Computational Social Science Institute

University of Massachusetts Amherst

Lecture series generously sponsored by Yahoo!

Videos of some CSSI seminars are available  here

Spring 2012

Understanding the World in Realtime [video] [slides]
Friday, February 3, 2012 • 12:30PM–2PM • Lunch provided
Computer Science Building, Room 150/151

Abstract:More communication than ever before is happening online, and for the first time, much of that data is available to be analyzed. Bitly began as a simple URL shortener and has evolved into a content sharing and analytics platform spanning the social web, with tens of millions of URLs and hundreds of millions of clicks on those URLs daily. I'll discuss the kinds of questions that we ask and the pieces that we need to find answers, which draw from many disciplines (machine learning, sociology, and engineering, to name a few), and discuss the big questions afforded by this kind of data.

Bio:Hilary Mason is the Chief Scientist at bit.ly, where she finds sense in vast data sets. Her work involves both pure research and development of product-focused features. She’s also a co-founder of HackNY (hackny.org), a non-profit organization that connects talented student hackers from around the world with startups in NYC. Hilary recently started the data science blog Dataists (dataists.com) and is a member of hacker collective NYC Resistor.

Forecasting the Locational Dynamics of Transnational Terrorism: A Network Analytic Approach [video] [slides]
Friday, February 10, 2012 • 12:30PM–2PM • Lunch provided
Computer Science Building, Room 150/151

Abstract:Efforts to combat and prevent transnational terrorism rely, to a great extent, on the effective allocation of security resources. Critical to the success of this allocation process is the identification of the likely geopolitical sources and targets of terrorism. We construct the network of transnational terrorist attacks, in which source (sender) and target (receiver) countries share a directed edge, and we evaluate a network analytic approach to forecasting the geopolitical sources and targets of terrorism. We integrate a deterministic, similarity-based, link prediction framework into a probabilistic modeling approach in order to develop an edge-forecasting method. Using a database of over 12,000 transnational terrorist attacks occurring between 1968 and 2002, we show that probabilistic link prediction is not only capable of accurate forecasting during a terrorist campaign, but is a promising approach to forecasting the onset of terrorist hostilities between a source and a target.

Bio:Bruce received his Ph.D. in political science from the University of North Carolina at Chapel Hill in summer 2010. In fall 2010, he started as an assistant professor in the Deptartment of Political Science at the University of Massachusetts Amherst. He joined UMass as one of the core faculty members of the Computational Social Science Institute. Bruce's recent research focuses on the development and application of methods for the analysis of political networks. Substantive applications of network analysis include international conflict, defense alliances, terrorist events, legislative collaboration, and intra-governmental communication networks.

Sampling Online Social Networks [video]
Friday, February 17, 2012 • 12:30PM–2PM • Lunch provided
Computer Science Building, Room 150/151

Abstract:Online Social Networks (OSNs) have recently emerged as a new Internet killer-application and are of interest to a range of communities, ranging from computer science and engineering to social sciences. OSNs are widely studied today based on samples collected through measurement of publicly available information. In this talk, I will give an overview of our recent work on sampling online social networks. First, I will describe a framework for obtaining a probability sample of users by crawling the friendship graph. I will provide practical recommendations including the choice of crawling technique, the use of online convergence diagnostics, and implementation issues. Second, I will introduce multigraph sampling - a technique that exploits different relations between users to efficiently sample users, even when the friendship graph exhibits poor connectivity or slow mixing. Third, I will present the stratified weighted random walk (S-WRW) - an efficient heuristic that preferentially crawls those nodes and edges that convey greater information pertaining to the target metric. Finally, I will report results from applying our techniques to real-life OSNs, such as Facebook and Last.FM, and from studying their characteristics, including user attributes and structural properties. This work is joint with Mina Gjoka, Maciej Kurant, and Carter Butts at the University of California, Irvine and with Patrick Thiran at EPFL, Lausanne. Parts of this work appear in IEEE INFOCOM 2010, ACM SIGMETRICS 2011 and IEEE JSAC 2011.

Bio:Athina Markopoulou is an assistant professor in EECS at the University of California, Irvine. She received the Diploma degree in Electrical and Computer Engineering from the National Technical University of Athens, Greece, in 1996, and the Master's and Ph.D. degrees in Electrical Engineering from Stanford University, in 1998 and 2003 respectively. She has been a postdoctoral fellow at Sprintlabs (2003) and at Stanford (2004-2005), and a member of the technical staff at Arastra Inc. (2005). Her research interests include network coding, network measurement, network security and online social networks. She received the NSF CAREER award in 2008.

Cooperation in Static and Dynamic Networks [video]
Friday, February 24, 2012 • 12:30PM–2PM • Lunch provided
Computer Science Building, Room 150/151

Abstract:This talk describes the results of a series of web-based, behavioral experiments designed to understand people's ability to cooperate in static and dynamic networks. In the context of static networks, it was previously thought that cooperation should fare better in highly clustered networks such as cliques than in networks with low clustering such as random networks. To test this hypothesis, we conducted a series of experiments, in which 24 individuals played a local public goods game arranged on one of five network topologies that varied between disconnected cliques and a random regular graph. In contrast with previous theoretical work, we found that network topology had no significant effect on average contributions. Since humans have a natural tendency to choose with whom to form new relationships and with whom to end established relationships, we also study cooperation in dynamic networks. Helping cooperators to mix assortatively is believed to reinforce the rewards accruing to mutual cooperation while simultaneously excluding defectors. Here we report on another series of human subjects experiments in which groups of 24 participants played a multi-player prisoner’s dilemma game where, critically, they were also allowed to propose and delete links to players of their own choosing at some variable rate. Over a wide variety of parameter settings and initial conditions, we found that endogenous partner selection significantly increased the level of cooperation, the average payoffs to players, and the assortativity between cooperators. Joint work with Jing Wang (NYU) and Duncan Watts (Yahoo! Research).

Bio:Siddharth joined the Human & Social Dynamics group at Yahoo! Research led by Duncan Watts in August 2008. Prior to that he was a postdoctoral associate working with Jon Kleinberg in the computer science department at Cornell University. He earned his Ph.D. in computer and information science from the University of Pennsylvania in January 2007 under the supervision of Michael Kearns.

The Social Stratification of Fame [video]
Friday, March 9, 2012 • 12:30PM–2PM • Lunch provided
Computer Science Building, Room 150/151

Abstract: Sociologists have argued that in contemporary society public attention to celebrities is generally short-lived. Fame would thus escape the classic forces that generate stable hierarchies in traditional stratification systems such as social structure and cumulative advantage. We investigate the tenability of this theory in a unique data source containing daily records of references to person names in 2,500 English-language newspapers and blogs. Mobility turns out to be minimal at all but the lowest strata. While the bottom of the public attention hierarchy exhibits fast turnover, at middle and upper tiers people receive stable coverage that persists around a fixed level and rank for many years. This pattern of “lock-in” characterizes stratification even in the domain of entertainment, in celebrity-oriented tabloids and on blogs, where fame is thought to be most fleeting. Joint work with Charles Ward (SUNY SB), Steven Skiena (SUNY SB) and Eran Shor (McGill)

Bio: Arnout van de Rijt earned an M.Sc. in Sociology from Utrecht University under supervision of Vincent Buskens and his Ph.D at Cornell in 2007, working with Michael Macy. He joined the Sociology Department at SUNY Stony Brook in August of 2007. In 2010 he received the Linton C. Freeman Distinguished Young Scholar Award from the International Network for Social Network Analysis.

Elena Erosheva

Elena Erosheva, University of Washington

Modeling Criminal Careers: The Case of Marijuana Smoking
Friday, March 30, 2012 • 12:30PM–2PM • Lunch provided
Computer Science Building, Room 150/151

Abstract: A major aim of longitudinal analyses of life course data is to describe the within and between individual variability in a behavioral outcome such as crime or drug use. Methods currently used for analyzing individual behavioral trajectories rely on growth models, mixture models, or combinations of these two modeling approaches. Most commonly, these models specify a polynomial relationship between the age of an individual and the behavior of interest and use corresponding polynomial random effects and latent class indicators to describe between individual heterogeneity. We take a different approach by relying on functional data analysis. We model heterogeneity in individual criminal careers via departures from the unimodal population age-crime curve that correspond to individual differences in their (a) levels of offending (amplitude variability) and (b) patterns of temporal misalignment (phase variability). We extend Bayesian hierarchical curve registration methods to accommodate count data and to incorporate influence of time-stable covariates on individual behavioral trajectories. Analyzing self-reported counts of yearly marijuana use from the Denver Youth Survey, we examine the influence of race and gender categories on differences in levels and timing of marijuana smoking. We find that our approach offers a flexible and realistic model for longitudinal crime trajectories that fits individual observations well and allows for a rich array of inferences of interest to criminologists and drug abuse researchers.

Bio: Dr. Erosheva is an associate professor in the Department of Statistics and the School of Social Work, and a core faculty member in the Center for Statistics and the Social Sciences (CSSS). Dr. Erosheva received her PhD in statistics from Carnegie Mellon, and joined the University of Washington in 2002. Her research focuses on the development and application of modern statistical methods to address important issues in the Social, Medical, and Health Sciences.

Ryan Acton

Ryan Acton, UMass Amherst

Salience and Perspective: A Return to Classical Balance Theory
Friday, April 13, 2012 • 12:30PM–2PM • Lunch provided
Computer Science Building, Room 150/151

Abstract: The theory of balance in interpersonal relations, developed by psychologist Fritz Heider (1946), sought to link a cognitive process within individuals with the formation and dissolution of signed social ties with others. Put succinctly, balance theory is a theory of social dynamics such that people are motivated to make changes to their relations with others in situations that produce cognitive dissonance (as when a friend of an enemy is a friend). Since Heider, numerous scholars have empirically tested, refined, and generalized notions of balance. This work revisits the details of Heider's balance theory, some of which are often missing in contemporary work on balance. Specifically, I model balance from the perspective of individuals embedded in networks, and thereby break from sociological tradition of starting with the generalized notion of balance of Cartwright and Harary (1956). By utilizing signed, relational data from an online community, I study the formation process of several kinds of dyadic and triadic configurations, and find mixed support in the extent to which actors behave in balance theoretic ways. Results and implications are discussed.

Bio: Ryan M. Acton earned his M.A. in Demographic and Social Analysis (2006) and Ph.D. in Sociology (2010) from the University of California, Irvine as a member of the Networks, Computation, and Social Dynamics lab under the supervision of Dr. Carter T. Butts. In 2010 he joined the Sociology Department and the Computational Social Science Institute at the University of Massachusetts Amherst. Trained primarily in methods and theories for social network analysis, Acton's work focusses on the collection and analysis of social behavioral data from web-based sources. His current work revisits the core intuitions behind balance theory using contemporary, large-scale data from users of online social networks.

A Social Network Approach to Political Party Dynamics
Friday, April 20, 2012 • 12:30PM–2PM • Lunch provided
Computer Science Building, Room 150

Abstract: A developing branch of the literature on political parties has begun to conceptualize the party as a network. In this paper, I argue that this is an incredibly useful and important theoretical framework, but that we still lack a carefully worked out set of implications that could be applied to social network methodology. I argue that, just as with causal inference in other network research, we need to consider rival explanations for empirical patterns, and that these rival explanations are similar to those that confound other network research. Otherwise, empirical work on parties as networks will not be able to test the theory of the party as a network. In this project, I am working on sorting out these issues in the context of the case of party endorsements for president. Substantively, this problem extends work by Cohen, Karol, Noel and Zaller (2008). The CKNZ argument is, in flavor, an argument about informal networks whose influence shapes the choice of a candidate. But the empirical analysis in CKNZ does not make use of explicit network structures. This work takes those theoretical claims and examines them in network data.

Bio: Hans Noel is an assistant professor of Government at Georgetown University, where he studies political coalitions, political parties and ideology, with a focus on the United States. He is the co-author of The Party Decides: Presidential Nominations Before and After Reform (Chicago 2008) and the author of a book on the role of ideology in party politics, which is currently under review. Noel has also written several articles on parties, networks and ideology in the Journal of Politics, Social Networks, Party Politics, and other journals.

Combining Human Judgments in General Knowledge and Forecasting Tasks [video]
Friday, April 27, 2012 • 12:30PM–2PM • Lunch provided
Computer Science Building, Room 150/151

Abstract: In this research, we build on ideas from cognitive science and machine learning to build aggregation models that combine human judgments in general knowledge and prediction tasks. We propose that a successful approach to the aggregation of human judgment requires a cognitive modeling framework that explains how individuals produce their answers, and that also allows for individual differences in skill and expertise of participants. In addition, we argue that it is essential to correct for any systematic distortions in human judgment when aggregating judgments over individuals. We present two case studies that highlight our overall approach. In our first case study, we present preliminary results from the Aggregative Contingent Estimation System (ACES), which is part of a project funded by the Intelligence Advanced Research Projects Activity. The goal is to develop new methods for collecting and combining forecasts of many widely-dispersed individuals in order to increase aggregated forecasts’ predictive accuracy. In the second case study, we apply a cognitive modeling approach to the problem of measuring expertise on ranking tasks involving general knowledge (e.g., ordering American holidays through the calendar year) and forecasting (e.g., predicting the order of football teams at the end of the season). Using a Bayesian model of behavior on this problem that allows for individual differences in knowledge, we are able to infer people’s expertise directly from the rankings they provide and without using any knowledge of the true answer.

Bio: Mark Steyvers is currently a Professor in the department of Cognitive Science and also has an appointment in the department of Computer Science at UC Irvine. He received his PhD from Indiana University and worked as a postdoctoral associate with Josh Tenenbaum at Stanford University. His recent work is on developing computational models for aggregating human judgments, including probability estimates, rankings, and problem solving behavior. He also works on probabilistic topic modeling, information retrieval, and knowledge representation. For his computational modeling work on understanding human activity, Dr. Steyvers received the New Investigator Award from the American Psychological Association as well as the Society of Experimental Psychologists.

Past CSSI Seminars :

Fall 2011

Spring 2011