In my research, I address fundamental questions about causation and causal inference using recently developed causal modeling methods. As a result of collaborations between philosophers, social scientists and computer scientists, it is now possible to answer previously unresolved questions about the relationships between causation, policy interventions and evidence. Yet there remains much work to be done in understanding the implications of these developments for philosophy of science and epistemology. Additionally (and independently) causal modeling methods often rely on idealized assumptions about one’s data that are known not to obtain in practice. There are consequently obstacles to showing how causal models are nevertheless useful for working scientists. The challenges of linking causal modeling to general philosophy of science and to the practice of science are typically addressed separately. I aim to show that they are related and that clarifying their relations is fruitful to both philosophers and scientists.
Over the past year I have pursued three projects with this aim in mind. First, I have explored the epistemological basis for principles for selecting among the causal models that are compatible with a probability distribution. This exploration resulted in two papers: one presently under review and one coauthored paper forthcoming the British Journal for the Philosophy of Science. The latter introduces a novel principle for choosing among causal models, compares it to existing parsimony principles in the literature, and argues that the best-known such principle (‘Faithfulness’) turns out not to be a parsimony principle. Second, I have developed a philosophical analysis of the concept of a path-specific effect – i.e. the effect of one variable on another as it is transmitted by intermediate variables. In one paper under review, I argue that probabilistic accounts of causality were unable to provide an account of such effects, which can instead be explicated using structural equations models. In another paper, which I will present at the biennial meeting of the Philosophy of Science Association in November, I identify the explanatory value of measuring path-specific effects, and draw a contrast between my account and accounts of mechanistic explanation. Third, I have worked long-distance with a reading group on issues related to causation and time. We are currently writing a paper arguing that in a diverse range of physical systems, the causal relationships that obtain among a set of variables are relative to the time-scale at which the system is evaluated.
The two sciences that I am presently most engaged with are neuroscience and epidemiology. This past June, my colleague Matteo Colombo and I presented a paper in Aarhus on modes of connectivity in the brain. We argued that the widespread belief that knowledge of anatomical information of the brain is essential for mapping the causal relations among brain regions is not supported by current scientific practice. We believe that there are potentially fruitful ways of using spatiotemporal information to build causal models of the brain, but that in order to do so one must first clarify the relationship between causal and temporal relations. My work on epidemiology is in collaboration with an epidemiologist at Tufts medical school and a public health informatics professor at Boston University. We aim to develop a framework that clarifies the concept of causation in epidemiology and provides guidance for scientists tracing the spread of disease in real time. We have a contract with Springer press to publish a book on this topic, which is tentatively titled The Philosophy of Causal Inference in Public Health Informatics.
Recent developments in the analysis of causal concepts have potential applications even beyond traditional philosophy of science. Last winter I presented colloquium at Tilburg University on using causal models to understand the concepts of racial and sexual discrimination. This coming fall I will be giving a presentation in Ghent on how to understand the causal role of future-directed attentions in action. These two projects illustrate the range of areas that can benefit from having clearer causal concepts.
I anticipate that in the next decade research into causation will lead to progress on several topics of practical and philosophical importance. Here are a few of the most difficult open questions that causal modeling can help to answer: How can we determine if a policy intervention that works in one context will work in another? What constraints are there on which variables can figure in a causal relationship, and how do we find and measure variables that are better for causal inference? Which assumptions made in causal modeling and statistical sampling are indispensible for acquiring causal knowledge, and which may be relaxed? I choose new projects with an eye towards answering these questions. My background in philosophy of science and causal modeling makes me well suited to address them and I am eager to collaborate with other scholars to expand the range of topics I can cover and make my research accessible to as wide an audience as possible. The next few years should be exciting.