Prof. Fredrik Sävje works on theoretical methodology and causal inference in the social sciences in general, and political science in particular. His work tends to be theoretical, focusing on describing the behavior of methods with precise and provably correct statements under known scope conditions, often with an eye towards foundational questions. Yet, his work is closely inspired by problems or questions in applied empirical practice, and is motivated by interactions with applied researchers. Prof. Sävje’s recent research concerns randomized experiments, with a particular focus on interference between experimental subjects and on experimental design.
What do you do with data science?
My work is currently focusing on methodology for randomized experiments. My main focus is so-called interference, which is when experimental units interact with each other during the experiment. The conventional approach to analyzing experiments assumes that no such interactions occur. However, interactions are commonplace in the social and medical sciences. For example, it is not reasonable to assume that information does not spread between participants when a political scientist investigates the effect of an informational campaign on voting behavior. Similar, it is not reasonable to assume that a virus does not spread between participants in a vaccine trial. In my work, I develop frameworks and methods that allow researchers to conduct studies when such interactions occur. This work tends to be foundational in nature in that it often require an overhaul of conventional frameworks (not only methods and approaches). Beyond my main interest, I’m working on other methodological questions related to randomized experiments (e.g., experimental design) and causal inference more generally.