Multiagent Systems

I came across the article Expecting the Unexpected: The Need for a Networked Terrorism and Disaster Response Strategy by Stephenson and Bonabeau. They propose we should use a more distributed (swarm) approach to handling terrorist attack and natural disaster responses.

This essay argues that terrorist attacks or natural disasters are likely to be so unpredictable that they frequently require improvised responses (as conventional hierarchical structures are ill-suited to such situations) and outlines a flexible and highly adaptive networked structure. Networked personal communication devices and applications that the general public can and will use in a disaster offer the possibility of a new networked strategy that can foster the ?swarm intelligence? needed in a disaster, in which a community, even an ad hoc one, is capable of a higher level of collective behavior than could be predicted from the capabilities of individual members.

This is, of course, a brilliant idea. It is how our economy works, it is how most multiagent systems work. The interesting challenge we face is that in these systems the agents are humans (with all their irrational and idiosyncratic behaviors) and the communications technology is highly varied and unevenly distributed. For example, in a situation like after Katrina, I can imagine people using their cell-phones to broadcast their location, status, and special needs (medication, ambulance), others responding to these calls, others aggregating it into maps, lists, or service queues, others using these aggregations to mobilize larger-scale efforts like deploying the coast guard, etc. It is fun to think about the specific technologies and incentives that would be needed to make something like this work.

By the way, Bonabeau is CEO Icosystem which implements simulations of these type of swarm scenarios.