Saturday, May 10, 2008

The Great Display Wall of China

In the "Event Webs" panel that followed, which was moderated by Bonnie Nardi, there was a terrific presentation by Ruth Mostern of U.C. Merced about the challenges of representing history with new digital tools. As a practitioner and partner in software development, she emphasized one of the major problems at issue: "How do we think about history, so that it works with a database?" She also showed Google maps to illustrate how spatial representations that use computer-mediated communication technologies tend to be much more sophisticated than temporal representations.

I particularly liked the fact that she was interested in ideology, politics, and state power in understanding how reconfigurations of historical events and geography take place. Her own research area has to do with space and place in China, where she showed how even seemingly discrete map markers could morph over time in relationship to other locales.

To demonstrate the kinds of "genre trouble" created by this work by digital historians, she showed examples of both "Digital Gazetteers" and "Historical Atlases" as preliminary models. She argued that ironically some of the best data representation of history is being done by law enforcement agencies that are creating crime mapping projects such as Instant Atlas and CopLink, which provides "geotemporal" visualization to locate perpetrators. However there are academic institutions also working on the problem of modeling discrete events, such as the Temporal Modeling Project at the University of Virginia or a number of initiatives at the Berkeley iSchool. I particularly liked the idea for a demo of SemTime, described at NewsBlip, for how various historical actors learned about Watergate that could encompass two different versions of time and the meaningfulness of events. Unfortunately, time limitations prevented Mostern from showing much of GapMinder or the SIMILE earthquake USGS mash-up.

Mostern claimed that the biggest constraint currently inhibiting this kind of scholarly work was the difficulty of having computers represent causality. As she pointed out with a diagram of the sequence for how to stop at a stoplight, the formal logic of Spatial Ontology Theory is far too cumbersome for more complex historical events and inadequate for the interpretive work of the historian with causality, narrative, agency, perspective, memory, emotional relationships, and themes. Although she did acknowledge that it might be possible, given our "socially authored world," for a computer to one day model a "Palestinian-authored world" and an "Israeli-authored world."

During the question-and-answer session, I was glad to see fellow panelist Lewis Lancaster of the Electronic Cultural Atlas Initiative bring up the work of Judea Pearl of UCLA. In addition to maintaining the Causality Blog, Pearl is known for his sophisticated work with Bayesian statistics to model causal relationships. (To learn more about this subject, you can also use "belief nets" or "belief networks" as keywords.) To her credit, Mostern was certainly aware of the importance of network graphs and pointed to the work of Lothar Krempel and papers such as "Diffusion Dynamics in an Historical Network" as very influential on her thinking.

Labels: , , , , ,


Post a Comment

<< Home