Modeling self-organization of communication and topology in social networks


Illustration of how to run the applet (click to enlarge).
If you don't see a network like this above, please install Java.

A black line between two nodes represents a link and make communication between the connected agents possible. The network can be modified by addition or removal of links (see figure). Click on a node, and while the button still is pressed down, drag the pointer to another node and release.

The ratio between communication and strategic rewirings in the network is set by the scrollbar under "Communication level".

The network can be shown in two different modes, and each mode in two versions where the number of links or the communication distance define the position relative to the center. The button under the "Information perspective" toggles between the different modes. When "Strategy" is selected, the agents can be chosen to chat, cheat or lie. Click on the corresponding button and then on a node.

When  "Information",
the size and the color of the nodes reflect the age of the information All agents have about the Selected node, the Selected node has of All nodes, or All nodes have on average about All other nodes.

To show the age and the amount of information that flows over the links, the links have a background color and a width. The width of these links reflects the relative amount of information that they transfer from the Selected agent to All other agents, to the Selected agent from All other agents, or between All agents. The color of the links represents the newest of this information (if
"All about all" is selected it represents the average).

The communication can be limited by links or nodes. Nodes communicate relative to the number of links they have in the first case and with equal amount as all other nodes in the second case. The agents' interests can be controlled by who they communicate about. By selecting "Age" their interests decay inversely proportional to how old the information about an agent is and with "No age" selected they talk about anybody.

If you have any questions or comments, do not hesitate to contact

Illustration of the communication event in the model (click to enlarge). Agents A, B, C... communicate with friends and get a global perception. This perception is illustrated by the memory bubble of A, before and after the chat with B. It consists of the agents (first row), the age of the information about the agent (second row), and from whom the information came (third row).

Illustration of the rewiring event in the model (click to enlarge). A asks B, the acquaintance that provided A with the newest information about H, from whom the information came from. B answers E and A open for communication with E by adding a link to E.

In the model applet here, we show that it is possible to build a reliable perception of the whole through repeated small talks. We simply let agents memorize the acquaintances that provided the newest information about other agents together with the age of this information.

Here we take that model one step further and give the agents a social mobility.  The agents can thereby get new acquaintances to meet different interests.

Agent A in the second illustration above would like to be better positioned relative to agent H. A therefore asks B, the acquaintance that provided A with the newest information about H, from whom B got the information. B answers E and A establishes a contact with E. At the same time a random agent looses a random link to keep the number of links constant. In this way the social mobility of the agents is limited to steps to an acquaintance's acquaintance.

The perception of the agents and the communication backbone differ substantially at low communication rates. At higher communication rates the rewiring become meaningful and the network topology no longer random.

The different strategies to become central have different effects on the network topology. For example, the more an agent chats with its surroundings, the better it and the surroundings perform. Contrary, cheaters and liars can at maximum gain themselves.

Valid HTML 4.01!