Geoscience Reference
In-Depth Information
settlement) are adapted to measure and explore phenomena of diffusion or durability.
Conventional statistical methods (notably, multivariate data analysis and regression
models) allow us to analyze such variables, to highlight how the differences between
the studied objects (as “statistical individuals”) are structured and to identify trends.
Such approaches can be called “data driven” to the extent that information is generated
from the manipulation of empirical data. It corresponds to the situation where the
problematics of research could be formalized in terms of statistical individuals and
variables.
The researcher may have other perspectives in regards to the empirical reality.
Let us consider the example of Schelling's model [SCH 71] which is emblematic in
the field of simulation in the social sciences. The empirical information consists of
remarking that there is ethnospatial segregation in most American cities. This
information is of a qualitative nature, because we simply observe that some
neighborhoods have a higher concentration of “black” people and others of “white”
people, and that this phenomenon is recurring from one city to another. Schelling's
aim is to use the model to reflect on the causes of these concentrations. The space is
represented by a grid on which a set of agents from two categories, that we will
name here A and B, is laid down. At each time step, each agent chooses to remain at
the same location or change cell. The rule is the following: if the proportion of
agents of the same type as itself in the neighborhood formed by the eight
neighboring cells (Moore's neighborhood) is lower than a given threshold
s
(Figure 4.1), then the agent “moves out” and settles down in an empty cell. The
point of this model is to show that it is sufficient for
s
= 1/3 to cause the simulation
to show a segregated spatial organization. Daudé and Langlois [DAU 07] have
explored in a systematic manner the effects of different thresholds of tolerance
according to the spatial density of the agents on the levels of aggregation that
emerge (figure 4.2). The conclusion is that there may be spatial segregation of the
population without a particular planning policy, and without individuals having
segregationist behaviors (since they accept to be a minority in their neighborhood, as
long as this minority is not less than one third). Even if the individuals have
preferences for a mixed neighborhood, the outcome is this segregated organization.
Such a model is said to be “concept driven”: the researcher has a hypothesis on the
role of a certain mechanism that she/he implements in the form of an interaction rule
between the agents and their environment. This rule operates for each agent, at each
time step, and the modeler observes the simulated situation (that is the spatial
organization of agents in the grid according to their A or B character) at the end of
n
iterations. Such computer simulation models are thus inherently dynamic, the driving
force of change residing in the interaction rule leading to the move of the agent, or not.
Thus, it refers explicitly to the
behavior
behind the change, which offers a different
perspective than the one that simply consists of giving an account of the change.
