How Simulation can Illuminate Pedagogical and System Design Issues in Dynamic Open Ended Learning Environments
Stephanie Frost · University Library - University of Saskatchewan (University of Saskatchewan) · 2017
A Dynamic Open-Ended Learning Environment (DOELE) is a collection of learners and learning objects (LOs) that could be constantly changing.In DOELEs, learners need the support of Advanced Learning Technology (ALT), but most ALT is not designed to run in such environments.An architecture for designing advanced learning technology that is compatible with DOELEs is the ecological approach (EA).This thesis looks at how to test and develop ALT based on the EA, and argues that this process would benefit from the use of simulation.The essential components of an EA-based simulation are: simulated learners, simulated LOs, and their simulated interactions.In this thesis the value of simulation is demonstrated with two experiments.The first experiment focuses on the pedagogical issue of peer impact, how learning is impacted by the performance of peers.By systematically varying the number and type of learners and LOs in a DOELE, the simulation uncovers behaviours that would otherwise go unseen.The second experiment shows how to validate and tune a new instructional planner built on the EA, the Collaborative Filtering based on Learning Sequences planner (CFLS).When the CFLS planner is configured appropriately, simulated learners achieve higher performance measurements that those learners using the baseline planners.Simulation results lead to predictions that ultimately need to be proven in the real world, but even without real world validation such predictions can be useful to researchers to inform the ALT system design process.This thesis work shows that it is not necessary to model all the details of the real world to come to a better understanding of a pedagogical issue such as peer impact.And, simulation allowed for the design of the first known instructional planner to be based on usage data, the CFLS planner.The use of simulation for the design of EA-based systems opens new possibilities for instructional planning without knowledge engineering.Such systems can find niche learning paths that may have never been thought of by a human designer.By exploring pedagogical and ALT system design issues for DOELEs, this thesis shows that simulation is a valuable addition to the toolkit for ALT researchers.Thank you to my supervisor, Dr. Gord McCalla, for teaching me how to be a researcher.Early on you explained techniques for reading papers and following references, and in later years coached me to evolve my writing beyond technical descriptions to speaking directly to the deep issues in AIED that fascinate me so.With this research area being so new, we didn't have much of a basis for comparison.Sometimes I'd lose confidence when my simulation output was different than expected.I always appreciated your insights, ideas and guidance toward a research problem that was manageable yet still an exciting and useful contribution.Thanks to my advisory committee: Dr. Jim Greer, for being supportive of the research I wanted to do on top of working at the ULC.Dr. Julita Vassileva, for your many past publications that have inspired me, and your constructive criticism that motivated me to improve my writing.I also want to acknowledge you and other women I've met in CS.I think that women should be allowed to either succeed or fail at things (like programming!)without anyone assuming that their performance is related to being female.But when you're one of the only ones, this is really hard to avoid!Dr. Jay Wilson, for your feedback ranging from grammar to important questions about the human element when AI is applied to Education.Thanks to others who have supported my studies: Dr. Mike Horsch, for the introduction to the wider field of artificial intelligence and insight into the reaches of AI with Prolog; Dr. Nate Osgood, for AnyLogic classes and introduction to system dynamics and agent-based modelling; Dr. Chris Dutchyn, for opportunities to improve presentation and research skills; Dr. Ray Spiteri, for the chance to look at one of my research questions under a new light (planning as an optimization problem); my colleague Dr. Carolyn Hoessler for the crash course on t-tests; Dr. Vania Dimitrova for helping me see how to connect real learning theories to my work, with Gord in Madrid!Thanks to my fellow students, lab mates, co-workers, friends, family members