dragan gasevic soed 2016
TRANSCRIPT
Using learning analytics to uncover learning strategies
Dragan Gašević @dgasevic
Shape of Educational Data Meeting April 7, 2016, Fairfax, VA
Freeman, S., Eddy, S. L., McDonough, M., Smith, M. K., Okoroafor, N., Jordt, H., & Wenderoth, M. P. (2014). Active learning increases student performance in science, engineering, and mathematics. Proceedings of the National Academy of Sciences, 201319030.
Categorization Deep and surface approaches to learning
Trigwell, K., & Prosser, M. (1991). Relating approaches to study and quality of learning outcomes at the course level. British Journal of Educational Psychology, 61(3), 265-275.
Poor choices of learning tactics and strategies
Bjork, R. A., Dunlosky, J., & Kornell, N. (2013). Self-regulated learning: Beliefs, techniques, and illusions. Annual review of psychology, 64, 417-444.
Significant role of instructions on approaches to learning
Trigwell, K., Prosser, M., & Waterhouse, F. (1999). Relations between teachers’ approaches to teaching and students’ approaches to learning. Higher Education, 37(1), 57–70.
Role of course design
To prompt active engagement and challenge higher order thinking
Bryson, C., & Hand, L. (2007). The role of engagement in inspiring teaching and learning. Innovations in Education and Teaching International, 44(4), 349–362.
Student profiling
Unsupervised approaches
Lust, G., Elen, J., & Clarebout, G. (2013). Students’ tool-use within a web enhanced course: Explanatory mechanisms of students’ tool-use pattern. Computers in Human Behavior, 29(5).
Sequences of activities
Sequence or process mining, HMMs, etc.
Reimann, P., Markauskaite, L., Bannert, M. (2014). e-Research and learning theory: What do sequence and process mining methods contribute? British Journal of Educational Technology, 45(3), 528-540.
Study context
Freshman course in computer systems at USyd
Enrolment: ~300 students
Assessment: midterm + final + project
Flipped learning design
Redesigned lecture – an active learning session requiring students’ preparation
Flipped learning design
Videos with multiple-choice questions (MCQs)
Documents with embedded MCQs
Problem (exercise) sequences
Exploratory sequence analysis
[1] (CONTENT_ACCESS,3)
[2] (EXE_IN,3)-(EXE_CO,1)-(EXE_IN,1)-(EXE_CO,1)-(EXE_IN,2)
[3] (CONTENT_ACCESS,3)-(EXE_IN,4)
[4] (MC_EVAL,4)
[5] (EXE_IN,5)-(EXE_CO,1)-(EXE_IN,3)-(EXE_CO,1)-(EXE_IN,2)-(EXE_CO,1)-(EXE_IN,9)-(EXE_CO,4)-(EXE_IN,4)-(EXE_CO,1)-(EXE_IN,2)-(EXE_CO,2)-(EXE_IN,3)-(EXE_CO,3)-(EXE_IN,1)-(EXE_CO,2)-(EXE_IN,1)
[6] (CONTENT_ACCESS,2)
Gabadinho, A., Ritschard, G., Müller, N.S. & Studer, M. (2011). Analyzing and visualizing state sequences in R with TraMineR, Journal of Statistical Software, 40(4), 1-37.
Agglomerative hierarchical clustering of sequences based on Ward’s algorithm and Levenshtein’s edit distance
Clusters of learning sequences
Pattern/strategy 1 (1354, 11.93%): focus on formative assessment, followed by metacognitive evaluation activities
Pattern/strategy 2 (4736, 41.72%): focus on summative assessment with indicators of trial-and-error learning
Clusters of learning sequences
Pattern/strategy 3 (3228, 28.44%): focus on reading lecture materials with tiny fraction of formative assessment Pattern/strategy 4 (2033, 17.91%): focus on the course videos, with not negligible amount of formative assessment activities; small fraction of metacognitive evaluation activities at the beginning of the learning sessions
Student clustering based on sequence clusters
All the cluster pairs, except for the 1-2 pair, are significantly different (even after applying the FDR correction for multiple testing) in terms of both midterm and final exam scores
Intensive/adaptive Strategic/effective Selective/efficiency Minimalist
Changes in learning strategy
Feature Feature description
MCQ.TOT.FACT Discretized count of completed formative assessment items (MCQs)
MCQ.PERC.CO.FACT Discretized percentage of correctly solved MCQs
EXC.TOT.FACT Discretized count of completed summative assessment items (exercises)
EXC.PERC.CO Discretized percentage of correctly solved exercises
VID.TOT.FACT Discretized count of play and pause video events
MCQ.SH.TOT.FACT Discretized count of requests for answers on formative MCQs
TG.DENS.FACT Discretized transition graph density
MC.EVAL.FACT Discretized count of dashboard and Hall of Fame views
CONTENT.ACCESS.FACT Discretize count of accesses to the lecture content pages
Changes in learning strategy
State Short description Correspondence to sequence-based student clusters
1 Low activity level; focus on lecture materials and summative assessment
Minimalists
2 High activity level; students are engaged with all the preparation activities and are experimenting with different learning strategies
Intensive / adaptive
3 Disengaged -
4 Moderate activity level; similar to state 2 in term of engagement and the diversity of learning strategies, but with lower activity level
Strategic / effective
5 Focus on summative assessment; low engagement with lecture materials and very rarely with the course videos; skipping formative assessment
Selective / efficiency-oriented
Process nature of learning - beyond coding and counting -
van der Aalst, W. (2012). Process mining: Overview and opportunities. ACM Transactions on Management Information Systems (TMIS), 3(2), 7.