My Insightful Thoughts on Learning Analytics

I feel like I don't have a lot to say about Learning Analytics even though I find the topic fascinating. Maybe it's because most of the articles talked about Learning Analytics from this high level view without really getting into the nitty gritty of what learning analytic engines have already accomplished and how and what we hope they might be able to eventually do. In short, a lot was said about something - I'm just not sure what exactly. I liked it however, and I'm excited to learn more of the details. 

From the excellent LAK 2012 lecture by George Siemens (which I found the most interesting of all the materials) the speaker describes the three main levels of learning analytics research and development:

1. Networks and social media analysis
2. Learning Analytics and Data Mining
3. The (digital) future of learning and learning institutions. 

I think his discussion of learning analytics and data mining the most related to what we have been we hope to accomplish in the class we will be designing. According to Siemen's framework, there are three major sub divisions within Learning Analytics and Data Mining itself: Classroom level data for helping teachers make better interventions, student level learning analytics to provide on-the-spot guided instruction, and institution level data analysis and analytics. 

Furthermore, the tools employed in each of these subdivisions can be roughly divided into several categories as well, although all of them are more of less intertwined one way or the other:

1. Data mining (looking for new and useful patterns not already known.
2. Predictive modelling (allowing for timely interventions)
3. Visualization (tools to help with the analysis of the data)
4. Social network analysis (it was never entirely clear how such analyses were going to potentially improve student performance)
5. Customizable dashboards allowing for personalized data analysis.
6. Intelligent curriculum and recommender systems.

What I find interesting is that each of these tools are created using almost entirely different methodologies. We have Bayesian analysis in one, good ol' software programming of basic visualizations in another, and in yet another we have a clustering algorithm from computer science. 

Moreover, these learning analytic tools are to be used for completely separate purposes and for completely separate people, ranging from student-centric analysis all the way up to aggregate data analysis for high level administrators. In some respects, sometimes it seems as if the only similarity between large domains of research classified as "learning analytics" is the use of computers and some usage of databases.

I cannot end my discussion of learning analytics without expressing my healthy skepticism towards much of the work and effort that has been given to learning analytics. Have you ever used Amazon's recommendation engine? Occasionally it gets something interesting, but very rarely. As an economics student and erstwhile statistician who's designed some data mining algorithms of his own, I've also come to realize that finding truly useful variables out of the vast array of possible data you could collect, or choosing the truly effective means of analyzing that data is a very imprecise process. Sad to say, the data does not naturally want to sing. Moreover, all learning analytics are implemented through algorithms. By their very nature algorithms are a fixed set of instructions for processing inputs to produce an output. They don't change or adapt - so for the foreseeable future we're going to still need humans

-------------------------------------------

I have to add a little on the bit from Creative Commons which was a good refresher on stuff I learned last year. I had never learned about the remixing rules and quite honestly it saddened me. In some ways, if it's not public domain or CC-BY I think it may not be worth your time to use the material. The author protests that regular copyright law has an even larger set of rules and he is undoubtedly correct in this, but in that case people aren't EXPECTING to remix it.  However, what Creative Commons has got going already seems like an excellent step in the right direction towards a greatly diminished copyright future.