| Defect | Cooperate | |
|---|---|---|
| Defect | ||
| Cooperate |
Background: This is a spatial version of the Prisoner's Dilemma. It is similar to the first model in that payoffs may vary, and pairs of individuals only interact a single time. However, in this model instead of individuals being able to interact with anybody in the entire environment, they can only interact with those individuals that are in a local “neighborhood.” This can result in a different outcome than with the same parameters in the first model.
Instructions: Provide parameters for the four possible meetings between cooperators and defectors, as in the first model. All parameters must be positive or zero, but there are no other restrictions. When the range is left at the default of 1, everyone interacts with only those individuals that are within one cell's distance. This may be increased to 2 or 3 cells in each direction (range 2 and 3). The default edge condition of “wrap” means that cells on the edges are effectively each other's neighbors. You can start with random initial conditions by clicking Load, or select one of the prepared starting conditions from the Initial menu. You may also modify the grid by clicking on squares; they will cycle through available colors. Each time you change parameters, re-initialize the grid by clicking Load. Make sure to click Stop before scrolling down or changing parameters.
Buttons: Start — begins the simulation, Step — advances the simulation one time step, Stop — stops the simulation, Load — loads the selected initial configuration.
Interpretation: Due to the relatively small grid size, there is a lot of random fluctuation in this model. When starting from a random grid you may want to re-randomize and run it again. On the grid, blue represents cooperators, red represents defectors, yellow represents defectors that were just cooperators, and green represents cooperators that were just defectors. Are outcomes necessarily the same (or different) than the same parameters in model one? Are there clusters of like types on the grid or is everything evenly distributed? To simplify things, try fixing three of the parameters and just varying one.
Based on a model by Nowak, May and Sigmund — Scientific American June 1995.
CA models based on terra.js