Introducing the latest in Data Science,

focusing on applications in social,

political and policy sciences.


2023/11/02 【Nov. 03】Using Machine Learning Methods to Classify and Compare State Repressive Actions

Dr. Rebecca Cordell, University of Texas at Dallas
Date: Friday, November 03 - 10:00AM~13:00PM (Taiwan Time, GMT+8); Thursday, November 02 - 21:00PM~00:00AM (US Central Time, GMT-6).
Topic: Using Machine Learning Methods to Classify and Compare State Repressive Actions 
Abstract: Why do states favor some forms of repressive actions over others? Human rights violations of physical integrity typically take the forms of imprisonment, torture, killing, and disappearances. In order to fully understand repression dynamics and the full scope of physical integrity violations, it is helpful to understand how states may vary in the use of particular forms of abuse. We present and evaluate an automated machine learning method that produces categorical information about the type of physical integrity rights violations actor responsible for the violation described in annual human rights reports using a supervised machine learning approach. The coding process builds on an existing text as data corpus of human rights allegations which extracts specific allegations of physical integrity rights violations from the original text of country reports on human rights. Each allegation sentence presents information about a specific type of physical integrity violation (disappearances, torture, killing and political imprisonment), the type of actor responsible for the violation (i.e., state versus non-state), location of the violation, and intensity of the violation. We present results for our violation type and actor variables using various machine learning algorithms and use cross validation to assess model performance. The results show that this automated method achieves high degrees of accuracy, with significant overlap between the classes assigned by the machine and the human coded data. We then provide an empirical illustration of how repressive repertoires change over time. Specifically focusing on the early period of the global war on terror (2001-2005), we show that states linked to the US extraordinary rendition program became more likely to employ torture as a more covert repressive measure.
Bio: Dr. Cordell is an Assistant Professor of Political Science in the School of Economic, Political and Policy Sciences at the University of Texas at Dallas. Prior to that, she was a post-doc in the School of Politics and Global Studies at Arizona State University. Dr. Cordell completed her Ph.D. in Political Science from the Department of Government at the University of Essex in 2017. Dr. Cordell studies the causes, dynamics and consequences of state repression, human rights and political violence using computational methods and quantitative text analysis. Her research has been published in International Studies Quarterly, Journal of Conflict Resolution, Journal of Peace Research, International Interactions, and Journal of Human Rights. Dr. Cordell has shared insights from this work with the public in articles for The Monkey Cage, Political Violence @ A Glance, and The Conversation. In 2018, Dr. Cordell received the ISA Human Rights Section's Steven C. Poe Best Graduate Student Paper Award.
Learn more about Prof. Rebecca Cordell at her website.