Humanities Faculty

Jiji Zhang

Assistant Professor of Philosophy

Research interests

Philosophy of Causation, Foundations of Causal and Statistical Inference, Epistemology

Research Statement

The primary research program I am currently engaged in --- with causal discovery and reasoning as its central subject --- has involved researchers from several different fields, including computer science, epidemiology, philosophy, psychology, social science, and statistics. As a philosopher, my participation in and contribution to this program is mostly motivated by my interest in the epistemology of causation and the methodology of causal inference. The major issue that concerns me is the possibility and extent of acquiring causal knowledge from statistical regularities, the kind of knowledge that supplies ground for causal explanation, counterfactual or subjunctive reasoning, and predictions of consequences of actions.

I take two inspirations from the philosopher David Hume in doing this. First, our beliefs about cause and effect have a lot to do with observations of statistical regularities (of which "constant conjunction" is but a special case). Second, to justify causal inference, Reason need be supplemented by extra assumptions, and the "right" assumptions may be imposed by nature. The first alludes to statistical causal inference. The second suggests a division of the question of justification into two projects. On the one hand, there is the normative, means-ends epistemology about what assumptions authorize what inference, or what kind of reliability can causal inference methods possibly achieve under what assumptions. On the other hand, there is the question of the epistemic status of various assumptions, which, in light of Hume's naturalism, should concern cognitive and developmental psychologists as much as they concern philosophers of causation. I am interested in both projects.

Besides the epistemology of causal inference, I have substantial interest in the so-called Bayesian epistemology and foundations of statistics. Recently I am concerned with the neo-Popperian, error-statistical theory of evidence, and how it can explain certain practices such as correction for multiple testing in statistical inference, as compared to a Bayesian approach. Inductive logic, of both the probabilistic brand and the non-probabilistic brand, is also part of my plan for future research.

Publications

Zhang, J. (forthcoming) From Correlation to Causation: A Completeness Result for Causal Discovery in the Presence of Latent Confounders and Selection Bias. Artificial Intelligence Journal.

Zhang, J. (forthcoming) "Error Probabilities for Inference of Causal Directions." Synthese.

Zhang, J. (forthcoming) "Causal Reasoning with Ancestral Graphical Models." Journal of Machine Learning Research.

Zhang, J. (2007) "A Characterization of Markov Equivalence Classes for Causal Models with Latent Variables." Proceedings of Uncertainty in Artificial Intelligence. (Runner-up for the best paper award.)

Zhang, J., and P. Spirtes (2007) "Detection of Unfaithfulness and Robust Causal Inference." Submitted to Minds and Machines.