I can’t resist the urge to brag a bit about a cool victory for MIT, my alma mater and professional home. A team from the Institute’s Media Lab won the DARPA Network Challenge, an experiment in distributed intelligence conducted in part to celebrate the 40th anniversary of the 4-node ARPANET network that blossomed into the Internet.
DARPA announced that on December 5 it would float ten large red balloons in fixed locations throughout the United States. The first team that submitted to the Agency the correct longitudes and latitudes for all ten balloons would win the $40,000 contest, which would be active for seven days.
It took the MIT team only nine hours to find all ten balloons.
Like most other groups that entered the contest, the MIT Red Balloon Challenge Team recruited a country-wide network of spotters; the goal, obviously, was to make this network as large and geographically distributed as possible. It also needed to be robust against spammers, saboteurs, and others who would submit false spottings in order to confuse a team.
So how do you quickly build a large, dispersed, sufficiently motivated, and fault-tolerant network of balloon finders? The MIT team’s approach was brilliant. It relied on the Web and its social networking utilities like Facebook and Twitter (duh), on multilevel and relatively high-powered financial incentives, and on a clever tracking mechanism.
Here are quotes from the team’s site explaining how it worked:
When you sign up to join the MIT Red Balloon Challenge Team, you’ll be provided with a personalized invitation link, like http://balloon.mit.edu/yournamehere.
Have all your friends sign up using your personalized invitation. If anyone you invite, or anyone they invite, or anyone they invite (…and so on) win money, then so will you!
We’re giving $2000 per balloon to the first person to send us the correct coordinates, but that’s not all — we’re also giving $1000 to the person who invited them. Then we’re giving $500 whoever invited the inviter, and $250 to whoever invited them, and so on… (see how it works).
It might play out like this. Alice joins the team, and we give her an invite link like http://balloon.mit.edu/alice. Alice then e-mails her link to Bob, who uses it to join the team as well. We make a http://balloon.mit.edu/bob link for Bob, who posts it to Facebook. His friend Carol sees it, signs up, then twitters about http://balloon.mit.edu/carol. Dave uses Carol’s link to join… then spots one of the DARPA balloons! Dave is the first person to report the balloon’s location to us, and the MIT Red Balloon Challenge Team is the first to find all 10. Once that happens, we send Dave $2000 for finding the balloon. Carol gets $1000 for inviting Dave, Bob gets $500 for inviting Carol, and Alice gets $250 for inviting Bob. The remaining $250 is donated to charity.
See how clever this is? People have incentive to sign up even though they know that chances they themselves will see a balloon are low. If someone they recruit sees a balloon, or even someone two or three levels of recruitment away does, they still make money, so why not take the few seconds to sign up? And thanks to the personalized invitation link URLs, it’s easy for the MIT team to accurately assess who signed up who, so the reward money can be divvied up automatically and unambiguously. I never thought the techniques of multilevel marketing would be good for anything except selling makeup; guess I still have a lot to learn.
I imagine that the personalized URLs also serve at least one other important purpose: they help assess whether a particular balloon sighting is real or bogus. Let’s say I’m a saboteur who wants to make the MIT team fail. So I recruit 50 people to send in the same bogus sighting. The MITers will see that all 50 sightings came from me, and so will probably trust those data points less than 50 spottings of the same balloon that have no obvious connection to each other. And the combination of personalized URLs and IP addresses provides, I’m thinking, a pretty good way to see how connected different groups of spotters are.
As is so often the case “God is in the details” when addressing a challenge like DARPA’s, and it feels like the MIT Red Balloon Challenge team did a great job of figuring out which details mattered, then building a solution that paid proper attention to them.
What do you think of the challenge and the MIT team’s victory? Am I missing anything important about it? Where else have you seen similar approaches? Were you expecting someone else to win, based on their smart approach? And what lessons do you take away from this about harnessing crowd energy? Leave a comment, please, and let us know.