My research goal is to combine the three major sub-fields of the Computational Social Science, including Social Media Data Analysis, Network Analysis, and Agent-based Modeling, to study various social science problems. Below you can find a summary of my past and current projects.
Identifying and Characterizing US Domestic Ideological Extremists
Between 9/11 and 2015, terrorist attacks by US domestic extremists have killed nearly twice as many people as those by Muslim Jihadists. Counter-terrorism strategies demand more systematic study of these homegrown extremists, important policy issues related to them, and ways to address those policy issues by using innovative approaches. I am collaborating with Professor Michael Macy from Cornell University, Dr. Ingmar Weber from Qatar Computing Research Institute, and Professor Claudio Cioffi-Revilla from George Mason University to identify active online individual extremists in the US and characterize the psychological and personality correlates of them using language expressed on Twitter. Our list of psychological features includes social relationships, disengagement, emotions, psychological engagement, anxiety, and anger. Moreover, we seek to develop a data-driven extremism lexicon that can be used in future research to identify individual extremists based on the content they post on Twitter.
Co-following Network of the US Domestic Extremist Groups on Twitter
Spatial Social Networks
While the importance of spatial proximity encourages us to consider the geographical properties of the individuals into the models, the availability of geo-coded data from social media enables us to validate the results of our computational models. Hence, incorporating the geographical properties of the agents into computational models of social behavior is a promising extension which helps us to analyze the spatial correlates of a given complex behavior. Co-authored with Professor Claudio Cioffi-Revilla and Dr. Andrew Crooks from George Mason University, I proposed spatial versions of the three classic network models including the Erdös-Rényi, Watts-Strogatz, and Barabási-Albert (Alizadeh, Cioffi, & Crooks 2016). We assume that nodes have geographical coordinates which are uniformly distributed over an m × m Cartesian space, and long-distance connections are penalized. This project aims to build a foundation for developing more spatially explicit agent-based models of social applications.
Connection Probability (P) is Penalized with Distance (d)
Intergroup Relations and Opinion Dynamics
Often policy makers and analysts are interested in predicting the effect of major events on the public opinions. Intergroup conflict escalation and in-group favoritism are two social processes that get triggered by some social and political events. For example, one might argue that the case of Ferguson Police Shooting triggered the in-group favoritism mechasnim among African-Americans. Building on the social identity approach to intergroup relations, I examined the effect of intergroup conflict escalation (Alizadeh et al 2014) and in-group favoritism (Alizadeh et al 2015) on the collective behavior of individuals’ opinion and radicalization. These two projects bring the empirical findings from social psychology and agent-based models of the opinion dynamics together to help public opinion policy analysts to get a better understanding of the effect of intergroup conflict escalation and in-group favoritism on the macro-level patterns of indivduals’ opinions.
Social Identity Approach to Intergroup Relations
Activation Regimes in Opinion Dynamics
Empirical evidences have supported the large heterogeneity in the timing of individuals’ activities. Moreover, computational analysis of the agent-based models has shown the importance of the activation regimes. In this project (Alizadeh & Cioffi-Revilla 2015), I applied four different asynchronous updating schemes including random, uniform, and two state-driven Poisson updating schemes on an agent-based opinion dynamics model. I compared the effect of these activation regimes by measuring the appropriate opinion clustering statistics and also the number of emergent extremists. The results exhibit both qualitative and quantitative difference between different activation regimes which in some cases are counterintuitive. In particular, I found that exposing the radical/moderate agents to more encounters decreases/increases the average number of extremists compared to other types of activation regimes.
Geography of Political Opinion Extremism in the US
I used spatial statistics and Geographical Information Systems (GIS) techniques to study the socio-economic and personality correlates of opinion extremism at the state-level in the US using the 2012 American National Election Study (ANES) time series data. Here, I used the Geographically Weighted Regression model to provide specific local regression estimates for each state with respect to association between opinion extremism and correlates of interest.