Normative Decision Analysis. Assignment.

by Samuel Ferrer

1- Brief explanation of the content of the course

Normative Decision Analys is the study of the mechanisms used by people in taking decisions, supposing that the decision maker and the situation are rational. In this course, the bayesian view point and the utility theory are presented, and also an intent for axiomatization, in order to represent rationality. The Allais' paradox is issued in oposition to the utility and the sure-thing principle.The basic literature discused was "Decision, probability and utility" by Peter Gärdenfors and Nils-Eric Sahlin.

2- Select a problem

In order to grasp the basic concepts accepted so far in Decision Theory, a problem is suggested and solved using, at least, two different decision tools, global available. The idea of the problem (The minefield) I got it after reading the chapter 13 (Reinforcement learning) from the book "Machine Learning" by Tom M. Mitchell.

3- DELTA's view point

The problem is presented in a way it can be solved using the DDT software developed by Johan Walter. DDT uses the Delta method, developed by Dr. Mats Danielson in his Ph. D Thesis.

4- DATA's view point

This tool implements decision trees, influence diagrams and Markov's chains. The basic characteristic exploited is its implementation of decision trees, which is not covered by DDT. However, DDT can handle undesirability, which is an important issue in decision making, while this is not directly possible in DATA. DATA instead gives a lote of graphic information to the user, so it can do some deductions out of a roll-back report. DATA is produced by TreeAge.

5- Conclusions


Let me know your comments