A couple of weeks ago, I had the pleasure of attending the xAPI Base Camp, to present on content strategy. While I was there, I remembered that I have some colleagues who don’t see the connection between xAPI and learning. And it occurred to me that I hadn’t seen a good diagram that helped explain how this all worked. So I asked and was confirmed in my suspicion. And, of course, I had to take a stab at it.
What I was trying to capture was how xAPI tracked activity, and that could then be used for insight. I think one of the problems people have is that they think xAPI is a solution all in itself, but it is just a syntax for reporting.
So when A might demonstrate a capability at a particular level, say at the end of learning, or by affirmation from a coach or mentor, that gets recorded in a Learning Record Store. We can see that A and B demonstrated it, and C demonstrated a different level of capability (it could also be that there’s no record for C, or D, or…).
From there, we can compare that activity with results. Our business intelligence system can provide aggregated data of performance for A (whatever A is being measured on: sales data, errors, time to solve customer problems, customer satisfaction, etc). With that, we can see if there are the correlations we expect, e.g. everyone who demonstrated this level of capability has reliably better performance than those who didn’t. Or whatever you’re expecting.
Of course, you can mine the data too, seeing what emerges. But the point is that there are a wide variety of things we might track (who touched this job aid, who liked this article, etc), and a wide variety of impacts we might hope for. I reckon that you should plan what impacts you expect from your intervention, put in checks to see, and then see if you get what you intended. But we can look at a lot more interventions than just courses. We can look to see if those more active in the community perform better, or any other question tied to a much richer picture than we get other ways.
Ok, so you can do this with your own data generating mechanisms, but standardization has benefits (how about agreeing that red means stop?). So, first, does this align with your understanding, or did I miss something? And, second does this help, at all?
David Glow says
I really like this, and I think folks will understand it extends further. What actions did I take to demonstrate X? Did I read books? Take stretch assignments? Get mentored? Take a course? …. And of those, which patterns led to the most users being able to perform (successfully)?
This is where analytics must go. It really isn’t much different than these activity trackers and diet logging really exposing things that significantly help, and/or sabotage a healthy life. I think we can apply the same rigor to L&D. Not as prescriptives, but as potential opportunities to consider for one’s own development, or to offer as an organization to achieve goals (and weed out less effective methods- no need to buy an deep, expensive course catalog for the LMS if those items do not have ANY correlation to results).
Thanks, David. It’s clear that a finer and richer granularity gives us greater insight. I like the analogy to fitness trackers; I may steal it ;).
Andrew Downes says
Thanks for taking a stab at this diagram. I think it’s certainly possible that an org could implement xAPI in this way. What we’re seeing in practice with the organizations we work with though is that it’s not just learning data that gets brought into the LRS, but performance data too, often with a connector to translate this data from non-learning systems into the xAPI structure. The kind of correlation between learning data and performance data you suggest then happens inside Watershed LRS. There’s a few case studies that illustrate that approach here: https://www.watershedlrs.com/success-stories
I hope that’s helpful!
Hi Clark, great post. Yes, this aligns with my understanding. At Learnovate, we did a small research project + case study on xAPI (blog on the VERY basics here: http://www.learnovatecentre.org/xapi-exploration/). However, what I would add to your graphic is a ‘visual’ layer on top of the analytics layer. After all, someone has to interpret the analytics and you need to be able to present your analytics in an effective user friendly way.
I totally agree with you that data mining is not the way to go, especially because xAPI is supposed to be LEARNER centric as opposed to DATA centric.
Also, another thing that I have learned when working on the project: when designing statements it is mostly the object and/or context that provide most meaning; not the verb.
Transactional capture to provide insights at the cross section of business data is one useful application of xAPI. The types of activities that can be captured are expanded by xAPI.
Typical systems of practice capture (if they capture at all) “solo” and “facilitated” events. But development opportunities are about more than events and the interaction scale extends into rich areas that our formal systems ignore. Solo > Pair > One-To-Many > Many-To-One > Work Unit Group > Community > Society > World.
The LRS is a database. One of the problems with the current iteration of xAPI / LRS is the way queries are run. To get the right data, an xAPI query is only going to work when you have a relatively simple query. To run a complex query using xAPI you’ll either need to return A LOT more data than you need and filter at the client or run multiple queries and combine them. I anticipate this will get better. To get good analytics out of an LRS, we currently need to fallback to the specific features of the LRS. It’s an interoperability shortcoming that I hope we see movement on sometime soon.
What I’m most excited about is using xAPI to personalize experiences and shape content based on previous experiences. So we could map one activity to another and present specific elements based on what someone has accomplished before. With most technology mediated learning activities, we are fairly limited in how we can shape “e-learning stuff.” Especially with SCORM content. By adding xAPI, we can get information outside of the sealed container. Declare an accessibility preference at some point? We can add that to your profile and pull it out anywhere. Pass a proctored exam last year that proves you know a lot about Y? When the module on Y appears, the module can know you did that and give you advanced topics instead of the basic stuff. You’re an instructor teaching a course on Z? You can get a little bit of background information right on your iPad and use that to provide extra help where it’s needed or to pair folks up in the lab.
Lots of possibilities to use insights both for management decision making and to help adjust interventions to meet the participant where they are. Exciting stuff.
Thanks, Mirjam and Andrew. Agreed, Andrew, that you could and should be able to bring the biz data into the LRS and use it as the analytics engine. I just wanted to highlight that xAPI *by itself* isn’t the solution. Thanks for the dose of reality. Similarly, Mirjam, I agree that making the outcomes visual would improve the ability to process it. You’d have to know what you want to visualize, but then yes it more aptly leverages our powerful visual processing systems.