Reuven M. Lerner, Sharona T. Levy, and Uri Wilensky

1
Making Modeling SocialIntroducing the Modeling
Commons

• Reuven M. Lerner, Sharona T. Levy, and Uri
  Wilensky
• Northwestern University University of Haifa
• Meital Conference
• October 14th, 2009

2
Models

• Model Reified theory about a system
• Often used in science and engineering
• Anatomy, molecules, DNA
• Also mathematical models
• Economics, cognitive science

3
Modeling and Constructionism

• Purpose of modeling Construction and revision of
  conceptual understanding (Jonassen, 2006)
• Papert Best way to construct knowledge is by
  creating sharing artifacts (1991)
• Modeling has been shown to help learning
  (Goldstone Wilensky, 2008 Blikstein
  Wilensky, 2008)

4
NetLogo
5
NetLogo

• Wilensky, 1999
• http//ccl.northwestern.edu/netlogo
• 100,000 users worldwide
• Includes 400 models, code examples
• Associated curricula (ProbLab, Connected
  Chemistry, MaterialSim, NIELS, BEAGLE)

6
Building vs. sharing

• NetLogo is successful for building...
• ... but existing community structures are not
  promoting sharing.
• Sharing in journals, Web not community
• netlogo-users Only 5.7 of 9,696 messages
  included a model (over 7 years)
• Limited community models (274 in 7 years)

7
Interactions are vital

• Vygotsky Learning leads to development when
  interacting with people in his environment and
  in collaboration with his peers. (1978, p. 90)
• Lave Wenger (1991) Legitimate Peripheral
  Participation, CoP
• Schön (1983) Reflection in Action

8
Collaboration is vital

• A skill and a perspective (Kolodner and Guzdial,
  1996)
• Much of our culture depends on collaboration and
  remixing (Lessig, 2008)
• CSCL sees collaborative communities as a learning
  paradigm (Stahl, 2006 Koschmann, 1994)

9
My research

• Encourage sharing, collaboration, communities of
  practice
• Encourage communication with models, not about
  them
• Identify patterns that emerge from such
  interactions

10
Modeling Commons
11
Modeling Commons

• Wikipedia of modeling
• View, share, collaboratively build models
• Run models in the browser
• Create families of branched models
• Discussion, requests for help
• Social tagging

12
Research to date

• Round 1 Winter-spring 2008
• 12 people, 3 interviews, 4 written reports
• Concentrated on design
• Round 2 Fall 2008
• 24 people, 10 interviews
• Combination of design and usage

13
Research to date

• Round 3 Spring 2009
• 36 participants in 3 classes at 2 universities
• Total of 90 models uploaded

14
Design Research

• Clinical interviews (Ginsburg, 1997) and defined
  tasks (Nielsen, 2000) to improve design
• Improve system usability
• Reduce threshold to sharing, collaboration,
  discussion

15
Home pages

• Users consistently reported a lost feeling
• It would be nice ... to find out what models
  were new, or what models there had been recent
  activity on. Both, actually... heres a new
  model, and people are talking about this model.

16
Solution Dynamic page
Your models
Most-viewed
Your tags
Most downloaded
Requests for help
17
Privacy

• 8/10 subjects said sharing my models with
  others has importance of 4 or 5 (out of 5)
• However, many expressed reservations for
  classwork or work-in-progress
• ...I wouldnt have been comfortable, and the
  people I work with wouldnt have been comfortable
  showing the models to the world in an
  unfinished form

18
Groups, permissions
19
Groups, permissions
20
Logfile analysis

• Every action in the Commons is logged
• Allows for analysis, understanding without
  looking over users shoulders

21
Logfile analysis LPP

• Predicted by Lave, Wenger (1991)
• 3 neither viewed nor uploaded
• 8 did view, but never uploaded
• 15 viewed before uploading
• 9 uploaded before viewing
• Some LPP behavior is seen in 63 (n 36)

Legitimate Peripheral Participation
22
Networks

• Explicit vs. implicit networks
• Networks of people
• via models, discussions, tags
• Graphs are from one university class, where
  students were instructed to tag and discuss
  models, as well as upload them
• Some measures from Krackhardt (1994)

23
Explicit membership
24
Models creators
Connectedness 0.097 Hierarchy 0.271
25
Models taggers
Connectedness 0.656 Hierarchy 0.0426
26
Models commenters
Connectedness 0.862 Hierarchy 0.085
27
Models all actions
Connectedness 0.731 Hierarchy 0.030
28
Social networking

• Explicit group membership does not necessarily
  indicate actual ties
• Connectedness is a useful comparison only when
  the number of nodes is the same
• Multiple communication channels would appear to
  be the key to a truly connected graph (and thus
  community)

29
Future work

• 3 courses already used it another soon
• Refining and extending logfile, network analysis
  methods
• Explore http//modelingcommons.org/
• Soon will be announced to the world

30
Acknowledgements

• Prof. Uri Wilensky
• Dr. Sharona T. Levy
• Members of the CCL
• Researchers, students, and TAs who participated