Social Foraging in Groups of Search Agents
with Human Intervention
Abstract
Intelligent agents coordinate and cooperate flexibly when rules and dynamics of interaction can change over time and across different tasks and environmental conditions. Loose coupling emerges among agents when the rules of interaction are weak enough for agents to act independently or interdependently, and patterns of interaction vary as a function of conditions. Here, we examine collective foraging among simulated agents with and without human intervention. We find that loose coupling among search agents improved group foraging success, and that human players improved performance partly by subtle, indirect effects on group interactions. Analyses of movement patterns showed that loose coupling enabled collections of agents to self-organize and reorganize into a greater diversity of ad hoc groupings.