created with NetLogo
view/download model file: Learning2.nlogo
WHAT IS IT?
-----------
Can agents learn about their environment by observing the actions and success of their neighbors? The agent in this model use the majority/plurality decision rule to determine which of two actions yields the better payoff. The two regions (black and white) can have different probabilities of rewarding actions A and B. The models tests whether agent behavior can communicate information about the agents' environment.
HOW IT WORKS
------------
Agents initially choose between action A and B at random. They then observe the actions and outcomes of their neighbors and themselves. In the next time step, agents choose the action that was more successful in the previous step. If plurality is on, then they do this by simply counting the number of winners who chose A and the number of winners who chose B and selecting the action with the higher number of winners. If plurality is off, then to switch to a new action, the strict majority of an agents neigbors would have to have won with a particular action.
A mutation rate can be set from 0% to 1% and flips the agents' choices at that rate. The mutation rate and the probablity of rewarding A or B can be altering while the model is running.
HOW TO USE IT
-------------
The screen starts all white. Decide on a neighborhood radius (high values slow the model down considerably) and whether to use the majority rule or the plurality rule. Click Setup to initialize to world and make the first step. The blue agents chose action A and the red agents chose action B. Winning agents appear in a lighter color.
Click Go or Step to see the model run some more. You can adjust the mutation rate and the probablity of rewarding action B while the model is running.
THINGS TO NOTICE
----------------
Notice the difference of behavior of agents under majority and plurality rules. Look closely. How frequently do majority rule-using agents change between red and blue? The border in the middle is saw-toothed, but the wrap-around border is smooth; does this difference have an effect on communication?
THINGS TO TRY
-------------
Let the agents turn all blue by cranking the percentage all the way up. Then set the mutation rate all the way to one percent. Now, how low do you think you have to set the percentage before you see a crimson tide? Now try it at half the mutation rate.
CREDITS AND REFERENCES
----------------------
The model description was provided by Scott Page and implemented by Aaron Bramson. Both individuals are located at the Center for the Study of Complex Systems at the University of Michigan. http://www.cscs.umich.edu/