This presentation is an overview of Open Social Learner Modeling project. It presents Mastery Grids interface, distributed personalized learning architecture Aggregate, and smart content for Java, Python, and SQL
Adaptive Navigation Support and Open Social Learner Modeling for PAL
1. Adaptive Navigation Support
and Open Social Learner
Modeling for PAL
Peter Brusilovsky
School of Information Sciences,
University of Pittsburgh
2. Key Goals
• Leverage large volume of data left by past
learners to improve learning process
• Better Interface
– Open Social Learner Modeling Interface for visual
learning analytics and content access
• Better Personalization Algorithms
– Enhancing personalized learning algorithms for
personalized guidance and content recommendation
3. Three Directions of Project Work
• Exploring Open Social Learner Modeling interface
– Diverse learning content
– Multiple domains
• Enhancing personalized learning algorithms for
personalized guidance
– Pro-active content recommendation
– Remedial content recommendation and guidance
• Develop architectural support and authoring tools
for Open Social Learner Modeling
– OSLM as a reusable component
– Content and course authoring tools
7. Accessing OSLM ADL Demo
• You can enter our at:
• http://adapt2.sis.pitt.edu/kt
• ADL Usernames:
– adl01, adl02…, adl10
– Passwords are the same as the usernames
• You will see links for courses in Java, SQL and
Python programming
• Note that most adlxx users are "empty", which
means that they do not have activity yet. User adl01
has some activity..
8. OSLM Experience
• Developed full-semester courses with smart
learning content for 3 domains, all accessible
online, used in several universities
– Java Programming
– Introduction to Databases with SQL
– Python Programming
• Evaluated Mastery Grids Interface in many
classroom studies
– Value of open learner model
– Value of social comparison
9. Term / Course Research set up / comments Active
Learner
s
Questio
n
Attempt
s
Annotated
Examples
Viewed
Animated
Examples
Viewed
IS0017 Fall 2013 Mastery Grids and Portal 38 3832 747 -
IS0017 Spring 2014 Social enabled
(preliminary version of MG)
41 2707 551 -
IS0017 Fall 2014 mixed
(random assign toSocial and non-social groups)
65 4563 1936 670
IS0017 Spring 2015 mixed
(random assign toSocial and non-social groups)
36 2146 947 281
IS0017 Fall 2015 mixed
(random assign toSocial and non-social groups)
58 7109 2689 1165
ASU Fall 2014 Social features enabled Recommendations
enabled
100 9285 4186 -
ASU Fall 2015 Different recommendation algorithms (2
groups)
74 4364 1175 505
CS 401 Fall 2015 mixed
(random assign toSocial and non-social groups)
68 2715 1267 606
WSSU Fall 2013 2 groups, Control / Social 22 876 340 -
WSSU Fall 2014 2 groups, Control / Social 20 1837 618 112
National Sun Yat-Sen
Univ. Taiwan Spring
2015
26 1550 889 420
JAVA Courses
10. Term / Course Research set up / comments Active
Learne
rs
Questio
n
Attemp
ts
Annotated
Examples
Viewed
IS 1022 Fall 2013 Mastery Grids and Portal 15 530 212
IS 1022 Fall 2014 Social features enabled 33 1194 793
IS 1022 Spring 2015 Social features enabled 18 224 277
IS 1022 Fall 2015 Social features enabled 32 526 787
IS 2710 Fall 2013 Mastery Grids and Portal 44 510 213
IS 2710 Fall 2014 2 groups, Control / Social 97 6819 2876
IS 2710 Spring 2015 Social features enabled 33 3616 1506
IS 2710 Fall 2015 2 groups, Control / Social 56 3486 1531
SQL Courses
11. Term / Course Research set up / comments Active
Learne
rs
Questio
n
Attemp
ts
Annotated
Examples
Viewed
Animated
Examples
Viewed
Parso
ns
AALTO Universty
Fall 2015
Social and non-social groups 490 7909 6187 4545 9158
IS0012 Fall 2015 Social features enabled 19 1301 586 548 1628
Python Courses
12. OSLM – Some Findings
• OLM/OSLM significantly improve learning
engagement, problem performance, learning
gain
• Social comparison in OSLM further increase
learner engagement
• OSLM helps students to work more efficiently
• OSLM preserves mastery orientation
13. OSLM Increases Engagement
Variable
OSM OSLM
U
Mean Mean
Sessions 3.93 6.26 685.500*
Topics coverage 19.0% 56.4% 567.500**
Total attempts to problems 25.86 97.62 548.500**
Correct attempts to problems 14.62 60.28 548.000**
Distinct problems attempted 7.71 23.51 549.000**
Distinct problems attempted correctly 7.52 23.11 545.000**
Distinct examples viewed 18.19 38.55 611.500**
Views to example lines 91.60 209.40 609.000**
MG loads 5.05 9.83 618.500**
MG clicks on topic cells 24.17 61.36 638.500**
MG click on content cells 46.17 119.19 577.500**
MG difficulty feedback answers 6.83 14.68 599.500**
Total time in the system 5145.34 9276.58 667.000**
Time in problems 911.86 2727.38 582.000**
Time in MG (navigation) 2260.10 4085.31 625.000**
14. Aggregate: The Architecture behind MG
• Extension of our original architecture ADAPT2
• Allows transparent connection of independent
smart learning content that is interactively
delivered by smart content servers
• Supports extensive tracking of learner activities,
learner record storage, learner modeling, group
modeling, social comparison
• Supports multi-domain course authoring,
content brokering and concept brokering
16. Smart Content and Content Authoring
• We created large volume of reusable smart content
– With activity tracing, content brokering, authoring
• Pittsburgh team smart content
– Interactive examples: Java, SQL, Python
– Java exercises
– SQL problems
– Python exercises
– YouTube video sections
• Helsinki team smart content
– Animated examples (Java and Python)
– Parsons problems (Python)
17. Full Support for Instructors
• Create course for any domain as sequence of any
topics
• Connect smart learning content of several kinds
from multiple content servers
• Create own content if existing content is not
sufficient
• Create groups and subgroups, assign to classes
• Observe class/group work with MG
18. Open Content, Open Source
• All developed content could be reused right from our
content server or by installing own content server
• All the sources are available in GitHub
– The Mastery Grids Interface, back-end Aggregate and documentation
can be found here.
– User model services can be found in here.
– QuizJET Interface, Authoring Tool, Content Brokering and
documentations can be found here.
– QuizPET Interface, Authoring Tool, Content Brokering and
documentations can be found here.
– Parson Problem Authoring Tool can be found here.
– Annotated Examples Interface, Authoring Tool, Content Brokering and
documentations can be found here.
– Animated Examples Authoring Tool can be found here.
– Videos User Interface, Authoring Tool, Content Brokering and
documentations can be found here
19. Algorithms
• Student Modeling
– Several data-driven student modeling approaches
– Most notable is FAST, an extension of BKT that can
use additional data New work: multi-content social
student modeling
• Recommendation
– Several algorithms for proactive content
recommendation and remedial example
recommendation
– Performed several studies demonstrating the value of
recommendation
20. Further Information
• Project Home page
– Explanations, demos, videos, flyers
– http://adapt2.sis.pitt.edu/wiki/ or http://bit.ly/1Ty5KOr
• GitHub
– Sources, installation, system documentation
– https://github.com/PAWSLabUniversityOfPittsburgh
– https://github.com/acos-server/
• Publications, conference presentations
– Interface, algorithms, studies, evaluation data
– Available from the project home page