Multi-agent decision making: lessons from the ants

David Sumpter

Department of Zoology, Oxford University, South Parks Road, Oxford OX13PS, UK

I discuss the problem of how a group of agents with different information but a common goal can make a collective decision about how to reach that goal. The problem is one of integrating the agents' independent information to arrive at a unanimous decision without losing the integrity of individual judgment. A balance must be struck between 'independence' and 'positive feedback', with independence improving the accuracy of the decision and positive feedback improving the speed with which it is taken. To illustrate these concepts with an example, I will look at how colonies of Leptothorax ants choose between alternative nest sites. These ants can choose the best of several nest sites, even when many of the active ants who organize the move visit only one site. By building an agent-based model I describe the decision rules that allow such fast and accurate decision-making. These rules are a combination of independent assessment and recruitment feedback with a quorum threshold determining when a collective decision is taken. I discuss how this last aspect, the quorum threshold, might be the mechanism that balances independence and positive feedback in a proportionate manner.