Using Participatory Paradigm to Learn Human Behaviour

Patrick Taillandier (1) , Thanh-Quang Chu (1)

1 UMMISCO - Unité de modélisation mathématique et informatique des systèmes complexes [Bondy]

Since the end of the seventies, the utilisation of multi-agents simulations has spread out. A typical use of these simulations concerns the modelling of human behaviour. In this application case, a key point to ensure the simulation realism is the definition of the agent behaviour. Unfortunately, designing such behaviour is often complex. In order to help the definition of such behaviour, we propose an approach based on the participatory paradigm. In our approach, a human actor directly plays the role of an agent in the simulation. Then, the analysis of the execution logs allows then to extract knowledge about the human actor behaviour. In this paper, we propose to formalise its behaviour by means of utility functions. An experiment, carried out in the domain of rescue simulation, is presented. This first experiment shows promising results for our approach.

Mots clés: agent behavior design; participatory paradigm; multi-agent simulations

Domaines: Modélisation et simulation; Système multi-agents [cs.MA]; Apprentissage [cs.LG]

Patrick Taillandier, Thanh-Quang Chu. Using Participatory Paradigm to Learn Human Behaviour. International Conference on Knowledge and Systems Engineering, 2009, Hanoi, Vietnam. pp.55-60. ⟨hal-00689151⟩