Architecture of Authority

Below is the working text of my talk to ACRL/NY’s symposium on Money and Power, along with more nuanced answers to questions and comments I received.

My ongoing thanks to the organizers, attendees, and many people who work behind the scenes to make events like this come together. Slides are available in Google Drive.

I’m a systems librarian. At Geneseo I’m one of two faculty lines outside reference and instruction. This informs a great deal of my thinking about the library and how it works. This is the end of my first year or so as a librarian. Nobody gets here alone, so here’s a partial list of people who influence, support, or encourage my thoughts today.

One of the most powerful ways vendors have influenced the academic library environment is through the systematic relocation of core values from libraries to vendors via our software. Libraries articulate those values–often at the expense of our salaries or better benefits–but the execution of our values manifest in library systems developed by third parties with competing interests.

Let me be clear, because there are vendors in the room: my job is to connect people with information. I make technology indistinguishable from magic for my users. Your job as a vendor is in fact to generate money. Although you’ll change the framework of how you want to accomplish this, the end game is the same. That’s not a bad thing, it’s an entirely sensible goal under capitalism. But it’s ultimately an agenda which competes with mine.

Before I get into how systems impact values, what do we claim libraries value? I’ll take a couple from ALA’s Core Values statement: privacy and access. I am choosing to ignore the value ‘Democracy’ because apparently that’s up for grabs even if we are calling it a ‘draft’.

The data we provide companies isn’t abstract. It’s rich and deep and most of all, even at smaller institutions it’s huge. Companies like Ex Libris know more about our users than we ever will, and we’re paying for the privilege of handing over that information. In many cases we’re paying unsustainable percentages of our resource budget to do two things:

1. Hand a third party with zero realistic accountability a rich data set that usually compromises user privacy.
2. Spend the rest of the year fighting with that system to function in a useful way for our patrons.

For all the data they have about our users, you’d think we’d have a beautifully customized product. It would perform and require very little instruction—if nothing else because I think reference and instruction librarians get tired of teaching interfaces which aren’t performing instead of the countless other things my colleagues do.

The other value I want to pick at is access. I interpret access to mean both straightforward access to information and the mechanism which provides that access, because we can’t uncouple these anymore. Specifically, let’s talk a little about Summon, a discovery layer like Primo or EBSCO Discovery Service, and a little about MARC records at Geneseo.

Summon aesthetically looks like Google. It’s a single search box in an empty field, which would be one thing if Summon behaved like Google. But it doesn’t. Summon acts more like Amazon once you initiate a search, and like Amazon, prioritizes the results from certain content over others. It’s in Ex Libris’ best interest to prioritize ProQuest databases, just like EDS prioritizes EBSCO products. Everything else is up to the underlying metadata. And, to get away from exclusively vendor criticism for a moment, I realized in getting this screenshot that we have some re-branding to do. Research isn’t easy. Here, the library has gone out of its way to say so. Failed searches then, demean the user.

Privacy isn’t determined by the library so much as it’s decided by the developers of systems we use. Take the example of Adobe Digital Editions in 2014. Adobe was gathering data on the eBooks you have, what pages you read in those eBooks, so not only did Adobe know you read Fifty Shades of Gray, it knew which pages. It knew you only read enough of Capitalism to impress someone at a party.

If you’re in the business of eBooks this is great data to have. Adobe’s problem is that all this information was being transmitted in clear text to their servers, which if you were a library using Digital Editions for your eBooks is a big problem. Adobe resolved the issue but things like this happen with distressing frequency.

I can value privacy all I want. My professional organization can say this too, but it’s workflows and systems that determine how accurate this is for my patrons. This for example, is the patron-facing record for Simone Browne’s Dark Matters: on the surveillance of blackness.

This is what happens when you click on the MARC record view. Now, it doesn’t matter what the intention behind this data collection is. And it doesn’t matter that this isn’t the most discoverable thing about me–yet. It turns out this was done entirely without malice, but it’s a vestigial workflow. Intent doesn’t absolve me of the potential harm here.

I found this because I am preparing for a trial of EDS. This requires a dive into how we use MARC. I’m willing to bet I could have asked EBSCO to index the 970 field, make it searchable.

On the subject of data I’m going to pick on Alma in particular. Not only does Alma’s parent company know more about our patrons than we do, but Alma analytics choose how libraries see and explore that data. Alma’s only going to show me what it feels is valuable to me as a user.

Alma defines users through roles and while you can get fairly granular with what individual users are able to do, there’s still going to be templates available to me, of Alma’s idea of what Systems Librarians do or what Catalogers do, when our daily tasks—those things we do to maintain those systems—are going to be very different between libraries. Alma Analytics functions the same way: I can create almost any report I want with the data Alma provides. We love assessment. We adore analytics. As a result, the more data you feed Alma, the more powerful it becomes. Analytics requires you to go all in, even though realistically this isn’t a solution for every library. Ex Libris representatives are routinely astonished that I don’t need my ILS to do everything.

One of the reasons I think Alma is increasingly adopted is because it makes provost candy easy to generate. But how many of the people we hand those analytics to for decision making are literate in the ways we need them to be to understand data visualizations, to give us FTE, a bigger collections budget, pay for interns?

If I don’t know how to do those things and want to learn, I can go to the community zone where Ex Libris will post the free labor they get from librarians paying to make their system better. This isn’t unique to Ex Libris: SpringShare has a community lounge and SpringyCamp. And those things aren’t bad: I use them all the time. SpringShare by the way, has recognized this and made their platforms free to library and information science program students.

While I admire SpringShare for this, it perfectly expresses the problem: vendors now have so much power in libraries their products show up in our job descriptions, as though systems don’t migrate or change. Vendors—and how accessible they make their training—influence who gets hired. Even the job I have now listed “experience with Intota” as a preferred qualification. Only a handful of librarians in the world have this training.

This image is from Matthew Reidsma’s work Algorithmic Bias in Discovery Systems, which is specifically about searches in Summon matching with the Topic Explorer to Wikipedia. The Topic Explorer works by suggesting encyclopedia articles from Gale or Wiki to match whatever I’ve searched for. Rather than use Wikipedia’s API for this, which would more cleanly match to articles, Summon has instead decided to add Wikipedia to the Summon index.

ProQuest had suppressed the inappropriate result–in this case, equating a search for ‘mental illness’ with the Wikipedia entry on the book ‘The Myth of Mental Illness’. Reidsma went on to customize Grand Valley State University’s instance of Summon to include the ability to report off-topic results. I like Matthew. I deeply admire his work. But this solution means his users are doing work ProQuest should already be doing, or rather, wouldn’t have to be doing if they’d use the API instead of the Summon index.

But I think we’re one of the only industries that pays for the privilege of improving products, just to get them to work the way we needed them to in the first place. In exchange for that privilege, vendors get the benefit and market those improvements back to us. Systems librarians hear “our current environment doesn’t support that” from vendors only to have a change the community requests roll out a few months later, I suspect after enough of us say we’re reconsidering a subscription.

Our trust issues with vendors are justified. Wiley uses trap DOIs to prevent automated downloading. In the process of implementing what is frankly an inelegant solution, Wiley cut off access to dozens of users, including Open Access articles. 1. Their engineers are better than that and 2. I don’t know how legal it is. Eric Hellman talks about this more in detail at his blog. Every time we give Wiley money, we’re saying this kind of solution is acceptable. During Geneseo’s last renewal cycle, I paid much closer attention to Wiley subscriptions because of this. They’ve lost money.

We didn’t develop an ILL module in Alma because we didn’t think ILL would exist in 4 years.
–Executive, Ex Libris

This quote above was astonishing news to everyone working in resource sharing. How can I trust any timeline this vendor gives me, if this is the kind of thinking that informs their development decisions? This same executive said if our link resolver wasn’t performing, we could switch to something else.

Toward Reclaiming Agency

I ended the talk with a few ideas for reclaiming agency.

• My budget doesn’t increase 6% yearly, why should theirs? How is that sustainable? If vendors need help, there’s a number of people in the room right now who know how to do more with less. I’m sure we’d be glad to consult for you.
• Ask for better contracts. EBSCO’s SLA with SUNY has 96.9% uptime. This sounds impressive until you realize it means they can be down for 11 unscheduled days before they breach their service level agreement. EBSCO also sent me an agreement to create the Open SUNY Textbooks database with a non-disclosure agreement. This is not EBSCO’s real intent, but it’s malice by boilerplate.
• Don’t reward poor execution of products with free labor and additional contracts.
• Encourage curious people to study things and give them permission to explore.
• Answer all curious people. Listen to who gets told “that’s not your job” versus who is allowed to know the full answer to questions.
• “Pay your interns” means a lot of things here. This could mean a salary or it could be a travel award to help them network. And, if you consider how much you’re paying for ineffective systems, it might be time to invest your money elsewhere.
• Advocate for maintainers and better documentation. I see you, open source software advocates, reading this text and waiving your hands like the student who knows the answer if only you’d just call on me. Your documentation needs to be better. When you write grants you need to include community managers and technical writers as part of the projects.

I want to end my talk with a modified version of this quote from Cody Hanson, Director of Web Development at Minnesota University Libraries:

We must face the fact that when we relinquish control over the software we provide, we effectively relinquish control of our most visible and effective public values.

//

At the end of the panel a number of questions came from both librarians and vendors. One, a woman who works for EBSCO/YBP claimed libraries do have all the power because without us, vendors don’t exist. My answer, “libraries don’t have any power in that equation” wasn’t as nuanced as it could have been. While libraries order print materials from many different sources I suspect YBP is so deeply embedded in local workflows that it would be difficult to extract acquisitions from that vendor. It’s a problem of scale, perception, and codependency.

I use the quote from a meeting with Ex Libris not because it is ‘wrong’ per se, though it is breathtaking in its misunderstanding of ILL. Resource sharing is changing, and how we analyze the landscape of resource sharing needs to change as well. ILL is not three columns of borrowing, lending, and document delivery numbers. Resource sharing is best understood through network analysis; less via Excel and more through d3.js or R because it is an expression of relationships between libraries. To claim “ILL won’t exist in X years” is both simplistic and a denial of those relationships. Worse, it negates the library’s ability to be flexible, re-skill, and change. What this claim really means is “ILL as I understand it will not exist in X years,” and you’d be correct.