Learning From The Data: A Theory Without Guessing

Department of Economics and Decision Sciences

Speaker:  Karl Schlag (Uni Vienna)

Room T-024

Abstract : 

A decision maker has to choose an action and observes a data set (small or large) that shows how different actions performed in the past. There is an exogenous mixed action that they would choose if they had no data set. The decision maker wishes to be better off with the data than without it in the environment they face. Standard (Bayesian) use of data generically will not satisfy this property as they "guess" which environment they face and hence might be wrong. We show what the decision-maker should choose when there are two actions.