THIS PAGE REDIRECTS FROM MIT.EDU/CLIO/WWW
What we're thinking about over here in Building 9, 77 Mass Ave. Read more [Technical Report : 'Informed Relationships'] (download pdf) (view as a google doc)How do we integrate personal relationships into geography? What would happen if there was better transit to see your friends and family? Have you moved somewhere because someone you like already lived there or was moving there? Are you friends with someone when you are 'in town', but not otherwise? Should you move somewhere where everyone has contacts elsewhere, or has contacts nearby? Has your relationship with someone improved because where you live and work has become convenient for meeting up? How has this changed your life? How can we measure this? Should we measure this? Tell me what you think. I am a Postdoctoral Researcher studying the convergence of Geospatial Networks and Social Complexity. My studies are a cross section of Geography and Computer Science.
RESEARCH INTERESTS
RESEARCH OVERVIEW In Plain English: Normally, when we measure distance, we measure in terms of miles, or minutes. For example, Boston is about 500 miles from Washington D.C., or, it takes 8 hours to drive. We call straight line distance (like on a map) "Euclidean Distance", because it connects two points usually at a diagonal. We call travel time (how long it takes to get somewhere) the "Cost Distance", where cost could be time, money or even wear-and-tear on a car. These two measures are very handy
for a lot of people, like airlines, or food distributors. But we now
may need a third measure: With new data and math software coming at us
from every direction, we have more information about how people
communicate, on the telephone or email, or how they commute from home to
work, and where they migrate. With this information, we can measure how
much two places "talk" to each other, by summing the phone call
minutes, or how much two places "trade people", by how many people have
moved from one place or another in the past year. High numbers show that two places are connected well, and low numbers show that they are not connected well. We could imagine that two neighboring towns have strong connections, and two distant towns have virtually none. But from our research, close distance can't always predict a strong connection. A Bostonian is more likely to know someone/or have visited San Francisco than San Antonio, Tx., even though they have the same population size, and San Antonio is closer to Boston. Now we can calculate "how much more likely" to get a feeling for the "social distance" between two places. This social distance can be important for measuring and predicting spreading of a lot of things: ideas, diseases, consumer behavior, voting preferences, or lifestyle choices like the adoption of a recycling program. I like this stuff a lot, and there's a lot more to say!
 In GIS/Mathy Language: (map: phone calls around London) From another perspective, social networks themselves can be coupled with spatial analysis tools for visual data mining and pattern recognition. Linking and Brushing techniques can HIGHLIGHT user-selected nodes (depending on a certain feature--all nodes in a module, nodes with high closeness centrality, female nodes, nodes with the most neighbors) ALONG WITH each selected node's corresponding location on a map. Of course, entities on the map could be selected to reveal participants in the social network. More dynamically auto-updated coupling views include regression plots and parallel coordinate plots (for showing trends and signals for nodal features--age, income, clustering coefficient, centrality, etc.), as already embedded in current open-source geographic environments. This system is important because most social network models only focus (if at all) on the straight line distance as the geographic separation between two people. This measure is not enough to show spatial clustering, autocorrelation, hot and cold spots and anisotropic spreading and diffusion patterns. Characterizing the entire network of places and their connectivity measures as a system allows us to see dynamics from many neighbors-away, so we can model the causes and effects of growth or shrinkage to all entities connected to somewhere, not just the place itself. These types of tools are inherently spatial, give us the benefit of a multi-scale analysis, sub-partitioning and fusing layers, and should be integrated with interdisciplinary GIS systems for the research community. 
EXAMPLE APPLICATIONS
|
