Homework: Spring is in the air
By Alexander van der Vooren(Utrecht University) and Yuntao
Long (IQTE)
We constructed a simple agent-based model of
people in different parts of a country, who begin to want a change in
government. We divided people into five types of political states, pro-active,
pro-inactive, neutral, against-inactive and against-active, and they interact
with each other, and take different reactions to different rules. Different
initial distributions reveal different results. We consider the three initial
distributions, such as random, perfect separation, clustering, and we find that
there is a really big disparity in the time it takes before the against-active
people dominate the country. Given the initial distributions and the initial
percentage of various agents, the model produces several results.
Problem statement:
People in different parts of a country, begin to
want a change in government. The people who hold different political states have
different attitude to their government, ant they interact with each other.
Question:
How does the initial location of the population of
a country affect the speed of changing the government?
In our model, we divide the population of a
country into five types of people. People differs in political states, this is whether
they are Pro or Against the current government and
whether they are Active or Inactive supporters. Some people are Neutral. The
specific classification of agents is as follows:
1)
Pro-active
2)
Pro-inactive
3)
Neutral
4)
Against-inactive
5)
against-active
People move around their country, represented by
a grid in Netlogo. If people bump into each other
while moving, they might change their political state. So, some agents cause
other to change their political state.
This is explained by the behavioral rules.
We adequately consider the interactions between
the different types of agent in the model, people can be influenced by other
people they meet with different political positions:
Pro-active
- If Pro-active meets Against-active, and if Against-active
dominates local neighborhood, then become Pro-inactive.
Pro-inactive:
- If Pro-inactive meets Pro-active, then become also
Pro-active,
- If Pro-inactive meets Against-active, then
they become Neutral.
Neutral:
- If Neutral meets active agents, then they become
inactive agents from the same type.
Against-inactive:
- Synchronic to Pro-inactive
Against-active:
- Synchronic to Pro-active
We first test which kind of people will dominate
the country under different initial presence of the different political states.
People are randomly positioned on the grid. The Netlogo
applet below enables you to replicate the different scenarios.
Scenario 1. There are initially no active agents
present:
Nothing happens, all Inactive agents stay
inactive and all Neutral agents remain Neutral.
Scenario 2. Synchronal presence of political
states and most people are inactive supporters of either Pro or Against. For example, Pro-active is 5% of population, 2) Pro-inactive
is 40% of population, 3) Neutral is 10%, 4) Against-inactive is 40%, and 5) against-active
is 5%.
Figure 1 illustrates a simulation run of
scenario 1. Inactive agents become Active, and Neutrals disappear as well. Two
strong fronts
emerge of
which one will eventually dominate the country. Since the set-up is synchronal,
there is an equal probability for Pro and Against to
dominate the market. We see that different clusters of Pro and Against agents emerge. As one of the types, Pro or Against, becomes a minority, then the clusters start to
disappear and one will dominate the country.
Figure 1.
Scenario 3. Against-agents are initially
slightly more present than Pro-active agents. For example, Pro-active is 4% of
population, 2) Pro-inactive is 40% of population, 3) Neutral is 10%, 4)
Against-inactive is 40%, and 5) against-active is 6%.
Figure 2 illustrates a simulation run of
scenario 3. Eventually, either Pro-active or Against-active agents will
eventually dominate the market. Since, Against-agents are initially slightly
more present than
Pro-active
agents, there is a higher probability that Against-active agents will dominate
the market.
Figure 2.
Simple test
Our research question is how the initial
positioning of the different agents on the grid affects the speed of one type
becoming dominant. Therefore we tested scenario 3, above, in three new scenarios.
The first is similar to the above, the agents are
randomly placed on the grid (left hand in Figure 3). In the second scenario the
Pro and Against agents are perfectly separated from
each other, and Neutral agents are randomly placed on the grid (center in
Figure 3). In the third scenario there are four clusters. Two of which consist
of Pro agents and two consist of Against agents.
Figure 3. Left is
random, center is perfect separation, right is clustering
We ran for each scenario 100 simulations runs
and calculated for each scenario the average number of time steps needed before
one of the types becomes dominant (2/3 of the population). The simulation runs
show that for random positions it took on average 2837 time-steps, for perfect
separation 5229 time-steps, and for the clustering 2797 time-steps.
These results show that the number of
interactions with agents of other types is an important factor in our model.
When there is perfect separation, the change of meeting other people is
initially much lower than for random and clustering. The almost similar results
between random and clustering illustrate that also the clustering is an
important factor. If Against-active people are clustered it might be easier to
dominate the neighborhood such that Pro-Active people become inactive.
Given the limited time in graduate workshop
available for homework, there are many insights and ideas we are thinking
about. A striking question is what will the government do if the against-active
population attempts to dominate the country and how the government can keep the
country stable?
The model can be extended by testing different
government policy, such as jailing Against-active people for a couple of years,
which is actually the same as making them inactive and make it impossible for
them to interact with other people. Another possibility, for a government that
wants to control their system, is to prevent clustering of Against-active
people. When the against-active people cluster a lot, the government could take
some steps to scatter the same type of people. Finally, this model may also be
extended to the election issues, international dispute and the spread of disease
(going from on state to another, if bumping into other people) so on.
The applet requires Java 5 or higher. Java must be enabled in your browser settings. Mac users must have Mac OS X 10.4 or higher. Windows and Linux users may obtain the latest Java from Sun's Java site.