Title :

How To Avoid Biases In Reactive Simulations

Author(s) :

Kubera, Yoann [Auteur]
Systèmes Multi-Agents et Comportements [SMAC]
Mathieu, Philippe [Auteur]
Systèmes Multi-Agents et Comportements [SMAC]
Picault, Sebastien [Auteur]
Systèmes Multi-Agents et Comportements [SMAC]

Scientific editor(s) :

Demazeau
Yves and Pavòn
Juan and Corchado
Juan Manuel and Bajo
Javier

Conference title :

7th International conference on Practical Applications of Agents and Multi-Agents Systems (PAAMS'2009)

City :

Salamanca

Country :

Espagne

Start date of the conference :

2009-03-25

Book title :

Proceedings of the 7th International conference on Practical Applications of Agents and Multi-Agents Systems (PAAMS'2009)

Journal title :

Proceedings of the 7th International Conference on Practical Applications of Agents and Multi-Agent Systems, PAAMS

Publisher :

Springer

Publication date :

2009

HAL domain(s) :

Informatique [cs]/Système multi-agents [cs.MA]
Informatique [cs]/Ingénierie, finance et science [cs.CE]
Informatique [cs]/Intelligence artificielle [cs.AI]

English abstract : [en]

In order to ensure simulations reproducibility, particular attention must be payed to the specification of its model. This requires adequate design methodologies, that enlightens modelers on possible implementation ambiguities ...
Show more >In order to ensure simulations reproducibility, particular attention must be payed to the specification of its model. This requires adequate design methodologies, that enlightens modelers on possible implementation ambiguities - and biases - their model might have. Yet, because of not adapted knowledge representation, current reactive simulation design methodologies lack specifications concerning interaction selection, especially in stochastic behaviors. Thanks to the interaction-oriented methodology IODA - which knowledge representation is fit to handle such problems - this paper provides simple guidelines to describe interaction selection. These guidelines use a subsumption like-structure, and focus the design of interaction selection on two points: how the selection takes place - for instance first select the interaction, and then select the partner of the interaction, or first a partner and then an interaction - and the nature of each selection - for instance at random, or with a utility function. This provides a valuable communication support between modelers and computer scientists, that makes the interpretation of the model and its implementation clearer, and the identification of ambiguities and biases easier.Show less >

Language :

Anglais

Peer reviewed article :

Oui

Audience :

Internationale

Popular science :

Non

Collections :

• Centre de Recherche en Informatique, Signal et Automatique de Lille (CRIStAL) - UMR 9189

Source :

Harvested from HAL