Keynote delivered by George Siemens (@gsiemens), Dragan Gasevic (@dgasevic), and Ryan Baker (@BakerEDMLab) at the 8th International Educational Data Mining Conference (EDM 2015) in Madrid, Spain on June 27, 2015 Educational data mining and learning analytics have to date largely focused on specific research questions that provide insight into granular interactions. These insights have bee abstracted to include the development of predictive models, intelligent tutors, and adaptive learning. While there are several domains where holistic or systems models have provided additional explanatory power, work around learning has not created holistic models with the level of concreteness or richness required. The need for both granular and integrated high-level view of learning is further influenced by distributed, life long, multi-spaced learning that today defines education. Drawing on social and knowledge graph theory, we propose the development of a Personal Learning Graph (PLeG) - an open and learner-owned profile that addresses cognitive, affective, and related elements that reflect what a learner knows, is able to do, and processes through which she learns best. This talk will introduce PLeG, detail required technical infrastructure, and articulate how it would interact with established learning software.

1. PERSONAL LEARNING GRAPHS
(PLeG)
George Siemens
Dragan Gasevic
Ryan Baker
Presented to:
International Educational Data Mining Conference
Madrid
June 27, 2015

2. The so-called “holy grail” of education:
personalized and adaptive learning

3. Personalized learning models
Keller Plan (Personalized System of Instruction)
Static learner profile (old school)
Objective based (adaptivecourseware)
Intelligent tutors (CMU OLI, cognitive tutor,
ALEKS)
Personalized (outer-loop, i.e. Knewton)
Smart Sparrow (teacher at center)

4. Parallel developing partners
Platform Publisher
Knewton Pearson
Smart Sparrow McGraw-Hill
Desire2Learn adaptcourseware
LoudCloud CMU OLI

5. Introducing PLG
Learner owned
API-like interface to systems that need
information
Related to existing work:
eportfolios
Personal learning networks
Existing toolsets (Learning Locker)

6. Elements
Cognitive
Process & strategy (meta-cognitive)
Affective/Engagement
Social

7. Why PLeG?

8. Jobs: disappearing & new
Automation
(Frey & Osborne, 2013)
Knowledge work
(US Bureau of Economic Analysis,
McKinsey & Co, 2012)
http://www.oxfordmartin.ox.ac.uk/downloads/academic/The_Future_of_Em
ployment.pdf
http://www.bea.gov/industry/gdpbyind_data.htm
http://www.mckinsey.com/insights/organization/preparing_for_a_new_era_
of_work

9. McKinsey Quarterly, 2012

12. Student profiles
Diversifying
(OECD)
Less than 50% now full time
(US Census Bureau)
http://www.oecd.org/edu/skills-beyond-school/EDIF%202013--
N%C2%B015.pdf
http://www.census.gov/prod/2013pubs/acsbr11-14.pdf

13. Complexification of higher education
Learning needs are complex, ongoing
Simple singular narrative won’t suffice going
forward
The idea of the university (and learning) is
expanding and diversifying

14. Granularization of learning
Competency-based degrees
(Chronicle, 2014)
Prior learning assessment
(Insider Higher Ed, 2012)
http://chronicle.com/article/Competency-Based-Degrees-/144769/
http://www.insidehighered.com/news/2012/05/07/prior-learning-
assessment-catches-quietly

15. Granularization of assessment
Cracking the credit hour
(New America Foundation)
Badges
(Mozilla & others)
http://newamerica.net/publications/policy/cracking_the_credit_hour
http://openbadges.org/

16. Something is needed that expands the
idea of a “course” and moves control
of learning experience/data from the
institution to the learner

17. Exploration
Learning is the exploration of the unknown…
… not just mastery of what is already known.

18. Compelling Questions
Habitable Worlds:
Are We Alone?
Contagion:
Can We Survive?

19. Astronomy
Chemistry
Geology
Physics
Biology
The questions we care about don’t fit in silos
Transdisciplinary

20. Smart Courses

21. What will PLeG enable?
Career transitions
Full spectrum of learning (hobby, work, formal,
personal)
Integrated & immersive learning
Foundation for personalized/adaptive learning

22. Multipartite graphs

23. Process & strategy
Login
Lesson
View
Discussion
View
Discussion
Post
Quiz
Discussion
Reply

24. Discussion
Reply
Login
Lesson
View
Quiz
Discussion
View
Discussion
Post
Lesson
Text
Discussion
Text
Discussion
Text
Quiz items
Discussion
Text
Cognition & Affect/EngageProcess & strategy

25. Login
Lesson
View
Quiz
Discussion
View
Discussion
Post
Lesson
Text
Discussion
Text
Discussion
Text
Quiz items
Discussion
Reply
Discussion
Text
SocialProcess & strategy Cognition & Affect/Engage

26. High computational complexity,
but suitable representation

27. Coding of nodes
(necessary) to describe PLeG
Information, learning processes, affective
states, social functions, etc.

28. Use of process/strategy graphs
Understanding of self-regulated learning

29. Capturing traces of SRL
Macro-Level SRL
Process
Micro-Level SRL Process Description Example SRL Event
Planning
Task Analysis
To become familiar with the learning
context and the definition and
requirements of a (learning) task at hand
Clicking on different competences under
duties or projects related to the user
Goal Setting
To explicitly set, define or update learning
goals
Drag and dropping an available
competence to a new or an existing
learning goal
Making Personal Plans
To create plans and select strategies for
achieving a set learning goal
Choosing an available learning path as the
path for a competence
Engagement
Working on the Task
To consistently engage with a learning task
and using tactics and strategies
Request collaboration for a competence,
learning path or learning activity
Applying appropriate
Strategy Changes
To revise learning strategies, or apply
change in tactics
Adding a new activity to an existing learning
path
Evaluation &
Reflection
Evaluation
Evaluating one’s learning process and
comparing one’s work with the others
Rating a learning path, learning activity or
knowledge asset
Reflection
Reflecting on individual learning and
sharing learning experiences
Adding a comment for a competence,
learning path or learning activity
Siadaty, M., Gasevic, D., Hatala, M., Winne, P. H. (2015). Trace-based Micro-analytic Measurement of Self-
Regulated Learning Processes. Submitted to the Journal of Learning Analytics.

30. Siadaty, M., Gasevic, D., Hatala, M., Winne, P. H. (2015). Trace-based Micro-analytic Measurement of Self-
Regulated Learning Processes. Submitted to the Journal of Learning Analytics.

31. Siadaty, M., Gasevic, D., Hatala, M., Winne, P. H. (2015). Trace-based Micro-analytic Measurement of Self-
Regulated Learning Processes. Submitted to the Journal of Learning Analytics.

32. Use of process/strategy graphs
Measurement of metacognitive monitoring

33. Learning strategy
-transition graphs-
Student A
(course 2 – graded)
Student B
(course 4 – non-graded)

34. Orchestration graphs
Process modeling and process mining (discovery,
compliance checking, and improvement)
Dillenbourg, P. (2015). Orchestration graphs. Lausanne, Switzerland: EPFL Press / Routledge

35. Information structure of content
Information extraction techniques such as topic
modeling (LDA) or name entity extraction

36. Connectivism as a
learning theory
Networked learning
Educational
technology
- Connectivism,
- Social media,
- Emergence,
- …
- E-learning,
- Complex
adaptive system,
- edtech,
- …
- Social network,
- Networked
learning,
- Social group,
- …
Connectivism in practice
- Collaboration,
- Knowledge,
- Thought,
- …
Joksimović, S., Kovanović, V., Jovanović, J., Zouaq, A., Gašević, D., Hatala, M. (2015). What do cMOOC participants talk
about in Social Media? A Topic Analysis of Discourse in a cMOOC," In Proceedings of the 5th International Conference
on Learning Analytics & Knowledge (LAK 2015), Poughkeepsie, NY, USA (pp. 156-165).
Topic extraction

37. Readings and Discourse Similarity
Joksimović, S., Kovanović, V., Jovanović, J., Zouaq, A., Gašević, D., Hatala, M. (2015). What do cMOOC participants talk
about in Social Media? A Topic Analysis of Discourse in a cMOOC," In Proceedings of the 5th International Conference
on Learning Analytics & Knowledge (LAK 2015), Poughkeepsie, NY, USA (pp. 156-165).

38. Coding learning processes from
unstructured sources

40. Cognitive presence
Triggering event
Exploration
Integration
Resolution
Garrison, D. R., Anderson, T., & Archer, W. (2001). Critical Thinking and Computer Conferencing: A Model and Tool to
Assess Cognitive Presence. American Journal of Distance Education, 15(1), 7-23.

41. Cognitive presence classifier
SVM classifier with the RBF kernel
Features: N-grams, Part-of-Speech N-grams, Back-Off N-grams, Dependency Triplets, Back-Off Dependency
Triplets, Named Entities, Thread Position Features, LSA Features, LIWC Features
Cohen’s κ = 0.42. Unigram baseline model: Cohen’s κ =0.33

42. Currently missing!
Integration of
LA/EDM & assessment
to guide learning progression

43. Promising development
Trace data based measures of
the crowd-sourced learning skill
E.g., Dreyfus model of skill acquisition
Milligan, S. (2015). Crowd-sourced learning in MOOCs: learning analytics meets measurement theory. In Proceedings
of the Fifth International Conference on Learning Analytics And Knowledge (pp. 151-155). ACM.

44. Progressions can build upon
• Models that represent prerequisite structure
and connections in knowledge
• Such as Partial Order Knowledge Spaces
(Desmarais & Pu, 2005)

45. Engagement in the PLeG
• Behavioral Engagement
• Affective Engagement

46. Engagement predicts learning

47. Engagement predicts long-term
participation

48. Engagement predicts long-term
participation
Engagement during middle school math predicts
– College attendance (San Pedro et al., 2013)
– College selectivity (San Pedro et al., in
preparation)
– College major (San Pedro et al., 2014, 2015)

49. Engagement predicts long-term
participation
Completing an EDM MOOC predicts joining the
EDM Society (Wang, Paquette, & Baker, 2015)

50. Community Factors Matter
Communities form during MOOCs like this one
(Brown et al., 2015)
Future work – study how these communities
persist into the future
(early evidence from CCK08 MOOC)

51. Use PLeG to
• Track what aspects of student engagement are
enduring
• As opposed to just pertaining to a specific
system or learning domain

52. Use PLeG to
• Determine when students are disengaged
• And track them to activities that can re-
engage them

53. Use PLeG to
• Find what does motivate a student
• And personalize less motivating content to
connect it to what motivates the student (cf.
Walkington & Bernacki, 2014; Walkington et
al., 2014)

54. Use PLeG to
• Figure out student long-term trajectories and
inform instructors and guidance counselors

55. Challenges
• Linking engagement models from different
learning systems to each other
– Models of different constructs
– Models with different reliabilities
– More and less aggressive models
• Figuring out how to decay engagement data
over time, and where it does and doesn’t
apply

56. Self-Regulated Learning
Similar challenges

57. We know…
• Scientific inquiry skills transfer across domains
(Sao Pedro et al., 2012)
– Essential if we are dealing with complex and multi-
disciplinary problems
• SRL skill that a student develops can be enduring
across a semester (Roll et al., 2011)
• These processes and strategies support the
development of cognition
– Can also support social skills, and affect and
engagement regulation skills

58. But…
• To what degree does SRL process skills in one learning
environment transfer to other environments?
• Are the same strategies and processes positive across
different learning environments?
– What behaviors are beneficial across learning
environments?
• Are the same strategies and processes effective for
different cultures and populations?
– Soriano et al. (2013) has found evidence that this is not the
case

59. Conclusion
Expansion of learning (for so-called knowledge age)
requires expansion techniques and methods for
learning
Learning controlled, owned
Personalized learning – by starting with learners
driving their learning
Resonance & activating latency
Labour market & related impact (rethinking “the
course”)
Need YOUR/EDM algorithmic and related expertise