It takes years of deliberate experience not only to learn anything of value in a field, but to be able to put it in simple, succinct words. This is what Moodle HQ Research Analyst Elizabeth Dalton gave us in her talk, Moodle Analytics Plans, last September.
Plans are all about connecting data, actions and results.
Admittedly, Dalton works from a position of advantage compared to other players in the LMS analytics arena. The founder of Moodle reached out to her specifically to lead the initiative, which she responded with initial hesitation. Not because she did not feel up to the task, but because she felt as too much of a disruptor.
Dalton believed the offerings of analytics spoke too much corporate and too little actual intervention. It surprised her when the community seemed to forget or overlook that “learning analytics are about learning” (emphasis hers). It is understandable that marketing materials and approaches are made in the context of a business case, especially for learning organizations focusing on the enterprise. But she fears the “metaphor is going too far“. In her review she discovered that most of the time the analytics product does little more than changing the tagline from “web” or “business” to “learning”.
They carry over certain assumptions (…) into the learning environment that do not necessarily apply there as well as they did in the original context (…)
Analytics should be about learning first.
Which takes us to a question we perhaps should be considering more often: What is learning? As dealing with this question would take her talk to a halt, Dalton simply refers us to research on curriculum theory and ideologies, made by Stephen Schiro, PhD. at Boston College. Each organization should identify their definition somewhere along Schiro’s proposed models:
- Academic Scholar
- Social Efficiency
- Learner Centered
- Social Reconstruction
I probably should dig further into each of the four. One article will not do. It is interesting to note that practical elements in the learning interventions depend on the approach. A definition of learning could stress teamwork, or unlocking and individual’s full potential. Among other desirable pairs of things that are inevitably at odds with each other.
This long introduction is necessary, in Dalton’s eyes, to understand its importance. Because only then learning operations could identify their critical action and apply quantitative sets of research and control. Each model gives a different priority to measurable elements, like retention or time spent by learner.
Moodle Machine Learning API
Supported by her doctorate candidate research, Dalton shares some details of this project, currently in development. (And without any reference I could find, either from official or informal sources.) This API would collect reams of data by courses and seasons from any Moodle setup that allows to. This would have the goal to
[P]ush data and retrieve predictions from machine learning tools.
Data would be anonymized and become available for Moodle core processing, as well as third-party services. It hopes to take advantage of all the data collection activities already present in Moodle, often out-of-the-box. The API would tap into, say, the Moodle Logs, and turn rows of values into practical knowledge.
Dalton lists three potential focuses of data processing, or “presences”: Social Presence, Cognitive Presence, and Teaching Presence. They could intersect, in fact it seems Dalton will encourage them to, as the trifecta is the completion of what she deems a “good Educational Experience”.
It is evident the API has a long road ahead. I will be attentive to any further developments, as well as patient. After all, if you intend to roll out a machine learning solution that benefits but also requires raw data from millions of users worldwide, the plan better be flawless, with its assumptions outlined as clearly as possible.