My research examines how scientists confront complexity and uncertainty in biomedical research, with a particular focus on preclinical research with animal models. The assumptions that scientists have about the complexity of the phenomena they study—for example, how many genes are involved in a psychiatric disorder, or how closely a mouse navigating a maze resembles human anxiety—are as much cultural as they are scientific. These assumptions reflect broader social and political commitments, are transmitted through informal means and laboratory lore, differ between groups of scientists, and change over time.

My first book, Model Behavior, is an ethnographic study of an animal behavior genetics laboratory where researchers were using mouse models to investigate the genetics of psychiatric disorders. In this laboratory, researchers worked from the expectation that psychiatric disorders were complex, situationally specific phenomena that were likely to be influenced by a number of genetic and environmental factors. I demonstrate how these assumptions shaped their knowledge production work in counterintuitive ways. For example, while these researchers were professionally committed to studying genetics, in practice they spent much of their time creating the controlled conditions that they believed were necessary to make elusive genetic effects visible. This meant that they produced as much, if not more, knowledge about how environmental factors shaped behavior as they did about genetic factors. For this work I won a First Book Award from the UW–Madison Center for the Humanities.

My current book project examines the “reproducibility crisis” in biomedicine, a recent phenomenon where scientists have found many supposedly stable findings to be difficult to replicate on subsequent investigation. In one landmark study, a pharmaceutical company reported that they could only reproduce findings from the published literature on cancer in 6 out of 53 (11%) of cases. The reproducibility crisis involves a radical reevaluation of the certainty of scientific findings, broadly construed. Whereas the scientists I studied in Model Behavior attributed the uncertainties in their research to the fact that they worked on behavior, reproducibility problems do not appear to be confined to any particular field or technique. I was awarded a fellowship at the Radcliffe Institute for Advanced Study at Harvard University to begin work on this project in 2018–2019. I presented some early findings at my Radcliffe public lecture in early 2019.

I also do research on new technologies in oncology research and clinical practice. As an embedded ethnographer in a trial investigating resistance to chemotherapy in breast cancer, I observed researchers as they introduced genomic technologies into existing clinical research routines. This work was conducted in collaboration with Alberto Cambrosio and Peter Keating. Currently I am conducting research with Pilar Ossorio on machine learning algorithms in oncology, where algorithms are now routinely used to estimate a patient’s risk of reoccurrence of their cancer. Our project aims to understand 1) where predictive algorithmic tests sit on a spectrum from merely informing to effectively deciding a patient’s treatment course, and 2) how different practitioners (e.g., community oncologists versus those at academic medical centers, experienced versus new practitioners) use these tests.