Challenges in Causality
Google Tech TalksFebruary, 11 2008ABSTRACTWhat affects your health, the economy, climate changes? And what actions willhave beneficial effects? These are some of the central questions of causaldiscovery. A "causal model" is a model capable of making predictions underchanging circumstances, corresponding to actions of "external agents" on asystem of interest. For example, a doctor administering a drug to a patient, agovernment enforcing a new tax law or a new environmental policy. It is oftennecessary to assess the benefits and risks of potential actions using availablepast data and excluding the possibility of experimenting. Experiments, whichare the ultimate way of verifying causal relationships, are in many cases toocostly, infeasible, or unethical. For instance, enforcing a law prohibiting tosmoke in public places is costly, preventing people from smoking may beinfeasible, and forcing them to smoke would be unethical. In contrast,"observational data" are available in abundance in many applications. Recently,methods to devise causal models from observational data have been proposed. Cancausal models thus obtained be relied upon to make important decisions? In thispresentation, we will challenge the hopes an promises of causal discovery andpresent new means of assessing the validity of causal modeling techniques.Want to play? Check the "causation and prediction" competition presently goingon: http://www.causality.inf.ethz.ch/challenge.phpDeadline April 30, 2008Speaker: Isabelle GuyonIsabelle Guyon is a researcher in machine learning and an independentconsultant. Prior to starting her consulting practice in 1996, sheworked at AT&T Bell Laboratories, where she pioneered applications of neuralnetworks to pen computer interfaces and invented Support Vector Machines (incollaboration with B. Boser and V. Vapnik). Isabelle Guyon holds a Ph.D. degreein Physical Sciences of the University Pierre and Marie Curie of Paris, France.She is vice-president of the Unipen foundation, action editor of the Journal ofMachine Learning Research, and competition chair of the IJCNN conference.