The reflections regarding risk and industrial catastrophes question
the complexity theories. Besides the new concepts which lead us to question
older concepts (risk, hazard and vulnerability), we propose to think about
more practical aspects, for example the modelling of human behaviour and
the confrontation in crisis situations. The link between concepts as: critical
self-organization, emergence, bifurcation, and the methods in the Distributed
Artificial Intelligence used to model them is however difficult. In this paper,
we present ongoing analysis on the key concepts of risk science, such as the
hazards and the catastrophes. We propose to enrich them with complex systems
theories. First, we present methodological perspectives of the DAI, for
example multi-agent systems, and compare them with other simulation methods
used in the context of the risks. Secondly, we present the MOSAIIC model
(Modelling and Simulation of Industrial Accidents by Individual-Based methods)
which gives possibilities to simulate the behaviour of individuals during
an industrial accident. The project and the MOSAIIC model aim to explore
the effects of a major industrial accident on public health. For instance, the
emission and the spread of a toxic gas in an urban environment may be a serious
danger for the human health. Thus we propose to study the consequences
of this type of event in order to reduce the vulnerability of the populations.
In the model, we emphasize both on spatial and behavioral dimensions (ie.mobility and perception of risk) All these questions lead us to use different
methodologies of analysis. For example, concerning mobility, the daily traffic
can be simulated at a meso scale: a road axis for example. In that way, we aim
to simulate the global dynamics of the network from the modelling of flows on
arcs of the network (modulated according to the time of day and the day of
week). Yet, we use classical models (for instance equilibrium models) because
they give an ”average image” of the flows of vehicles on the arcs. Based on
this first structural mobility, it is then possible to consider ”a change of level”
regarding both the representation and the analysis: if a risk occurs or if a
specific context disrupts the structure. As a consequence, from a management
of flows on the arc, we turn to an analysis of the individual behaviours in a
multi-agent system