reciprocal exchange and resilience of social networks

Some quick thoughts on:
Reynolds et al. (2003) The Effects of Generalized Reciprocal Exchange on the Resilience of Social Networks: An Example from the Prehispanic Mesa Verde Region, Computational & Mathematical Organization Theory

Summary

The authors of this paper claim that their experiment shows the reciprocal exchange of resource led to larger populations and more resilient social network. In their state-based model, agents in different states can exchange resource across the kinship network. Resource can be donated, for example, from agents in satisfied state to agents in critical state, based on three cooperation methods: requester-initiated, donor-initiated, or both. Their social network links are established or updated based on rules such as children’s marriage and moving out, sibling, remarriage, relative relationship, friendships and neighbors, etc. The model has two spatial constraints: the distance constraint of immigration and distance constraint of searching an available donor.

Queries

Their claim about social network resilience is base on the number of social links, volumes (nodal degree) of social networks and agent population, which is not convincing to me. The causal link between the reciprocal exchange of resource andthe change of social network links is not clear. On the other hand, the change of social network links is more obviously affected by the change of relationship. For example, when an agent is in critical state, he may move away and then build new relationship with other agents. The change of relationship can be affected by environmental change. When the environmental change leads to death of agents, agents could remarry and thus build new relationship. Such of changes of relationship could lead to increase of social network links (even an agent’s spouse die, the relatives are still alive), but this doesn’t mean that the social network is more resilient.

The concept of the cultural algorithm seems to be a elegant framework to facilitate agents’ learning ability, but the authors didn’t specify how they exploit this concept in their model.