MIT floats ideas in DARPA balloon challenge (Q&A)
The head of the team that conquered the geolocation contest reveals the winning techniques and tells how the Internet can be harnessed to tackle real-world problems.
MIT's Riley Crane only found out abouta few days before it started. Yet his team went on to through its savvy use of the Internet.
The challenge posed by the Defense Advanced Research Projects Agency asked people to find the coordinates of 10 red weather balloons floating above the U.S. in one day. Since no one individual could plot the location of all 10, participants had to figure out how to work with others to solve the puzzle.
Team MIT's strategy was to build a Web site designed to attract more and more followers--people who might know the balloons' locations themselves and those could bring aboard others who knew the coordinates, essentially creating a chain effect.
The five-member MIT Red Balloon Challenge Team consisted of group leader Crane and Manuel Cebrian, both post-doctoral research fellows at the Massachusetts Institute of Technology. Rounding off the team were three students--Galen Pickard, Wei Pan, and Anmol Madan.
Crane holds a Ph.D in physics, while Cebrian has his in computer science. Both Pickard and Madan are Ph.D candidates, and Pan is preparing for his MS degree. As researchers in the Human Dynamics Group at MIT's Media Lab, the five study the science of how people interact with technology.
On Monday I spoke with Crane to discuss his thoughts on entering and winning DARPA's offbeat challenge.
First, congratulations on winning the challenge. Second, can you tell me how you entered the contest and what methods or strategies you used to find the positions of the balloons?
Crane: Sure, I think the key factor as to why our team ended up able to pull this off--I mean, we had some great competitors that we only learned about after the fact--is that we really designed a system that allowed people to [see it] as sort of a recursive incentive. So on the surface, I think some people see it as sort of a pyramid scheme, when in fact, it's exactly the opposite. I can illustrate it with an example of a friend of mine in Europe who wrote back and said "There's no reason I should join because I'm in Europe. I'm not going to find one of the balloons." And that's exactly the point. We designed this sort of recursive incentive. Of course, he's not going to find the balloons. But he might know someone in the U.S., and that person may find a balloon. And so if he can put us in contact with that person, then he should get some type of a reward as well.
So the system we built is as simple as that. If you heard about our Web site and went to sign up directly, and you found a balloon, you would get $2,000, and $2,000 would go to charity. If instead you signed up and then you told your friends, and one of your friends found a balloon, that person would still get $2,000 because they found the balloon. And you, because you signed someone up who found the balloon, would also be rewarded with $1,000, and then $1,000 would go to charity.
We think this recursive incentive allowed people to feel as though they were a part of a team helping to solve this proverbial needle in the haystack challenge. And they could participate in a meaningful way, even if they really had no chance of finding a balloon themselves. We had a lot of feedback of "We won." And people really felt they were a part of the team because of this.
One of the interesting things is that in trying to understand what we actually did, a lot of people might think of viral marketing. But again, this is the wrong point because it's not that our approach was to get a message out. It was more that we wanted people to send information back to us. You could really see a fun way of engaging people that they can see how influential and how resourceful they could be at getting people to join this challenge and recruiting them for the greater good.
So it was kind of a snowball effect?
Crane: Absolutely. The key is to try to make this as self-propagating as possible so that we can really get as many people engaged.
Riley, do you have a sense of how many people were engaged through your site?
Crane: The shocking thing on our side is that I actually only found out about the challenge four days ago. A friend of mine sent me a link and said, "Oh, I guess you must have a team over there at MIT." And we were kind of joking around within our group and very quickly hit upon this idea. In two days, I built a site and got it running. And then we only officially launched the site on Thursday evening. We sent out five e-mails because there are five members of the team. And within 48 hours, we had gone up to 5,000 [e-mails]. And we had something on the order of a few hundred thousand page views of people who may not have signed up but came and took a look. So it was an incredible bootstrapping approach that really got us up to speed.
Do you have a sense of what tools or sites other people used to spread the word? Did they use traditional social-networking sites like Facebook or Twitter or other resources?
Crane: I'm not entirely sure, but I do know of a few other competitors, one of which used an iPhone application. I think there was another one, a videologger, who had a few hundred thousand followers and a pool of thousands of people that they could call and ask them to verify [the locations of the balloons].
How were you were able to verify the responses from people who sent you the coordinates of the balloons?
Crane: That's the only thing we can't discuss at this point. I can tell you that we received a lot of spam and had indirect, private communications with one of the other teams who was happy they had been so successful at spamming us. We received a lot of Photoshopped images of balloons in different places in the U.S. It was a very exciting Saturday for us in our lab up here in the Human Dynamics group. But as for the actual way we did it, we'll discuss in a forthcoming release.
As far as the prize money, it sounded like whatever amount goes to a single individual, that same amount would also go to charity?
Crane: Not the same amount. Each balloon had a value of $4,000. If you came directly to us without a referral, you got $2,000, and the charity got $2,000. If you came with one referral, if one person referred you to us, then you still got your $2,000, the referrer gets $1,000 and the charity gets $1,000. This goes on from $2,000, $1,000, $500, $250, $125, and so on. This goes on down the chain. It essentially gets to the point where you would be splitting pennies if the chain were extremely long.
So if you took the $40,000 cash money, can you extrapolate how much would go to participants and how much would go to charity?
Crane: We would be able to, but at this point we haven't looked at the data. We're trying to be as transparent as possible. So we've actually asked the MIT auditing department within the greater Massachusetts Institute of Technology to verify our results and make it official so that there's complete transparency in how the process is handled.
Do you know yet which charity would be the recipient of any of the funds?
Crane: No. That will be announced at some point, how that would be done. But at this point, we have no comment on that.
And if I could add, from our point of view, what the message of this was. I think it's important to point out that there's a tremendous scientific opportunity in all of this, and from our side, we were never in it for the 10 balloons. Of course, that was the challenge, and that was exciting. But from a broader scientific perspective, we were in it to understand how to mobilize the vast resources of the human network, to face challenges and explore opportunities in living in such a connected society. And as a footnote to that, I think some of the applications that might come out of this would be: Can we use this technology we've developed to find missing children or something along those lines where there's an incentive for people to really participate and help out? Often, the police will offer a reward for finding a missing child. Can we restructure that in a way that we tap the vast resources of this network? Again, maybe you don't live in the state where a child was abducted, but maybe you know someone who does. Or during an emergency, maybe we need to find 10 people in a region who can operate heavy machinery, maybe a building collapsed. And how can we use these new tools to solve those challenges to help society? That's kind of the broader message that comes out of this from our side.
I know DARPA seemed to go into the challenge with the same point of view. Their goal was not to have the positions of the balloons uncovered. Their ultimate goal was to gauge how the Internet and social networking could be used and harnessed for more widespread issues and problems and puzzles.
Crane: Yeah, I think there's a subtlety in there. I think it actually only works because there's a benevolent or greater good. You know, somebody had asked us if they thought we could use this to do something bad. And I think it really wouldn't work. The incentives were designed specifically so that people feel good about the fact that they're participating, that maybe if they don't solve it, that somehow they're helping charity or helping science in the greater good.