A few months ago Dan McFarland and Jim Moody asked me to create a movie showcasing some of the work done in SoNIA as submission to the netsci06 visualization contest. We didn’t win, but here it is any way for your viewing pleasure.
The winer (Diversity and Complexity of Ecosystems: Exploring Balance and Imbalance in Nature) definitely had the most interesting story, and some really cool features. It was especially nice to see the evolution of the sim parameters alongside the changing network with edges growing, shrinking, and nodes disappearing. But in my opinion the actual network display was quite poor. I think a better layout could have made the hierarchy much clearer and I’ve never been very fond of 3D visualizations for networks, but that is a discussion for another time.
Anyway, some more details on our movie for anyone who is curious about the various clips:
The Key, and Classroom Data
It starts out with a sort of animated key to explain how to read all the bouncing bubbles and lines before complicating the picture with any real data. Then we cut to a clip of some of the actual data collected by Dan, showing observed conversations among teachers and students in a high-school classroom. (more on that here) You can see the contrast between recitation (dominated by the black teacher-centered star) and free time, which shows up as small cliques and groups of interactions. Contrast this with the next classroom, in which the teacher has pretty much lost control and is shouting at everyone (red arrows) to try to get them back on task. The third class shows students taking turns giving presentations, standing up and talking to the entire group – very stable but not too exciting.
Then I tried to change gears and explain the concept of aggregating on different timescales. The first movies are shown with 2.5 minute rolling windows onto the interactions. When we go back to the data for the first class period and we cut that window down to 30 seconds, the network is just the dyads and triads of turn taking. Almost more sequence than network. When we go the other way, showing an aggregate network that is formed by a 10 minute history of the conversations, we what looks more like a “traditional” social network: a dense group with a periphery, some nearly isolated pairs (4-13 and 7-19) all linked by some moderate talk by the teacher.
Ben Shaw’s Collaborative Design
The next section is a very short clip from Ben Shaw’s doctoral dissertation work on shared representations in design teams Although it is difficult to to read the node labels in this scaled down version, it shows a network of interactions between individuals, representations (screens, whiteboards, etc) and concepts evolving in time alongside the video recordings of the JPL design team. So when one of the engineers walks up to the sketch of the radiator on the whiteboard and begin discussing it you can see the arcs linking the display and the radiator concept appear in the network discourse diagram and begin to pull together. All of which you can see much better in Ben’s full versions.
Moody’s Well-Balanced Sim
We then jump to the other extreme: purely simulated data. A few seconds from Jim Moody’s simulation of a social balance process that was included in the SoNIA paper in AJS.
At each of 200 iterations, five randomly chosen nodes evaluate their local network with respect to transitivity, intransitivity, and reciprocity; nominations are changed if doing so increases the comfort of the actors’ overall position with respect to these characteristics. Actors favor relations that are transitive, avoid those that are intransitive, seek to reciprocate nominations, and avoid making long-term asymmetric nominations [Moody, McFarland, Bender-deMoll 2005]
Blue ties are asymmetric nominations and green ties a symmetric relations.
CSDE’s Sexy Networks
The last segment shows some of the most recent work in SoNIA. A toy version of the models of disease spread produced by the statnet/Network Modeling team at the Center for Studies in Demography and Ecology in the University of Washington. The team has been working with all sorts of folks to create realistic statistical models of networks and dynamics than can be scaled up large enough to do simulations of the spread of diseases such as HIV. (Check out Dr. Morris’s NIH lecture…) This model shows sex contacts in black between nodes (colors representing different races) Infected nodes have red borders and infecting edges turn red. We used a 5 step time window so that part of the time-connected-component would be visible.
Thats all folks, fade to credits.