Research

I am a postdoctoral associate in the Department of Sociology at Cornell University and a member of the networks research team led by Michael Macy. My research brings together data mining, statistical modeling and computer simulation with the aim to advance our knowledge about dynamic processes in social networks.

I received my PhD from Columbia, where my advisor was Duncan Watts. My dissertation included a case-study of an e-mail communication network in a large university (about 100,000 faculty, students and staff), where social interactions, affiliations and attributes of individuals were all recorded over time. I developed a methodology to recover network structure from discrete communication events and quantified the process of tie formation by simultaneously analyzing interactions, individual attributes, and shared groups (Science 311:88-90, 2006). The results are especially relevant for building realistic models of evolving social networks, an important task for sociology, information science and epidemiological research.

I also studied the impact of missing data on network properties by simulating non-inclusion and non-response of individuals as well as missing social groups (Social Networks, 28:247-268, 2006). The results show that missing data may dramatically affect the estimates of network statistics, which could lead to erroneous conclusions about the diffusion properties of such network and, consequently, to inappropriate interventions (for example, in epidemiological control). Missing data in networks is an important methodological problem on which I hope to do more work in the future.

Since arriving at Cornell, I have continued the exploration of large-scale dynamic networks. In a recently submitted paper with Jon Kleinberg and Duncan Watts, we have described a new set of results regarding the temporal properties of social communication. The goal of the study is to better understand how information diffuses in organizations and distributed communities, for example, via e-mail and instant messaging.

My most recent project is centered on Wikipedia - a free online encyclopedia written and edited by hundreds of thousands of volunteers. It has been observed that many participants in online communities contribute very little and leave, whereas very few contribute a lot. I am interested in the sociological underpinnings of volunteers' productivity and longevity. By combining event history modeling, natural language processing and network analysis, I plan to study how the contributors' output depends on the negative and positive feedback they receive from their peers, and also on their positions in the community's social network. In collaboration with Michael Macy, Jon Kleinberg, Ted Welser, and Dan Cosley, I plan to extend this study to compare Wikipedia contributors to product reviewers on Amazon.com and participants in a popular online discussion forum. I envision that this research will improve our understanding of the dynamics of participation in online communities from both the theoretical and practical point of view.

Publications

G. Kossinets and D. J. Watts. Empirical analysis of an evolving social network. Science 311:88-90, 2006.

G. Kossinets. Effects of missing data in social networks. Social Networks 28:247-268, 2006.

G. Kossinets. True colours of the allies: Analysis of voting patterns in the Polish parliament. In: S. Kapralski and P. V. Smith (eds), Democracies, Markets, Institutions: Global Tendencies in Local Contexts. IFiS Publishers: Warsaw, Poland, 2002.