A very mild preference for being surrounded by people like onself will
transform a random population into a clustered/segregated one.
A 50% preference corresponds to a wish to have at least half one's neighbors
be like onself or, stated another way, a willingness to have 50% of one's
neighbors be different from one. Nonetheless, a clustered/segregated population
results. This is not the desired state for any individual but instead results
from the fact that some people in the randomly distributed population will
be unhappy because more than 50% of their neighbors are different from themselves.
It is the random movement of these people that triggers a cascade of movements
and this cascade of random movements that results in the segregated/clustered
The relation between similarity preference and degree of clustering/segregation
is not a straightforward linear one but rather one in which fairly large
changes in preference have little effect on clustering over some ranges
of values and fairly small changes in preference produce fairly large changes
in clustering over other ranges of values.
Such non-linear relationships characterize many phenomena resulting from
"cooperative behavior", ie behavior in which the behavior of each individual
element is influenced by the behavior of other elements. In the present case,
there is an important practical implication.
It is very hard to produce an unclustered/integrated population from a
random or a segregated population by altering the magnitude of the preference
to have people like onself in one's surroundings.
Segregated populations result from values through most of the range.
Integrated populations much more easily result of one changes the character
of the preference from a preference for similarity to a preference for difference.
Under these circumstances an integrated population occurs for a wide range
of preference values starting with either a random or a segregated population
An interesting question: Is a random unclustered population the same as
an integrated unclustered population (as emerges when the preference is
How might this be tested within the constraints of the model?
Good models raise as many questions as they answer. What additional new
ones does this one raise?
Some that can be explored within the constraints of this model?