of 13
Current View
Global Manhunt Pushes the Limits of Social
Iyad Rahwan,
Sohan Dsouza,
Alex Rutherford,
Victor Naroditskiy,
James McInerney,
Matteo Venanzi,
Nicholas R. Jennings,
Manuel Cebrian
Masdar Institute of Science & Technology, UAE
University of Edinburgh, UK
University of Southampton, UK
NICTA, Melbourne, Australia
Massachusetts Institute of Technology, USA
The Tag Challenge, sponsored by the US State Department, required locating and
photographing 5 target persons in 5 different cities in the United States and Europe in
less than 12 hours, using only their mug shots. We report on how we used social media
to win this challenge, finding 3 of the targets. We reflect on the difficulties we faced,
the lessons we learned, and the implications of this capability.
1 Introduction
Back in 2009, to commemorate the
th anniversary of the creation of the Internet, the US
Defense Advanced Research Project Agency (DARPA) launched its Network Challenge
(also known as the “Red Balloon Challenge”). The challenge aimed to test the limits of
social mobilization and rapid information gathering using social media. It required com-
petitors to locate 10 weather balloons tethered at random locations all over the US. The
winning team, based at MIT, found all balloons within 9 hours by mobilization through so-
cial media [18]. The winning strategy relied on a novel incentive scheme in which people
were rewarded both for reporting balloon sightings, as well as for recruiting their friends to
look for balloons [17]. Further theoretical work proved that the strategy is in fact optimal
in terms of minimizing the investment to recover the information [4], and is also the most
robust to misinformation [16].
In March 2012, the “Tag Challenge,” funded by the US State Department, raised the bar
significantly higher (
). The challenge set the unprecedented
task of locating and photographing 5 people in different cities, across two continents, within
12 hours. There was a reward of $5000 for the winning team. Only a single mugshot of
each target person (or “suspect”) was released to the teams at 8am local time on the day of
the competition (Figure 1). Each volunteer target wore a t-shirt with the competition logo
(the appearance of which was also not known until the first mugshot was released) and
was instructed to follow a 12-hour itinerary designed to reflect a normal day. For example,
the New York City suspect started at Columbia University, had breakfast at a cafe nearby,
took the subway to the World Trade Center site, then went shopping, and so on. As such,
the task represented a realistic search for an individual following a characteristic mobility
pattern. The other suspects were in Washington D.C., London, Stockholm, and Bratislava.
Figure 1: Only information given was a single mug shot per city released at 8:00am local
Now, the Tag Challenge is significantly harder than the Red Balloon Challenge because
it required locating people in extremely populated cities, where the pace of life can help
people “hide in plain sight” [15]. Therefore, the target people were much harder to spot
than large red balloons. Furthermore, people are mobile, making it difficult to rule out
locations that have already been visited. While the suspects were not explicitly hiding or
in disguise, searching for one moving about in a city like New York, with a population of
8.2 million people, seemed nearly impossible. In particular, it was an open question as to
whether it is possible to use social media to accomplish such a difficult task, as evidenced
by the difficulty of finding suspects in police investigations [19].
Despite these challenges, our team won the challenge by locating 3 of the targets using
a Web platform, a mobile application, and an incentive scheme. This was accomplished
without any of the team members being located in any of the target cities. Specifically,
Figure 2 shows the targets our team found, and the approximate local time at which we
submitted their photos to the organizers. The targets in London and Stockholm remained
at large, although pursuing them after the allotted 12 hours was not part of the competition.
Other teams used Web sites and social media in combination with a wide range of ap-
proaches, including attention raising measures (e.g., search engine optimization + broad-
casting a high volume of messages on social media), standard incentive measures (e.g.,
promising shares of revenue to participants + pledging winnings to charity), and decep-
tive measures (e.g., impersonating the challenge organizers + attempting to sabotage rivals
teams with a virus). Clearly, there was a wide range of possible strategies in this competi-
tion; our approach was based on an understanding of the key challenges of rapid mobilisa-
tion of crowdsourcing teams. In the next section, we give an overview of these challenges.
2 Challenges
There are at least three distinct problems in crowdsourcing rapid information gathering: (1)
of participants; (2) the
of information; (3) and the
of information. We discuss each one in turn.
2.1 Mobilization
The success of search in social mobilization clearly requires individuals to be motivated to
conduct the search, and to participate in the information diffusion. In an attempt to repli-
cate Stanley Milgram’s “small world experiment” [14], it was observed that the majority
of message forwarding chains observed empirically terminate prematurely. Specifically,
Dodds et al. conclude that “
although global social networks are, in principle, searchable,
actual success depends sensitively on individual incentives
” [5]. In other words, a key chal-
lenge in social mobilization is the
incentive challenge
. It has also been observed that the
success of crowdsourcing mechanisms, in general, can vary depending on the details of the
financial incentive scheme in place [12].