Keynote

Title: Revealing the Cement of the Universe

Presenter: Professor Clark Glymour, Alumni University Professor of Philosophy at Carnegie Mellon University

David Hume, the 18^th century philosopher who called causation “the cement of the universe,” is conventionally cited in support of scepticism about causal relations, which was not his intent. Attributions to Hume, combined with a true but misleading slogan, “correlation is not causation,” and with R.A. Fisher’s influential conviction that only randomized experiments could reveal causal relations, made the very idea of algorithms that can discover causal relations from non-experimental data seem fanciful to most of the statistical community throughout the last century—the long history of causal discovery from purely observational data in astronomy, biology and climate science notwithstanding. The last 25 years have produced a revolution challenging that skepticism, with important new research from the United States and Europe appearing almost monthly, and actual and prospective applications to low and high dimensional data in many scientific domains. I will survey the basic ideas of several of these developments, and describe some empirical applications and open problems.

Biography:

Clark Glymour is a Senior Research Scientist at the Florida Institute for Human and Machine Cognition and Alumni University Professor of Philosophy at Carnegie Mellon University. He is the founding member of Carnegie Mellon’s Philosophy Department, anomalous in philosophy for its focus on statistics, decision theory, machine learning and mathematical logic.  Glymour has been a Guggenheim Fellow, Fellow of the Center for Advanced Study in the Behavioral Sciences, and a Phi Beta Kappa lecturer. His work with Peter Spirtes and Richard Scheines on automated methods for discovering causal relations spans four decades. His applied work has concerned regulation of gene expression, robotic mineral identification, ecology and forest fire prediction, and, currently, recovering neural causal connections from fMRI data.