Statistics Seminars: Utilityless decisions
5 December 2008 14:00 in CM221
In this talk, I will question the usual strong (and often unrealistic) assumptions which most theories of decision take as a prerequisite. In order to relax these assumptions, one approach is to go back to decisions in a for that is as simple as possible. Traditionally, this is taken to be a preference ordering, however I will argue that choice functions a more suitable means for modelling decisions, especially when uncertainty is serious. Finally, I will study the minimal structure required to arrive at the modelling of decision trees, and examine necessary and sufficient conditions on choice functions under which normal form solutions can be derived by backward induction. Surprisingly, this turns out to be possible without ever invoking any concept of utility.
Contact email@example.com for more information