Apr. 20 2009 2:00PM

Apr. 20 2009 3:30PM

Counterfactual reasoning is central to many scientifc desciplines. For example, if we perform a statistical experiment to determine if working during the night causes cancer, then we face a great philosophical problem: we cannot study persons that work at night as well as at day time. Instead we study groups of people that either work at day or night time. But the conclusion that we wish to infer is counterfactual: would the person have developed cancer if he had not worked at night *given* that he worked at night. Such statements have so far been impossible to analyse formally, but new work on Causality pioneered by Pearl is very promising. In this talk I will introduce Shpitser and Pearl’s UAI 2007 article by the above name. I will start of with a gently introduction to causality, and then try to explain the algorithms of the paper in detail.

Pravidelný seminář Inteligentní systémy (Kontaktní osoba: Tomáš Kroupa)