Policy and Complex Systems
the most influential neighbors are those lying at the two extremes of the awareness scale : evangelist and active agents at the top and blind agents at the bottom ( Fig . 2 ).
The awareness of the agents is updated at each process iteration ( Fig . 3 ). A basic design principle of this mechanism is context-dependence : the influence of neighbors depends on their type . In addition , the rate of awareness change and thresholds are different for each type . Another important design principle is saturation : the “ greener ” the agents , the higher the threshold they have to reach for moving to a new type . A third principle is hysteresis : once agents become “ green ” ( i . e . join the active or evangelist types ), their awareness never decreases . Finally , transitivity of influence supports a kind of cascade effect , limited by the influence sphere of the agents .
Each agent has a reduction goal concerning the limited resource and progresses toward its goal at a given rate . At each run , the number of agents belonging to a type can change , while total number of agents is constant . Main global variables include the cardinality of each type of agents , the current resource consumption , and an overall reduction goal .
The model evolution stops when the global consumption reduction goal is achieved . With each iteration , agents look around to verify how many neighbors they have and what type . According to its neighborhood , each agent then changes its awareness level . The rules to update awareness are different for each agent type . This models general communitybased social pressure . Blind agents can change their awareness only if they have completely green neighborhoods and even then , their awareness increases very slowly .
55