James Kennedy, PhD

I am a social psychologist, employed by the federal government in Washington, DC. My professional work involves research into survey measurement error, and development of survey data-collection software.

My independent research is mostly concerned with the paradigm known as the Particle Swarm algorithm. Populations of individuals interact in a simple fashion: they evaluate, compare, and imitate, eventually converging on optimal regions of the problem space.

The very complicated and technical theoretical assumption is this: when people have a problem, or when they're trying to understand something, they talk to other people about it. After they communicate awhile their beliefs become more similar, and they are probably better off than before they talked. You can write computer programs that do that, start from a population of random guesses and communicate until they've found a good answer.

I've been doing particle swarms since 1994, with the first papers on the subject by Russ Eberhart and me coming out in 1995. Researchers have tried lots of new things, sometimes making it simpler, sometimes more complicated, and the current state of the art hardly resembles the original. But the paradigm is young, and even though it is extremely simple I have to say there are very few people, still, who completely understand how the particle swarm works. Maybe nobody.

(I'm the big one)