From Santa Fe Institute Events Wiki
|CSSS 2006 Santa Fe|
Network Growth Simulation
Objective : simulate a range of social behavior
Agents are nodes linked by edges.
Resource Allocation. Each agent has an attributed flow of R resources: Ai, Bi, Ci, etc. Base Case: R=2 Variant:: R=3, 4, 5, … The sum of resources is constant per agent : Ai+Bi+Ci+…=C
Dynamics. Permanently, an agent search potential alter to link to. Base case: search randomly on all other agents
Agency. For each potential target, evaluate the opportunity. Each agent brings it resource set. The combined set resource constitute the dyad resources. The utility is evaluated on the pair. Each agent compares his utility from the joint, to the utility in its previous case. If joint utility is greater for both, then link. If joint utility is lower, then drop link. Base Case: utility is split half Variants: agent could propose an unbalance split The utility function is a classic production function. Base Case: u(A, B, C…)=sqrt(A)sqrt(B)sqrt(C)… Variants: • u could be a Cobb-Douglass with other coefficient (not very interesting) • u could vary slightly by agent (not very interesting)
Resource Sharing. Every so often, resources are distributed among agents. Base Case: after every change in the network Variant: space the update (algorithm complexity!) Every agent takes half of its resources, and splits it equally among its alter. Base Case: the split value is spread 2 Levels Variant: L=3, 4, infinite (beware complexity) Each agent has therefore to remember the resources coming from alters. This will be taken into account for all subsequent utility evaluation. When exhibiting agency, the total utility available should take into account the “leak”!!!
Cross-level analysis: a total utility can be computed for the group (connected agents) by applying the utility function to the sum of all resources available. Base Case: same utility fct Variant: different utility fct
- 09 06 06 Summary of our ideas so far (please correct them if necessary)
- Explore system where network grow due to individuals benefit for joining , and where both individual and group level of performance can be differentially studied
- Possible research objective: study social dilemma of network growth
- Initial model: Douglas-Cobb function for production function. Allocation of welfare is equalitarian, to be later differentiated.
- Next Step: all members learn basics of netlogo (NL).
- Next meeting: meet for lunch saterday