Presentation on social learning analytics for online professional learning by Kathleen Perez-Lopez and I at Learning Analytics and Knowledge, May 2, 2012 in Vancouver.
Towards a Social Learning Analytics for Online Communities of Practice for Educators
1. First Steps Towards a Social
Learning Analytics for
Online Communities of
Practice for Educators
Darren Cambridge
Kathleen Perez-Lopez
2. Community Cultivation
• Now
– Assess4ed.net
– ConnectedEducators.org
• Coming soon
– EPIC-ed Dropout
prevention and recovery
3. Outreach
• Connect & Inspire
• Briefs
• Community directory
• Innovations blog
• Connected Educators
Month
4. Research
Evolution Tracking evolution of five emerging online communities; examining
critical decisions made by leaders and the ways in which decisions
are informed by data, resources, and people.
Value creation Collecting value creation stories and survey data from a range of
established communities to determine which online activities,
content, and interactive features best support learning and
provide value to educators.
Engagement Beginning design-based research in new EPIC-Ed community.
Current focus is on design interventions to increase
“connectedness” among educators.
Social roles Exploring the use of social network analysis in four communities to
identify and better understand the connecting patterns and social
roles of online community leaders.
5. Research Team
• Researchers • Case study partners
– Darren Cambridge, AIR – Al Byers, NSTA
– Kathleen Perez-Lopez, – Sheryl Nussbaum-
AIR Beach, PLP
– Rachel Crossno, AIR – Sharon Roth, NCTE
– Sherry Booth, NCSU – Lia Dossin & Geoff
– Shaun Kellogg, NCSU Fletcher, SETDA
– Bobby Hopgood & Lisa
Hervey, NCSU
– Jim Burke, English
Companion Ning
– Andrew Gardner,
BrainPop
6. Learning Analytics Goals
• Small set of visualization methods and tools
simple enough for regular, direct analysis by
community managers
• Practitioner question driven
• Support reflective dialog about what to do next
• More efficient use of expert community
moderator judgment
• Actionable intelligence Actuated intelligence
7. Social Learning Analytics
Approaches
• Focus on three of Ferguson and Buckingham
Shum’s five:
• Social learning network analysis
• Social learning content analysis
• Social learning context analysis
8. NSTA Learning Center
• 8,300+ PD
Resources and
Opportunities
• 100K+ users
• Badges and
leaderboards
• Learning plans and
portfolios
• Expert advisors
• Forums
9. Learning Needs of Science
Teachers
• Science teachers need to learn continuously and broadly
– To address mandates to teach “out of field” (particularly grades
6-8)
– To address topic focus of coming standards that cross disciplines
– To incorporate changing body of pedagogical content knowledge
• Teachers often come to the Learning Center initially to
address an immediate challenge
– I need to teach students the difference between weather and
climate tomorrow morning
• What activities lead to broad and sustained
engagement?
• How can we lower barriers to entry in conversation while
maintaining connections between people?
10. Year of NSTA LC Posts 9/24/2010 - 9/28/2011
6978 posts
21 forums
492 members
557 topics
SNA using NodeXL
http://nodexl.codeplex.com/
11. Quintile 1 9/24/2010 to 1/9/2011
Early Months:
Very little activity from
these members
12. Quintile 2 1/10/2011 to 2/26/2011
2nd Quintile:
Activity building here,
but still light
13. Quintile 3 2/27/2011 to 5/7/2011
3rd Quintile:
Lots of posts to
one private forum
14. Quintile 4 5/8/2011 to 7/25/2011
4th Quintile:
Private forum
died out, but much
more activity from these
members
15. Quintile 5 7/26/2011 to 9/28/2011
5th Quintile:
Activity concentrated
among these members,
and healthy activity among
lower posters.
16. Repartitioning Topics
Find Fn , a partition of topics, that yields:
1. VERY segregated Topic network, Tn
474 x 281 474 x 474
281 x 474
X
Member-Topic
Topic-Member Tn
2. UN-segregated member network, Mn
281 x 20+ 281 x 281
20+ x 281
X
Fn-Member Mn
Member-Fn
17. Clustering Algorithms
• Clauset-Newman-Moore groups (NodeXL)
• Wakita-Tsurumi groups (NodeXL)
• M-slices and k-cores (Pajek)
• Wakita-Tsurumi on a reduced dataset
• Wakita-Tsurumi on member network
Perez-Lopez, Cambridge, Byers, & Booth (2012) Sunbelt XXXII
18. Adding Content Analysis
• Better to have a different way to represent the
natural clustering of topics than by those who
post to them
– Textual content analysis to locate concepts: LSA + ?
• Filtering out non-contextual content
– Friendly banter
– Useful for other purposes, but interference here
19. Adding Context Analysis
• Pre-hypothesis narrative
research using
CognitiveEdge
SenseMaker Suite
• Narrative fragments +
quantitative classification
by author
• “Filter questions” indexed
to Wenger, Trayner, &
DeLaat’s (2011) five
cycles of value creation
• Authors linked to usage
data
21. Key Questions We’re
Thinking About
• Significant differences in purpose, context, and
theories of learning
– Are the managers' questions likely to be similar enough?
– Is there likely to be a set of visualizations that can be
useful across contexts?
• Can techniques of sufficient power to tell managers
something they don’t already know be made
sufficiently accessible that they actually use them?
• Which techniques are most likely to be worth
focusing on next?
22. We’d Love to Hear From
You
• connectededucators.org
@edcocp
• Darren Cambridge
dcambridge@air.org
+1-202-270-5224
@dcambrid
Editor's Notes
Add AIR & ED? logos, new colors and fonts
Difficult to see many patterns beyond the fact that participation is highly skewed. There are a few dense dark edges in the upper portion of the network, between relatively few members and topics. In fact, a few forums received most of the posts; a few members initiated most of the topics and made most of the posts, and a few topics received half of the posts). The 59 large heavy-posting members – those with 20 up to 582 posts – are allotted almost two-thirds of the vertical space. If members had been spread uniformly, the heavily skewed activity would have been even more apparent.However, we can see that the distribution of edges is not smooth from the upper portions of the network to the lower, and there appear to be a number of concentrations in mid-figure. There might be some interesting activity by members in those regions, but with this static view, it is difficult to see what that could be.
1- post nodes removed: topics with only one post made to them; members who posted only onceEdge color (black to yellow) and opacity are logarithmically proportional to the number of times a member posted to the topic
Looks Like Healthy Evolving CommunityIt might show how time bounded activity targeted at some subgroup could be leveraged into more sustained and general engagement.
Goal = no member islandsGo from bi-model to member-member and topic-topic Look for repartition where topics are maximally distinct and people are maximally connected Add cite
Topic network of 556 nodes is very dense: 0.48; average degree is 267. Doesn’t seem to decompose to many separable subnetworks. Group
Unique identifiers for respondents index to SNA/content analysis data