Although we know a lot about learning and eduction, there is still a lot to be discovered. In particular, we need to understand human variability better and to take into account how differing human cultures impact what we want to learn and how we learn it.
Of course, we also need to take advantage of each advance in the digital technology and to carefully craft the interfaces we build that impact interaction both with the machine and the information it delivers.
KEPLAIR founders, myself and Stefano Ferilli have been outlining the components that will make up the KEPLAIR system and uncovering the research topics that will enhance development. I will return to this blog often and add to it. For now…
Knowledge Graph AI/Ontology
Fundamental to KEPLAIR is knowledge graph or semantic AI. This is a reasoning system based on concepts and how they connect via specified logical rules. Compare this technique with machine learning that relies on large networks of switches “trained” in a limited set of data. With knowledge graphs a human being can follow the way the AI makes its decisions. Machine learning systems are opaque to humans.
Knowledge graphs are informed by “ontologies”, highly structured formal descriptions of the concepts and connections. A great deal of research on knowledge graphs has been done in the fields of medicine, and computer science – not so much in learning and education. We are working on formal descriptions for three of the four major KEPLAIR concepts: Goals, Learner Profile, and Learning Environment. The fourth, Learning Objects (the physical and digital tools we use to learn with such as books and videos) is already quite well developed.
Learner Profile
What do we need to know about a person in order to identify learning activities that feel like play while advancing them toward self-chosen educational goals? Pervious knowledge, primary language, and digital access are obvious parameters. But what about a learner’s preference for social context: solo, duo, small group, large group? Do they like to be highly challenged and fail often or are they more comfortable when they know how to succeed at a task most of the time? How much “agency” or self-determination motivates them? Do they feel angry and hemmed in when told what to do or lost at sea without direction from an outside authority? This is a different way to approach recommendation then algorithms that create classes of people who bought the same item.
Much of the current research on “personalizing” instruction assumes that
- goals or learning objectives are predetermined
- success or failure of an activity can be measured by testing
- there is a preexisting order in which skills and concepts must be mastered
- the learner is a “student” in a school, university, or employment context
These and many other aspects of a learner’s personality and preferences affect what, when, where, and how they will learn. As KEPLAIR developers we cannot take on this kind of research challenge ourselves. Instead, we are keeping up with psychological, sociological, educational, and vocational breakthroughs made by other investigators. Of special interest are validated instruments that are “gamified” and might be integrated into the KEPLAIR system. The result will be a sophisticated, individual, learner profile generated with a minimum number of dry surveys and test batteries.
Next?
My notes from the last 4 years contain many other R&D topics I want to add to this blog – – and, I would like to hear from you as well. Could some aspect of what you are working on be incorporated into the KEPLAIR concept? When you contemplate your ideal “guide-on-the-side” for your own learning, what would you like it to be able to do or recommend. Please email your ideas to me: lizaloop@loopcenter.org
