Cooperating search agents explore more than defecting search agents in the Internet Information Access

In the Internet Information Access Problem, information-seeking agents (software or humans) are selfishly rational in obtaining the information sought. From a single agent's perspective, sending out as many queries as possible maximizes the chance of achieving the information sought. However, if every agent does the same, the information servers will be overloaded and most of the search agents won't be able to retrieve the information. Our previous results suggest that when behaviorally similar information-seeking agents cluster together, cooperation (i.e., sending moderate number of queries) is promoted. In these experiments, the ranges of query (i.e., maximum logical distance from the information-seeking agents to potential information severs) is fixed for each search agent; agents only inquire the severs within the distance. In this paper, we attempt to evolve the range of the access distance. When similar agents - cooperators with cooperators and defectors with defectors - cluster together, cooperato rs tend to access diversified information sites while defectors tend to access only common information sites, resulting high congestion. This phenomena can be seen in human agents as well. When an agent sees too much competition or overuse of resource, it considers alternative choices. For example, when people see a congested highway, they tend to take other routes even if the routes may be longer. A similar phenomena is observed in our experiments. The results of the research can be used to help designing the Internet search agents that are efficient and less burdensome to information servers.