This is a slightly expanded and more linear version of remarks I made during the ALA Midwinter 2016 Top Tech Trends panel in Boston on January 10, 2016. I think I covered all this, and probably more, but these are the main points as I see them. Please note — I speak only for myself, not for the University of Michigan Library or the University of Michigan.
I’m going to talk a bit about learning analytics — some might say “library analytics” — in the context of academic libraries and their parent institutions. I hope the themes I raise will be of broader interest than just to academic librarians in the audience; I think there is resonance for other libraries, as well. But first, I want to provide a bit of context from libraries more broadly about user privacy, rightly focused on ensuring that users’ privacy rights are protected and knowable.
There have been several initiatives; I’d like to bring two to the forefront. First is the recently-published Consensus Framework to Support Patron Privacy in Digital Library and Information Systems, the output of a Mellon-funded grant to NISO (the National Information Standards Organization) for “a series of community discussions on how libraries, publishers and information systems providers can build better privacy protection into their operations and the subsequent formulation of a framework document on the privacy of patron data in these systems.” This months-long process of virtual and in-person meetings of library and content providers provided a forum for participants to hash out a set of principles for ensuring that patron privacy is maintained through the collection of personal information related to information access. (I was a small part of these conversations.) The result of this voluminous series of conversations was a remarkably succinct set of 12 principles [PDF]. If you haven’t seen this document, take a look — it’s worth a read.
The other recent development is the “HTTPS Everywhere” initiative of the Electronic Frontier Foundation. This is a set of browser plugins (for Firefox, Android, Chrome, and Opera) that make sure that your browser is actually communicating via a secure, encrypted https method rather than the insecure http. It makes sure that if https is available (even if you don’t follow an https link, or the website has misconfigured something). Conceptually related to this is the “Let’s Encrypt” initiative, a free service to provide secure (SSL) certificates to web sites at no cost, so that the browser-to-server connection for more of the web will be secure from eavesdropping. Incidentally, this site’s ISP, Dreamhost, recently implemented Let’s Encrypt as a free add-on service and I’ve taken advantage of it.
With that as the landscape, I’d like to talk more about the sorts of things that can be done at an institutional level to better understand what factors make for successful student outcomes. This is what I mean by “learning analytics”: It’s the process of looking at all aspects of student life and trying to find the factors that indicate that a particular student is going to achieve a successful outcome (that is, graduation). And, conversely, what factors might indicate that things are going awry. And, most importantly, to be able to identify students who are not on the right path well before real problems hit. In short, what is it that makes a successful student a success, and how can those things be fostered through analysis and recommendations of individualized corrections?
Some campuses are embarking on large-scale efforts to implement learning analytics programs across the entire campus. My campus, the University of Michigan, is in this group; for more information, see the Learning Analytics pages within the office of Digital Education & Innovation. Within the University of Michigan’s understanding of learning analytics processes, the library has a great deal to contribute. And I, as a librarian and technologist, feel there is a tremendous amount to be learned to improve educational outcomes by examining closely the study, research, and interaction patterns of students so that we can improve the tools we offer and identify the specific areas that actually need special education, as opposed to improved interfaces.
Striking the balance between these two forces — users’ (and librarians’) broad desire to increase privacy — with institutions’ goals of improving the educational process and outcomes — will be the challenge. The trick will be to find ways to understand specific user behaviors and match them to broader data sets from other parts of the campus (which in many cases have less restrictive policies about data retention, with similarly lower expectations for privacy).
This is a two-edged sword; we (in academia) have a fantastic opportunity to prove our contribution to campus in a way that was all but impossible until recently. Being able to make these connections between students and faculty, resources, and outcomes of the educational process will be powerful, more for the campus to have proof positive of what we’ve been saying all along about the power of libraries. (At least, I hope that’s what we can say; that’s why we need to do this research.)
At the same time, we have a corresponding opportunity to help guide campuses in developing better information use and access policies for the data they are already collecting and using.