Statistics Seminars: Normal and Extensive Form Equivalence in Sequential Decision Problems
11 January 2010 14:15 in CM221
Sequential decision problems are usually expressed in one of two forms: the normal form, where the subject must specify all his actions in all eventualities in advance, or the extensive form, where the subject's choice at a decision node is determined only if that node is reached. When maximizing expected utility, it is well known that the standard normal form and extensive form solutions are equivalent (where "equivalent" means that the strategies implied by the extensive form solution are exactly the strategies of the normal form solution). When a different choice function on gambles is used, we can find normal and extensive form solutions as usual, but the two are not usually equivalent. I shall detail the properties a choice function must satisfy for equivalence to hold, and make links with related concepts such as Selten's subgame perfectness and Hammond's consequentialist theory.
Contact email@example.com for more information