YUNKYU SOHN FPSE WASEDA UNIVERSITY
Thank you for visiting my website. I am Associate Professor at Faculty of Political Science and Economics, Waseda University. My primary research interests are developing methodological tools for understanding large-scale, unstructured, dynamic datasets and enhancing statistical models for answering core questions in social sciences by incorporating recent advances in data science. My substantive interests lie on legislative politics, polarization and collective action. I am open to various types of methodological approaches including behavioral models and large-scale psychological and economic experiments.
You can reach me via ysohn_at_aoni.waseda.jp.
Jong Hee Park, Yunkyu Sohn.
Bayesian Analysis (advance access)
We propose a hidden Markov latent space model for longitudinal networks using Bayesian inference.
Yunkyu Sohn, Jong Hee Park.
Network Science 5 (2017)
We propose a multi-layer stochastic blockmodeling framework that uncovers layer-common node traits and factors associated with layer-specific network generating functions. The method uncovers common coalition structure for a pair of mutualistic and antagonistic interaction networks.
Yunkyu Sohn, Myung-Kyu Choi, Yong-Yeol Ahn, Junho Lee, Jaeseung Jeong.
PLoS Computational Biology 7, no. 5 (2011): e1001139.
The paper proposes a novel statistical framework for graph clustering with non-topological constraints.
Computational Social Science
Rooni Lee, Yunkyu Sohn, Wonjae Lee.
Leonardo 50 (2017)
We characterize the ideological heterogeneity in Fluxus, the first artistic collaboration movement recorded in the history, and disprove the long-held leader-follower framework in art history, called the Maciunas-based Paradigm.
Jaimie Y. Park, Yunkyu Sohn, Sue Moon.
International Conference on Web and Social Media 2016
We validate the effectiveness of friend-tagging as a new targeted and personalized online advertising strategy as opposed to traditional paid advertising methods.
World Wide Web 2014
Lorenzo Coviello, Yunkyu Sohn, Adam DI Kramer, Cameron Marlow, Massimo Franceschetti, Nicholas A. Christakis, James H. Fowler.
PloS One 9, no. 3 (2014): e90315.
With data from millions of Facebook users, we show that online social networks may magnify the intensity of global emotional synchrony.
Experimental Political Science
Yunkyu Sohn, Jung-kyoo Choi, T.K. Ahn.
Journal of Complex Networks (advance access)
By conducting pairwise prisoner's dilemma experiments with partner selection option, where subjects had freedom to interact with multiple individuals, we show conditional cooperation/defection strategy and partner selection strategy of the players, and structural characteristics associated with their strategy types.
Jooa Julia Lee, Yunkyu Sohn, James H. Fowler.
PloS One 8, no. 12 (2013): e83143.
A series of social psychological experiments conducted online and offline demonstrate a critical link between emotion regulation and political attitudes.
Philosophical Forum 34: 171-197. (2006)
The paper provides a functionalistic rationale for indicating common mental properties (e.g. pain) of different species. (in Korean)
Resolving Identification Problem in Multidimensional Item Response Theory Model [PDF]
I resolve the prevalent identification problem in 2 parameter normal ogive multidimensional item response theory model using a matrix variate distribution called Matrix von Mises-Fisher distribution, and propose a reasonably fast fully Bayesian estimation algorithm. The method produces highly accurate estimates of the convex combination weight parameter in the quadratic-normal spatial voting model.
Dimensionality and Detectability of Voting Blocs
I propose a coalition-driven voting model for legislatures, and prove that the minimum number of dimensions required to represent the preference of N coalitions in the euclidean space is N-1. I append this finding using the random matrix ensemble approach, and derive the detectability threshold for discrimination of the voting blocs, as a function of the binomial intra-coalition and inter-coalition voting parameters, and the size of the voting matrix.
Automatic Dimensionality Selection Algorithm for Linear Factor Models: Evidence from Random Matrix Theory
I propose a Hidden Markov Model approach for the automatic discovery of noise-free spectral bands in the eigenspectra of linear factor matrices including covariance matrices and graph Laplacians. The proposed method corrects flaws in the popular automatic discovery method based on profile likelihood comparison, and replaces the conventional elbow-detection visual inspection scheme. I demonstrate the performance of the method using image, network, voting and educational testing datasets.
The paper proposes an ideal point estimation method for non-voting datasets under coalitional game theoretic foundation.
Yunkyu Sohn, Jeongin Ju, Haewoon Kwak, Sue Moon.
We provide meta-analysis of partition spaces of various network community detection (i.e. clustering) methods, and examine critical factors affecting solution heterogeneity. We also find a pervasive over-fitting problem associated with the most popular modularity-based agglomerative algorithms.
Ideological Preference, Issue Dimensions and Party Polarization in Korean National Assembly from 1948 to the Present
The paper shows issue-specific ideological spectrum of Korean National Assembly from its birth to the present.
On the Measurement of Party Polarization [PDF]
Yunkyu Sohn, James H. Fowler.
We discuss axiomatic features of party polarization and validate the performance of a number of measures. A newly proposed metric for matrix segregation outperforms all other metrics of polarization. We identify distortion in party polarization trend of US Senate and House roll call voting records captured by conventional measures.
This paper provides two types of evolutionary equilibrium analysis of Palfrey-Rosenthal voter turnout game. The deterministic stability analysis shows that all equilibrium turnout solutions derived in the previous studies are unstable and extremely vulnerable to the invasion of idiosyncratic voting propensity. In contrast, frequency dependent stochastic learning formulation provides steady state equilibrium solutions which are largely in accord with actual patterns of turnout.
Parabolic Small World Effect: A Trade-off Effect of Cohesion and Brokerage on Information Diffusion
Wonjae Lee, Jungmin Kim, Yunkyu Sohn, Haewoon Kwak.
We analyze political Tweets of 6 million Korean users during the 2012 presidential election season . Our analysis suggests that the absolute value of tweet volume is not necessarily correlated with its level of endogenous proliferation, but the proliferation is simultaneously associated with within-community coherence and inter-community connectivity.
Education and Employment
Associate Professor (Tenure-Track), Faculty of Political Science and Economics, Waseda University, 2018-
Pre/Postdoctoral Research Fellow, Department of Politics, Princeton University, 2016-2018
Ph.D. Political Science, University of California San Diego, 2017
Compulsory Military Service at Computer Science Division of Korea Advanced Institute of Science and Technology, 2012-2015
Predoctoral Research Fellow, Institute for Quantitative Social Science, Harvard University, 2011-2012
M.S. Computational Neuroscience, Korea Advanced Institute of Science and Technology, 2008
B.A. Philosophy, Seoul National University, 2007
B.S. Physics, Seoul National University, 2007
© 2015 by Yunkyu Sohn