Skip to content

What are Learning Analytics?

Last week, I announced an upcoming conference in Banff on Learning Analytics (call for papers can be found here). We have also set up a Google Group for discussions about learning analytics and the conference.

What are Learning Analytics? (LA)

Learning analytics is the use of intelligent data, learner-produced data, and analysis models to discover information and social connections, and to predict and advise on learning. EDUCAUSE’s Next Generation learning initiative offers a slightly different definition “the use of data and models to predict student progress and performance, and the ability to act on that information”. Their definition is cleaner than the one I offer, but, as I’ll detail below, is intended to work within the existing educational system, rather than to modify it. I’m interested in how learning analytics can restructure the process of teaching, learning, and administration.

LA relies on some of the concepts employed in web analysis, through tools like Google Analytics, as well as those involved in data mining (see educational data mining). These analytic approaches try to make sense of learner activity (through clicks, attention/focus heat maps, social network analysis, recommender systems, and so on). Learning analytics is broader, however, in that it is concerned not only with analytics but also with action, curriculum mapping, personalization and adaptation, prediction, intervention, and competency determination.

What does this look like?

Here is my vision of learning analytics and the role they *can* play in education (while simultaneously demonstrating the paucity of my visual/image creation skills).

This is the process of learning analytics:

Learners constantly off-put data – sometimes explicitly in the form of a tweet, facebook update, logging into a learning management system, or blogpost, and other times unintentionally while in the course of daily affairs (or data that is provided by someone else – such as being tagged in Facebook or with the new privacy-pushing Facebook Places). This data happily sits in a database, waiting for some type of analysis. The data that learners off-put is supplemented by the learner’s profile. My profile, for example, is all over the internet in all kinds of formats and sites: Linkedin, Facebook, Elgg, Google, blogs, institutional services, and others. Facebook (with Connect), Twitter, and Google are trying to reduce the confusion by becoming the default sign-in service on other sites. To date, online identity and profiles are a mess.

With data becoming increasingly intelligent (semantic or linked data), learner data, profile information, and curricular data can be brought together in some form of analysis. I’ll return to the curricular implications of this in a moment. For now, it’s sufficient to state that our data trails and profile, in relation to existing curriculum, can be analyzed and then used as a basis for prediction, intervention, personalization, and adaptation. I’ve addressed this process partly in Technologically Externalized Knowledge and Learning. It’s important to emphasize that the adaptation is not exclusively technological – sensemaking and wayfinding through social systems have demonstrated their value over the last several years through recommender systems, small network clusters, and so on. Adaptation and personalization needs to be holistic and multi-faceted, incorporating technology, socialization, and pedagogy.

A simplified image of the process looks like this:

Effective utilization of learning analytics can help schools and universities to pick up on signals that indicate difficulties with learner performance. Just as individuals communicate social intentions through signals well before they actually “think” they make a decision, learners signal success/failure in the learning process through reduced time on task, language of frustration (in LMS forums), long lag periods between logins, and lack of direct engagement with other learners or instructors.

This final image is fairly simple, but it gets to the heart of where I think learning analytics can help to transform education:

Curriculum in schools and higher education is generally pre-planned. Designers create course content, interaction, and support resources well before any learner arrives in a course (online or on campus). This is an “efficient learner hypothesis” (ELF) – the assertion that learners are at roughly the same stage when they start a course and that they progress at roughly the same pace. Any educator knows that this is not true and will eagerly resist the assertion that their teaching assumes ELF. But systems don’t lie. How educational institutions design learning is due for dramatic restructuring – the model of curriculum design, development, and delivery currently employed by schools, corporations, and universities is strongly antagonistic to what actually works in the learning process (even a brief review of learning sciences reveals the failure of ELF).

Learning content should be more like computation – a real-time rendering of learning resources and social suggestions based on the profile of a learner, her conceptual understanding of a subject, and her previous experience. Competence (as measured by a degree or certificate) need not be explicitly pursued. For example, an integrated learning system should be able to track my physical and online interactions, analyze my skills and competencies, and then compare my life-long skills against a discipline or field of knowledge (this comparison will be possible because a discipline will utilize intelligent/semantic/linked data to define its knowledge). Then, the learning system should inform me that I am “64% to a achieving a phd in psychology, 92% to achieving a masters in science, 100% to achieving a certificate in online learning” and so on. If I do decide to pursue that phd in psychology, the learning system should offer a personalized path forward that adapts constantly to knowledge I acquire in the course of work, parenting, or generally living my life.

I recognize that a few of these assertions are a bit future focused. However, I think all of them are showing early signs of promise in advanced data analysis, prediction and probabilistic models, learning research, intelligent data, and social learning theory. Given the painfully slow process of change in education or learning institutions, now would be an ideal time for institutions to start asking questions about the structural impact of networked technologies on the future of learning and knowledge-making in their systems. I’m convinced that education tomorrow will look far more like the model detailed above than what we currently see in traditional institutions.

2 Comments

  1. When the method for analysis is usable, it will compliment existing processes inside formal education, such as Recognition of Prior Learning (RPL), or assessment of prior learning (APL) that theoretically help people find accelerated pathways through curriculum. This really only works where assessment is standardised, such as the national unit standards used in Australia or New Zealand. HE resist this standardisation, making RPL and APL impossibly inefficient.

    The ultimate would be a method that assists people to do their own LA and develop evidence for arguing for more precise RPL or APL.

    I imagine this tool as a browser add on, something that could collect, track and save communicative and informative activity.

    Wednesday, August 25, 2010 at 2:53 pm | Permalink
  2. Ken Anderson wrote:

    So,

    Is learning analytics essentially a data-mining (and analysis etc) of students extra-school learning/activity, akin to data-mining by web social networks? Or would it only be data-mining of the info that LMS’s gather? Could a student choose what extra-school data would be mined? Your diagram shows a wide source for the data. What are the privacy implications of that?

    Thursday, August 26, 2010 at 6:02 pm | Permalink