All Times EDT
Keywords: International Conflict, Machine Learning, Network Science
There is a rich tradition in Political Science of quantitatively modelling topics in international politics, such as international: trade, alliances, legal disputes, conflict, and more. Indeed, the majority of modern scholarship in international politics draws upon and benefits from cutting-edge methods in data science. In my dissertation, drawing on techniques from machine learning and network science, I produce novel estimates of every country's sentiment toward their international standing from 1816-2012. While generally understood as consequential, a country's satisfaction or dissatisfaction with the distribution of valued international goods is also fraught with measurement challenges. Issues include, but are not limited to: class imbalance in variables of interest, evaluating model quality in an unsupervised context, and non-independent observations. Beyond its value for subsequent scholarship, the measure also demonstrates considerable ability to predict the onset of international conflict, providing conceptual focus about current trends in world politics.