Katherine A. Keith

  kak5 (at) williams (dot) edu

  curriculum vitae (CV)

I am an Assistant Professor in the Computer Science department at Williams College. Previously, I was a Postdoctoral Young Investigator at the Allen Institute for Artificial Intelligence. I received a PhD from the Manning College of Information and Computer Sciences at the University of Massachusetts Amherst in August 2021, and I am grateful to have been supported by a Bloomberg Data Science PhD Fellowship 2019-2021.

My work lies at the intersection natural language processing—programming a computer to understand, classify, and generate human language—and data science—finding patterns in data that help us better understand the world. Specifically, I work on methods and applications that help make quantitative social science conclusions from large-scale text-based datasets, aligning with the subfields of text-as-data and computational social science. I focus on language because it is one of the most salient forms of expressing thoughts, sharing ideas, and recording history. My work often involves collaborations with other scholars in political science, legal studies, education, sociology, and other social science fields.

One of my primary research goals is to expand the set of text-specific causal estimation designs available to data scientists. Causal inference is the theoretical framework underlying policy and interventions: what would happen to variable A if we intervened and set variable B to something different? I focus on causal inference (primarily using non-experimental/observational datasets) because causal inference frameworks allow us to make our assumptions explicit before moving to statistical estimation.

Incorporating variables with text into causal inference is difficult because language is high-dimensional and complex, and often even our best natural language processing methods result in measurement error. Improved text-based causal methods could help many disciplines that want to understand the mechanisms of human behavior that intersect with speech or text communication.

Outside of research, I am also interested in projects that build bridges between natural language processing and the social sciences. In line with this goal, Ian Stewart and I organized the NLP+CSS 201 Online Tutorial Series and I have helped organize the NLP+CSS Workshops. I also hosted the podcast Diaries of Social Data Research with Lucy Li in which interdisciplinary scholars described the "behind the scenes" diaries of their published papers and tips for success.