Change search
ReferencesLink to record
Permanent link

Direct link
A Decision-Theoretic Framework Using Rational Agency
KTH, School of Computer Science and Communication (CSC), Numerical Analysis and Computer Science, NADA.ORCID iD: 0000-0002-2677-9759
2002 (English)In: Proceedings of the 11th Conference on Computer-Generated Forces and Behavioral Representation, 2002, 459-463 p.Conference paper (Refereed)
Abstract [en]

Maximizing expected utility has been the foremost method for decision-making for centuries and has been applied to numbers of decision tasks of various kinds. The ideas of using utility matrices in a tree structure to predict behavior among intelligent agents is however new with several contributions during the last decade. We have investigated such a decision-theoretic framework, the Recursive Modeling Method, which is originally applied within intelligent agents. This framework includes a data structure that holds information regarding the surrounding environment, and a model for computation that takes advantage of the mentioned data structure. The data structure is based on utility matrices used for storing information regarding preferences, the environment and other agents. These utility matrices are organized in a tree structure that also contains probability distributions representing beliefs regarding the current situation picture. The probability distributions are used recursively together with the utility matrices in order to solve decision tasks. We conclude that the investigated framework needs to be extended to be fully functional for Command and Control decision-making. Therefore we outline an extended framework where we introduce the “attribute domain”, which will be used throughout the model. The main idea is to keep track of different utility variables, one for each attribute, throughout the recursive process so that information can be used for various decision tasks. We believe that different utility functions will be used from time to time and therefore the utility cannot be combined into one single variable. Instead the data structure must be designed to hold sets containing one utility value for each attribute, rather than one single utility value describing all kinds of profit.

Place, publisher, year, edition, pages
2002. 459-463 p.
Keyword [en]
Decision Theory, Game Theory, agent, decision-making, multi-attribute, Command and Control, rationality
National Category
Computer Science
URN: urn:nbn:se:kth:diva-89606OAI: diva2:503359
The 11th Conference on Computer-Generated Forces and Behavioral Representation
NR 20140805Available from: 2012-02-15 Created: 2012-02-15 Last updated: 2012-02-15Bibliographically approved

Open Access in DiVA

No full text

Search in DiVA

By author/editor
Brynielsson, Joel
By organisation
Numerical Analysis and Computer Science, NADA
Computer Science

Search outside of DiVA

GoogleGoogle Scholar
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

Total: 27 hits
ReferencesLink to record
Permanent link

Direct link