Tana Johnson May 7: "Contemporary Dynamics among Industrialized & Developing Countries"

Minnesota International Relations Colloquium is excited to host Prof. Tana Johnson of Duke University on May 7, at 3.30pm. She will be presenting a paper entitled "Contemporary Dynamics among Industrialized and Developing Countries: Textual Analysis Using Machine Learning". This presentation will close out the MIRC schedule for the spring semester.

Please find below the abstract, and we will circulate the paper mid-week upon request (email MIRC). As always, the colloquium will take place in Lippincott Room in Social Sciences, and coffee will be served.

We hope to see many of you there,
MIRC Organizers

Contemporary Dynamics among Industrialized and Developing Countries: Textual Analysis Using Machine Learning

Inter-governmental organizations (IGOs) face numerous challenges. Many deal with more than one policy area, include states that are equal in sovereignty but quite unequal in characteristics such as development level, and involve emerging economies that have grown richer rapidly. For these and other reasons, it can be difficult to deductively predict what various states will do in international fora. We show how more inductive, machine-learning techniques can help. With 3,678 paragraphs of government statements made between 1995 and 2012 in negotiations at the nexus of environmental and trade policy within the World Trade Organization (WTO), we use text as data to answer three questions: 1) which topics are states discussing? 2) are some topics strongly associated with poor states, and others with rich states? 3) as the “BASIC” states of Brazil, South Africa, India, and China have grown richer since the 1990s, have they increased or decreased their emphasis on particular topics? Our Structural Topic Model produces three findings that are likely to hold in other IGO contexts – but are unlikely to be revealed by purely deductive approaches. First, institutional mandates can be incomplete, or even misleading, for anticipating what states actually discuss. Second, some (but not all) discussions are associated with characteristics of the speakers themselves. Third, the BASIC states are not moving in lockstep. We demonstrate the promise – and detail the process – of using computer-assisted approaches to investigate the numerous IR research contexts in which theoretical expectations are opaque or incomplete.
© 2015 Regents of the University of Minnesota. All rights reserved. The University of Minnesota is an equal opportunity educator and employer. Privacy Statement