A hierarchical search strategy with economic incentives

DARPA ran a contest recently. They placed ten red weather balloons at secret locations all over the United States. The first person to correctly identify the locations of all 10 balloons would win the prize.

The winners (a bunch of postdocs from MIT's Media Lab) organized it as a pyramid scheme (the concept is useful beyond Madoff and Amway!). The person who reported a balloon to a central website would get $2000; the person who invited him to participate would get $1000, and the person who invited that person would, in turn, get $500, and so on. So, there was an incentive, not just to find the balloons, but to canvass your social network for people who were spread out, and into this sort of thing.

What I like about this is that it takes advantage of two concepts: hierarchical, parallelized search (each person looking in their neighborhood independently, sort of like a genetic algorithm) and a smart economic design of incentives (to increase the number of agents conducting the search).

Ultimately, it took just 9 hours to locate all ten balloons.

No comments:

Post a Comment