Text-as-Data and Historical Foreign Policy Decision Making

Session Submission Type: In-Person Full Paper Panel

Session Description

The advent of text-as-data methods for social science coupled with the new availability of primary source historical material provides a compelling opportunity for IR scholars to bring new methods and data to the study of classic questions related to foreign policy. This panel convenes a set of innovative works in this emerging research program, demonstrating a range of creative techniques and novel findings. The first two papers investigate patterns of bureaucratic speech to understand how national security policymaking organizations carve out niches and frame concepts consistent with their own interests and expertise. Eric Min and Austin Carson examine bureaucratization in the U.S. intelligence community by charting the lexical diversity of their reporting to senior officials. Using a new dataset of the President’s Daily Briefs, the authors demonstrate that intelligence agencies craft the language in their reporting to acquire influence through the impression of technical sophistication. Don Casler explores how officials from different national security bureaucracies within the U.S. government conceive of credibility, which is often believed to the essential currency of international politics. Drawing on records from the Foreign Relations of the United States and Declassified Documents Online collections, he shows that military and diplomatic officials harbor divergent understandings of what credibility means, which in turn has implications for how each group of officials advocates for policies related to the use of force. The second pair of papers broaden the conversation by considering how states use speech and rhetoric to establish or signal political preferences. Erik Voeten, Michael Bailey, and Cathy Lee develop and test a model to estimate state preferences from United Nations Security Council speeches. They combine features from two popular wordscoring models to demonstrate that countries have not adopted rhetoric preferred by China over that favored by the United States, suggesting that China remains relatively isolated at the UN. Lastly, Sarah Dreier, Sofia Serrano, Emily Gade, and Noah Smith advance a method of natural language processing known as word embeddings that deal more systematically with complex topics requiring contextual interpretation, otherwise known as “know-when-I-see-it” concepts. They use this method to parse archived British government rationales for internment without trial during the Troubles in Northern Ireland. While these papers only scratch the surface of what we can learn by deploying new methods against timeless subjects, this panel will speak to various historiographic debates, engage historically-minded IR and foreign policy scholars, and address key theoretical questions by accessing the “inside view” of decision makers’ internal perceptions (through primary materials) rather than what is intended for public consumption (through published statements or web-scraped media).

Sub Unit

• Division 19: International Security 

Individual Presentations

• Keeping it Professional: Lexical Diversity and Bureaucratizing National Security - Eric Min, UCLA; Austin Carson, University of Chicago

• The Organizational Semantics of Credibility in International Relations - Donald Casler, Columbia University

• Estimating State Preferences from United Nations Security Council Speeches - Erik Voeten, Georgetown University; Michael A. Bailey, Georgetown University; Cathy Sona Lee, Georgetown University

• Troubles in Text: Finetuning NLP to Analyze Rationalization of Rights Violations - Sarah K. Dreier, University of New Mexico; Sofia Serrano; Emily K. Gade, Emory University; Noah A. Smith, Carnegie Mellon University

Chair

• Justin Grimmer Stanford University

Discussants

• Rochelle Layla Terman University of Chicago

• Abby M Fanlo Stanford University