com/QOZUgmknu7
— François Renaville (@f_renaville) August 21, 2018
Assuming we continue to live in a hybrid world where there is both large amounts of OA (including varying versions from preprints , accepted versions to published versions) and subscribed versions, this is a need that RA21 alone can't help.
Imagine a user searches for an article and lands on a ScienceDirect article. But he doesn't have access to the subscribed version. But a Green OA - accepted version is available in a IR (Institutional repository). Would RA21 solve this problem alone?
I would guess not. I can see two possibilities that might work. First, he would login and the library link resolver on the ScienceDirect page (assuming there is one) would direct him to the OA copy. That is in a world where all library link resolvers have OA discovery capabilities.
Alternatively Sciencedirect itself embeds a link to the Green OA version, perhaps via OA discovery tools like Unpaywall. But would publishers do this? Possibly not given that Elsevier has decided not to link to Green OA with unpaywall in Scopus despite it's availability in Unpaywall dataset.
The thing to note is that RA21 alone does not solve the traditional appropriate copy problem. In other words, if a user googled and lands on Sciencedirect, even after logging in he may still not have access as he may have access via JSTOR or another aggregator like Proquest.
In a way pointing to different OA versions is a new special case of this old problem. But let's consider library link resolvers which are designed to solve this.
3. Poor user experience of Library Link resolvers
The poor user experience of library resolvers, which were initially Openurl link resolvers is a well known problem ever since they started to be used in the early 2000s and was formalized as NISO standard in 2005.
Part of it was simply the instability of linking using the Openurl protocol. As any librarian tasked to trouble shoot broken links for OpenURL link resolvers and discovery services can atest, the links can seem to break at the drop of a hat and it's not just a selection effect caused by minority of users only reporting broken links. Studies have shown results as bad as 30% errors(particularly for non-article content) for a variety of reasons (though poor or none-matching metadata in Knowledge base and article indexes is typical the biggest issue).
The library world responded in the 2010s with projects like IOTA (Improving OpenURLs Through Analytics) and KBART (to improve KB standards) but it's unclear to me how much this helped.
That said, there have been improvements in library link resolvers and today's link resolvers are not purely OpenURL based. The wide use of dois (which usually work more for publishers than aggregators) and direct linking methods such as Summon's IEDL (which often require partnerships with the full text provider) have helped reduced broken links.
Beyond this though, Library link resolvers also have a user experience problem in the way there are implemented and some libraries have attempted to improve on this.
For example, library link resolvers often lead you to a immediate library link resolver page with multiple possible sources and studies have shown many users get stuck on that intermediate page. Many link resolvers provide a way to bypass this page by going directly if there is only one source, or always go to the first ordered source (so called 1 click linking), but because of the unreliable linking, one school of thought is it is always better to link the immediately resolver page first. Admittedly workaround for this exist like in 360link iframes sidebar helpers which have problems of their own.
Another problem which I have with link resolvers has nothing to do with the link resolvers themselves but with how they are displayed in search results.
Take Scopus, as it works now, the library link resolver link will appear next to every result whether the user has access to it. The user has to click first and see if he had access to it (also known as click and pray).
Library link resolver links appear next to each result in Scopus, whether the user has access to it or not.
This has led in my experience to some users getting confused and librarians have attempted different text labels to indicate that the link is just a try at full text and not a guarantee.
In my opinion Google Scholar and Pubmed (Link out) are the only two (and now Dimensions with Anywhere Access) to handle this the right way by only showing the library link resolver link if there is full text.
This is doable only because both these services are sent the holdings of the library monthly so it knows which articles are available to the user without the user clicking on it.
The library link resolver link appears only if the user has access to it, as opposed to all the time
I'm unsure why Google Scholar and Pubmed are the only discovery services that implement link resolvers that way. One possibility is that despite the much better UX, it takes too much effort to send each library's holdings to all possible search partners. Another thought occurs to me, would libraries want to send their full holdings to partners like Elsevier which might use the data for competitive reasons. Might there be a cleverer technological method that avoids the need to send all your holdings?
As a sidenote, with databases like Web of Science and Scopus adding their own links to open access material, I suspect many users will just click on the link to the OA version rather then click to the library link resolver link which might yield nothing even though the later might have the same OA version (assuming OA capabilities are added) or even a better published version.
As a user would you click on the "Free accepted article" that always works or the Findit@SMU Libraries' link that may or may not yield an article?
Beyond all this, there's another distinguishing fact of some of these new "Access brokers". They capitalise on the fact that library link resolvers generally don't link directly to the PDF but to the article landing page. Let's look at their marketing material and tag lines
Kopernio - "Fast, one-click access to millions of research papers."
Anywhere Access - "Nothing is simpler than 1-click PDF Access"
Kopernio tag line - "Fast, one-click access to millions of research papers."
Anywhere Access tag line - "Nothing is simpler than 1-click PDF Access"
This may sound like a small detail but in practice once you have used this, you will be wondering this isn't the default. While I have seen on Twitter some researchers argue they prefer to link to the article landing page rather than directly to the PDF, I suspect they are in the minority.
To summarize these access brokers improve on library link resolvers in the following ways
More reliable direct linking (probably doi based, abandoning of openurl linking)
Direct 1-click PDF access
ability to be triggered on almost any page, not only on sites that implement library link resolvers
The last is important. While RA21 has non-publishers included, I think it is fair to say it is publisher lead. Would that lead to some blind spots? RA21 definitely doesn't solve the appropriate copy problem. Think of the following scenario, the user is on a blog post, say mine and I point to a article with a doi.
Scenario 1 - User clicks on doi, lands on publisher page. He logins using Openathens (RA 21 does it's magic and he doesn't even need to choose an institution or type in his password). Finds he still doesn't have access to it from the publisher site. Hopefully there is a library link resolver. Clicks on it, hopefully it works correctly and sends him to a aggregator site where he has access. On the aggregator's article landing page, he clicks and he finally gets to the PDF.
That's at least 4 clicks by my count.
Scenario 2 - Browser extensions detects dois, detects he has access via an aggregator and with a click goes straight to the pdf.
To be fair, aggregator direct linking in many of these tools are still currently being developed so scenario 2 might not be possible for many yet.
As a sidenote not all the "Access brokers" offer one click to pdf. Interestingly Browzine has recently offered Libkey, a API that offers one click to PDF directly (taking into account your institution holdings of course)
"LibKey is based on the years of experience Third Iron has in managing all of the complexities required to connect users to articles, including journal entitlements, authentication and linking. The result is an almost magical user experience of clicking on a link and getting immediately to a PDF. No link resolver, choosing sources, navigating platforms or any other aggravating steps that can derail your researchers."
Currently, you can embed it into your discovery service like Summon, Primo etc.
Embed Browzine's Libkey into Primo for one click to PDF
I don't think a browser extention like Kopernio exists for Libkey, but it does seem possible to do with the API available.
4. Competition from Sci-hub
One of the more interesting findings from analysis of the released Sci-hub logs was that a lot of the access was coming from ip ranges belonging to large university systems. The implication (see this analysis of downloads from Dutch University ip ranges) is that many users are using Sci-hub not because they have no access via their institutions but they find Scihub much easier to use for the reasons already mentioned above.
This is of course the major reason why there is this big interest in RA21 by publishers now, despite the fact that many of the issues above are decades old.
Kent Anderson writes "More to the point, what if someone were to make a plugin as simple and effective as Unpaywall or Kopernio or LibKey or Anywhere Access, and point this at Sci-Hub? Would users defect to this? Would libraries approve? Would publishers sue? The answer to all three may be affirmative."
I hate to burst his bubble, this already exists, I have seen no less than 3 versions of this, but I shall refrain from linking to it.
Would RA21 alone solve this issue? My suspicion is no. As mentioned above even with federated ID solutions, Sci-hub would still be more convenient for the reasons I stated above.
Why click on a doi, that leads to a publisher page, login (pray you have a copy via your institution) , then click on a library link resolver (pray again) which may or may not lead you to a appropriate copy?
Kopernio and other access brokers while not perfect would work more closer to the way Sci-hub plugins work.
Of course nothing beats Sci-hub which does not need to worry about authenication and will almost always have access but with the Kopernio component for handling the appropriate copy problem, RA21 alone isn't sufficient.
Conclusion
As always this has been a long winded think out loud piece to try to clarify in mind the position of RA21 via the latest crop of browser extensions. In general it looks to me that RA21 is just a authentication system and while it helps with quite a lot of pain points (particularly if users are constantly hitting the publisher pages and have access there), it doesn't quite replace completely what RA21 calls Access Brokers (which themselves can work with Ezproxy, SAML based methods etc) as they provide some complimentary access features mostly around discovery and access to the copy appropriate to them.