Save the date. August 28th will be the time when the Moodle Learning Analytics Working Group (working title) will gather online for the first Community Call. In the agenda there are introductions, a call for R&D partnerships, and a “Setting up the ‘Community of Inquiry’ model” seminar. Also, to agree upon a few definitions. It could be a little chaotic, and that’s okay. The current state of the debate underscores the importance of diverse voices and lines of work, at the beginning as well as throughout the life of open source development collaboration.
Define not too narrow goals
Moodle HQ’s Learning Analytics leader Elizabeth Dalton recognizes the perils of premature compromises. It is in fact the subject of the first case of the “Analytics in the News” series, an initiative she inaugurated with an article about Charles Goodhart and the paradox carrying his name. Shortsighted, indicator-maximizing goals can easily compromise the purpose or sustainability of a system. Most career researchers, educators, (or fans of The Wire, Season 4) alike, are aware of the problems arising from focusing exclusively on high scores to evaluate students, organizations districts and programs.
Speak a human and machine-readable language
Still, collaborative teams must find common goals, affinities and a language over which to progress. For now, most of the process of codification and data standardization has fell in Dalton’s shoulders. For the subject matter of courses in her current dataset she has, for example, adopted UNESCO’s standard of education and training. She developed a Competency Framework for Moodle courses willing to submit to them. (It’s available for download here.) In the future, Moodle’s machine learning models can take advantage of the codes and the competency framework to optimize models for specific cases and subject matter. Furthermore, it can help in the reproducibility of the models, by nurturing from educators everywhere running them under the same parameters.
Standards are an unassailable part of any collective roadmap. On another discussion, Dalton has proposed two, stemming form Data Mining Practice
- PMML, or “Predictive Model Markup Language”
- PFA, or “Portable Format for Analytics”
While they offer a broad intersection, their use case is subtly different. Among other standards likely to come up, the most obvious one has to be xAPI with its slowly, yet certain growth rate of adoption.
The specific discussion on representation standards for Moodle Analytics models continues to take place in Moodle Tracker Issue MDL-60944. It’s worth reminding that Issue activity (“Votes”, “Watches” and “Favorite” stars) helps its case as a development priority by the Moodle HQ core development team. Here are some of the Issues for which the team would like your support.
Make it visually usable
While few Moodlers might find discussions about code and data standards riveting, a more approachable debate, and one that is just as important, involves the visual interfaces for analytics. A more friendly interface to manage a model and its variables could be the key towards a higher adoption of the powerful analytics engine shipped since Moodle 3.5. The roadmap could be split into two parts of the User Interface: a UI for model creation, and a UI for the display of quantitative results.
Little attention seems to be given to best practices, layouts and wireframing tools in visual model design, evidence of which is visible in the “clunky” diversity across analytics dashboards. (Moodle developers use Balsamiq, Among the few, the Shiny package for RStudio, the development environment for the open source scientific programming language R, stands out, with a wide variety of interfaces for the display of quantitative information. Admittedly, working with Shiny requires some programming skills, but it is a versatile and elegant solution, in no small part due to the influence of Edward Tufte.
Translating the impeccable interfaces of Shiny to Moodle is no small feat. But Dalton offers a tempting proposition:
R and Shiny depend on Python, which is one reason they haven’t been considered for integration with Moodle core– we don’t like to add dependencies to Core features. I wonder if a Shiny visualization plugin for Moodle would be of interest?
Attend Moodle Learning Analytics first community call on Tuesday, August 28, 14:00 UTC following this link.