Open Learner Models Working in Symbiosis With Self-Regulating Learners: A Research Agenda
- PDF / 338,678 Bytes
- 14 Pages / 439.37 x 666.142 pts Page_size
- 105 Downloads / 164 Views
Open Learner Models Working in Symbiosis With Self-Regulating Learners: A Research Agenda Philip H. Winne 1 Received: 15 January 2020 / Revised: 29 July 2020 / Accepted: 12 August 2020 # International Artificial Intelligence in Education Society 2020
Abstract Learner modeling systems so far formulated model learning in three main ways: a learner’s “position” within a lattice of declarative and procedural knowledge about highly structured disciplines such as geometry or physics, a learner’s path through curricular tasks compared to milestones, or profiles of a learner’s achievements on a set of tasks relative to mastery criteria or a peer group. Opening these models to learners identifies for them factors and relations among factors. Open learner models tacitly invite learners to regulate learning. However, contemporary learner models omit data about how learners have and should process information to learn, understand, consolidate and transfer new knowledge and skills. What to do with information opened to learners is opaque. I propose incorporating trace data about learning processes in learner models. Trace data allow generating learning analytics that inform self-regulating learners about potentially productive adaptations to processes they have used to learn. In a context of big data, such elaborated learner models are positioned to collaborate with self-regulating learners. Together, they can coordinate symbiotically, creating a platform for the system to improve its models of learners and for learners to more productively self-regulate learning. Keywords Open learner models . Trace data . Self-regulated learning . Learning analytics
In 1994, Greer and McCalla edited a volume reporting state-of-the-art work on learner modeling presented at a NATO-sponsored workshop. At that time, the basics of learner modeling involved recording a learner’s correct and incorrect responses to a series, sometimes a branching series, of questions or steps taken to complete a task. Those data were processed to instantiate a representation of the learner’s state of knowledge about that topic or task. An early representation was updated as successive learner responses became available. The model of a learner’s knowledge or skillset was compared to a standard model. Usually, the standard was a representation based on a logical analysis * Philip H. Winne [email protected]
1
Faculty of Education, Simon Fraser University, Burnaby, BC V5A 1S6, Canada
International Journal of Artificial Intelligence in Education
of prerequisites in a discipline or an expert’s (or several experts’) account(s) of correct or efficient performance. Approaches to modeling differed mainly in formalisms used to represent knowledge and methods for comparing learner data to the standard model. Learner models of these sorts were prevalent in intelligent tutoring systems (ITS) learners used to study subjects ranging from algebra to medicine to law to reading. A recent meta-analysis reported positive effects of ITSs compared to various alternative instructional arr
Data Loading...
