Invited Talks

Deborah M. Gordon
Department of Biology, Stanford University


Interaction Networks as Distributed Algorithms in Ants

An ant colony operates without central control. Each ant uses only local information, mostly odor, and no ant can make global assessments about what needs to be done. Through the local decisions of individuals, colonies accomplish task allocation, adjusting the numbers of ants devoted to each task to the current situation. An ant decides where to go and what to do based on its recent experience of brief interactions with other ants. Most interactions consist of antennal contact, in which one ant smells the other. In an interaction, one ant perceives the chemical signature of another ant’s task, but does not give or receive any instructions. Interactions are not targeted toward particular individuals. The rate of interaction, rather than any information transferred, influences task decisions. Interaction networks explain how seed-eating harvester ants organize where they forage and how they regulate the intensity of foraging according to the availability of food. Seeds are scattered and colonies adjust foraging activity to food supply without pheromone trails and without reference to any specific location. Whether an ant leaves the nest on its next foraging trip depends on the rate at which ants return with food. Since most of a forager’s trip is spent searching for food, and a forager does not return until it finds food, the rate of forager return is a measure of how quickly food can be found. Task allocation in harvester ants depends on colony age and size; older, larger colonies show more stable dynamics than younger, smaller ones. We investigate how colony size affects interaction networks and thus colony behavior. We also investigate how interaction networks function in other ant species that collect food by continually modifying an established highway system rather than by individual searching.