Suthers, D. D., & Chu, K.-H. (2012, April 29-May 2, 2012). Multi-mediated community structure in a socio-technical network. Paper presented at the Learning Analytics and Knowledge 2012 conference
Multi-mediated community structure in a socio-technical network
1. Paper presentation at Learning
Analytics and Knowledge 2012
Modified (and background
removed) for SlideShare
Multi-mediated
Community Structure
in a Socio-Technical
Network
Dan Suthers & Kar-Hai Chu
University of Hawaii
Supported by the National Science Foundation
2. Learning in Socio-Technical Networks
Learning in university settings, professional
communities, and virtual organizations is increasingly
technologically embedded
– “online,” “distributed,” “networked,” “blended”
Fundamental question in all of these settings:
how learning and other enhancements of knowledge,
skill and capital take place through the interplay
between individual and collective agency
Demands analyses that connect learning activity in
specific times and places with the larger socio-
technical network contexts in which they take place
3. Levels of Learning in STNs
How do social settings foster learning?
Agency Epistemologies
Who or what is the agent What is the process of
that learns? learning?
Individual Acquisition
Small groups Intersubjective meaning-
Networks (communities, making
cultures, societies) Participatory
Learners participate in all “levels” simultaneously
Need to identify the social settings (“communities” or
networks) within which learners participate
Suthers (ijCSCL 2006)
4. Tapped In
SRI’s Network of education professionals: PD and peer
support (Mark Schlager, Patti Schank, Judi Fusco)
Since 1997: longest running educational online
community
8 years of data (7.4G)
20K educators/year
800 user spaces QuickTimeª and a
decompressor
are needed to see this picture.
50 tenants
40-60 volunteer-run
community-wide
activities per month
Chats, threaded discussions, wikis, resource sharing ...
5.
6. Empirical Community Identification
Schlager: “I don’t know what communities are there”
– Organizational “tenants” and individuals who come for their
own enrichment
– Multiple forms of participation and mediational means by
which participants associate with each other
An empirical matter:
– Don’t assume that the network constitutes one community
– Don’t assume that external communities are replicated within
the sociotechnical system
Our Approach:
– Identify clusters (cohesive subgroups) of participants
– Interpret clusters using affiliations and other information
– Note media (chats, discussions, files) through which they
interact
7. Relevance of Mediation
Multi-mediated: TI and other environments offer
multiple means of participation, each with their own
interactional and social affordances
Choice of technologies by which people keep in
touch both reflects and reaffirms the relationship
between interlocutors (Licoppe and Smoreda, 2005)
Apply this idea to collective rather than dyadic level:
communities are embedded within and make use
of technological media for interaction in ways
that reflect and reaffirm their nature
Our approach identifies the mediational means
simultaneously with identification of cohesive
subgroups (candidate communities of actants)
8. Traces Analytic Hierarchy
Basic needs
Activity is distributed across media:
– Reunite traces of interaction into a unified analytic artifact
Logs may record activity in the wrong ontology:
– Abstract event data to other appropriate levels of description
(interaction, mediated associations, ties)
Sequential interaction analysis and aggregate network analysis
are complementary:
– Enable mapping between these descriptions both ways
The Traces analytic hierarchy addresses these issues
Abstract transcript representation that collects relevant events
into a single analytic artifact
Analytic hierarchy that supports multiple levels of analysis
Suthers (HICSS 2011)
Suthers & Rosen (LAK 2011)
9. Interaction Affiliations
Uptake Ties
Contingencies
Mediated Associations
11. Tapped In Data
Selected 2 year period of high activity
Parsed and filtered logs of user activity involving files,
asynchronous threaded discussion forums, and quasi-
synchronous chat rooms
Events: accessing (reading and downloading) or
contributing (posting and uploading) to one of these
three artifact types
Filtered out:
– Private chats
– Activity in the K-12 (student) campus
– Guest accounts
– Indirect file access (portrait displays)
14. Tapped In Associogram
40,490 vertices = actants:
– 19,842 actors (49.00%)
– 12,037 discussions (29.73%),
– 5,862 files (14.48%)
– 2,749 chat rooms (6.79%)
229,072 edges = events
20,431,944 events (sum of weighted degree)
Average path length: 4.398
– This is bipartite graph: actor-actor path length about half!
– Largely due to Tapped In Reception (R1), normalized
betweeness centrality 0.665; weighted degree 2,511,057;
unweighted degree 18,810
– When R1 removed, average path length = 6.02
15. Finding Communities in Associogram
TI is a network; communities are embedded
Community: cohesive subgroup with identifiable
common identity, purpose, and/or task
“Community detection” (modularity partitioning)
algorithm due to Blondel et al.
– Maximize intra-partition connectivity in relation to inter-
partition connectivity (NP Hard)
Computed and visualized in Gephi (gephi.org):
– beta OSS for network analysis and visualization
– handles large graphs
Examined properties (e.g., organizational affiliation) of
high degree nodes in each partition to interpret as
communities
16. Visualization of Partitions in OpenOrd
171 Partitions
Modularity: 0.817
Blondel, V. D., Guillaume, J.-
L., Lambiotte, R., & Lefebvre, E.
(2008). Fast unfolding of
communities in large networks.
Journal of Statistical Mechanics:
Theory and Experiment,
http://dx.doi.org/10.1088/1742-5
468/2008/10/P10008.
Martin, S., Brown, W. M., Klavans, R., & Boyack, K. (2011). OpenOrd: An Open-Source Toolbox for
Large Graph Layout. Paper presented at the SPIE Conference on Visualization and Data Analysis
(VDA).
20. Interpreting P3
2565 actants (6.33%) P3 (red-brown)
– 851 actors
– 103 chat rooms
– 1286 discussions
– 325 files
Examine high degree actants …
All media used, with intensive
reading of discussions:
– Chat: 272,865 in, 226,561 out SRI Colleague:
– Discussion: 355,656 in, 5,976 out CoP mentoring
– File sharing: 4717 in, 325 out of new teachers
in a Midwestern
school district
22. High Degree Nodes
Actants of unweighted degree greater than 282. Vertex size scaled by weighted
degree. Radial layout with a non-overlapping filter
23. Interpreting P1
Top actants (unweighted)
– R1, Tapped In Reception
– R4, public room for Tapped In’s After
School Online (ASO) events
– R10, the Floor Lobby by which one enters
rooms on the Tapped In Groups floor
Top actants (weighted)
– R1, the Tapped In Reception: 12.29% of
all chat events in the network
– R3, the personal office of Actor F, an
educational researcher
Not a community with
– R4, the ASO Public Room.
its own purpose, but
Many highest ranked rooms are owned rather a network for
by Tapped In and function to welcome
Legitimate Peripheral
and route newcomers or as venues for
public events open to all Participation by which
18% of actors associated only with R1 other networks are
approached
Overwhelmingly chat based (1M to 1K)
24. Interpreting P2
Top Two Actants (weighted)
– Actor A: volunteer with normal
account and no affiliation, is the
most active account in the system.
– Actor B a volunteer with facilitator
status. The real-world actor was
given a second account B’ Taken
together, the real word actor B/B’
is as active as Actor A.
Most highly ranked actors are
help desk volunteers Chat-Based After
School Online Events
Top ranked discussions and
(tightly associated with
chats are group rooms, all of P1, separated to meet
which are used for ASO events requirements of non-
A and B regularly facilitate overlapping modularity
these events partitioning)
25. Summary: Community Interpretations
P1 P2 After School
LPP via TI
Reception and Online
other public Events
rooms
Mixed Media Chat-based
P3, P4 P5, P6
CoP in a Language
Midwestern Arts in the US
school district; Midwest;
Discussion- Pre-service
based program in
professional Western US
development
in the Southern
US
27. Summary & Comments
Purely structural (graph theoretic) computations
identified cohesive subgroups that have interpretations
as communities
Demonstrates vibrancy of Tapped In as “transcendent
community” ( Joseph et al., CSCL 2007)
Value to learning analytics: identify social units that
are the setting or agent of learning
Need to try with algorithm for overlapping cohesive
clusters
– Not clique percolation ( Palla et al., Nature 2005)
– Edge communities promising ( Ahn et al., Nature 2010)
Can “dive in” to examine activity of high-degree
actors, structure of chat sessions in rooms, etc.
28. Related Work: Chat Analysis
Find evidence for relatedness (“uptake”) between chat contributions
Structural analysis of the resulting graph
“Folding” into social network
29. Discussion
Dan Suthers
suthers@hawaii.edu
lilt.ics.hawaii.edu
Notas del editor
Explan title. (shoutout to Devan)
I said “Tapped In online community” and …
Shout out to Latour
(Could replace these two slides with simple explanation of associograms.)
Explain the term “Associogram” and its structure in detail. (Could replace these two slides with simple explanation of associograms.)
Skipped the Filtered Out -- too much detail
Perhaps the detailed numbers are OK here but not needed later.
Of course I don’t discuss the detailed numbers but I need to mention the role of R1 at some point.
Really don’t need all the numbers!!! Could delete the breakdown of actants . Or leave them there in case needed?
Could skip the demo animation; go direct to this. But it’s fun
Say something about the significance of the media distribution!
Comment on how the partitioning forces nodes into different subgroups even though they have high connectivity. (Stuff at the bottom is only paraphrased briefly.)
Did not have time to discuss the actants. Just summarized the third bullet and the results. I did not realize I had two small bullets at the end and talked ahead of them. In the talk I said “within which public communities are embedded” but they aren’t really embedded in THIS network. Maybe put an image from the high degree actant here.
Maybe put an image from the high degree actant here.
Just to show there were others …
Last item is last to transition to next slide, but would otherwise be second to last.