For my social networks class, I had to post a response to some in-class readings and talk generally about the research that I would like to be doing. It ended up as a pretty concise summary of my thoughts on what I want to do, so I’m going to post it here for the world.

As a Computer Science student, my research is primarily centered on applying methods and tools of computer science to problems in the social sciences; particularly economics and information. It is an inherently interdisciplinary challenge. Social network analysis and applications fall right in the intersection of these fields. Using the tools of computer science and mathematics (mostly but not exclusively statistical hypothesis testing) I want to help develop better tools for social science researchers.

Currently, I’m most interested in problems of computational complexity and algorithm design (particularly randomized algorithms) to these applied fields. Thus, I found Hanneman’s “note on statistics and social network data” to be an interesting insight into some of the challenges of computational complexity of social networks data. Particularly, it leads to the following computational issues/questions:

A useful tool for statistical hypothesis testing on social networks is comparing a given network versus a randomized network. What happens when the network is enormous, with millions of nodes? Simulation of a random network will have a great degree of computational complexity. How can we reduce the complexity? Are there sufficient approximation algorithms we can use? What about random sampling, is that a sufficient tool for analysis of a network? It certainly is for other statistical research; what makes sampling different in a network context?

The issue of computational and algorithmic complexity is frequent in the social sciences. Economists make mathematical assumptions about consumer and producer decisions without considering the often extremely demanding computational resources required. In the “real world”, agents usually make decisions without advanced computational resources; this demands a new model of decision making under uncertainty or with limited resources. Certainly, models like this exist for decision making on a social network, particularly in the case of diffusion and information theory.

For example, assume I exist as a node on a large social network. I am considering adopting a new technology if some of my neighbors (other nodes, connected by an edge) have the technology already. Furthermore, I weigh the edges to my neighbors by some value based on their influence over my decision making process. My decision making process is a computational problem based on my preferences (i.e. utility), the proportional weight of my connections, and how the technology has diffused on the network already. This is a computational challenge; what algorithm do agents use to make a decision, how complex is this algorithm? Is it reasonable to assume that agents can even run this algorithm?

These are the type of questions that I intend to research in this class and in general. As an aside, I am always looking to make connections and do research with social scientists, so if anyone is interested in doing work together in a very interdisciplinary field, feel free to email me. My background is in the social sciences (communications and information science) but I’m in EECS program here at NU.

This entry was posted on Tuesday, January 12th, 2010 at 11:24 am and is filed under Uncategorized. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.