Geoscience Reference
In-Depth Information
entities, whose definition can be varyingly complex. Such an approach provides an
alternative to a recurring position of methodological individualism when using
multi-agent systems (MAS) in social sciences. Indeed, most of these applications
mobilize in a privileged way the simplest elementary entities, notably individual
beings or households, whether in human geography or in archeology. For Benenson
and Torrens, such a choice allows escaping the problem of the modifiable areal unit
problem (MAUP), which makes the results (of statistical model or gravity model, for
example) dependent on the definition of the entities on which the treatment operates
[BEN 04]. Formalizing the rules of evolution at the agents' level corresponds to an
epistemological positioning of methodological individualism. The interactions
between simple entities (at the most basic level; here the people) may however not
correspond to the same principles as the interactions between the composite entities
that they are part of (whether they are social or spatial aggregated entities). From a
methodological point of view, MAS corresponds to an individual-centered approach
but this does not mean that the epistemological framework is that of the
methodological individualism. An individual-centered model formalizes the
behavior of elementary entities rather than that of the aggregation of these
elementary entities, whether these elementary entities are: individuals, houses, firms,
but also parcels, settlements or cities. Methodological individualism, for its part,
systematically refers to individual beings in the sense of economic or social agents
as the elementary entities of an analysis. The MayaSim and EuroSim models
presented in Chapter 4 are hence methodologically individual-centered models, but
the elementary entities are composite (city-agents and group-agents) and the
approach does not relate then to methodological individualism.
A number of discussions on this issue are conducted in economics and sociology.
Lars Udehn [UDE 02] therefore proposed to explicitly distinguish the approaches of
methodological individualism depending on whether they are targeting the definition
of social concepts, the explanation of social phenomena or the reduction of
social laws into individual laws (physical or psychological). These epistemological
nuances would allow going further in the construction of the explanation of spatial
phenomena, or at least giving a complementary viewpoint to that given by most
microsimulation models or MAS models developed at the individual level. The
elements of Chapter 1 allow us to conduct a reflection in this direction to formalize
in a pertinent manner the links between different geographical levels: the examples
discussed in Chapter 4 illustrate the interest of using MAS to simulate the effects at
a higher level of organization, of interactions operating at a lower level of
organization. Two types of examples have been mentioned:
- the effects of interactions between agents representing human individuals on
the spatial configuration of a territory (typical example of Schelling's socio-spatial
segregation model);
