DTIC ADA581803: Harnessing the Crowdsourcing Power of Social Media for Disaster Relief

1.7 MB·English

by  Defense Technical Information Center

C Y B E R - P H Y S I C A L - S O C I A L S Y S T E M S
Editor: Daniel Zeng, University of Arizona, [email protected]
Harnessing the
Crowdsourcing Power
of Social Media for
Disaster Relief
Huiji Gao and Geoffrey Barbier, Arizona State University
Rebecca Goolsby, US Office of Naval Research
Social media has recently played a criti- challenges that must be addressed to make crowd-
sourcing a useful tool that can effectively facilitate
cal role in natural disasters as an informa-
the relief progress in coordination, accuracy, and
tion propagator that can be leveraged for disaster
security.
relief. After the catastrophic Haiti earthquake on
12 January 2010, people published numerous Disaster Relief Systems
texts and photos about their personal experiences and Crowdsourcing
during the earthquake via social media sites such Coordination is a central, challenging issue in
as Twitter, Flickr, Facebook, and blogs, and vid- agent-oriented distributed systems.3 Disaster relief
eos were posted on YouTube. In just 48 hours, systems focus primarily on designing coordination
the Red Cross received US$8 million in donations protocols and mechanisms to manage government
directly from texts,1 which exemplifi es one benefi t and non-governmental organization (NGO) activi-
of the powerful propagation capability of social ties. Research has shown that it is possible to lever-
media sites. age social media to generate community crisis maps
Survivors also use social media sites to keep in and introduce an interagency map (see Figure 1) to
touch with the world after a disaster. The jammed allow organizations to share information as well
mobile phone network in Japan caused by the re- as collaborate, plan, and execute shared missions.4
cent tsunami and earthquake made it diffi cult for The interagency map is designed to allow
people to communicate with each other. In re- organizations to share information if they ope-
sponse, they used Twitter, Facebook, Skype, and rate on the same platform or use similar data-
local Japanese social networks to communicate representation formats.
and keep in touch with their loved ones.2 Recently, the effi ciency at which government
Although social media can positively impact and NGOs are able to respond to a crisis and
disaster relief efforts, it does not provide an in- provide relief to victims has gained increased
herent coordination capability for easily co- attention. Researchers are exploring and devel-
ordinating and sharing information, resources, oping applications to leverage crowdsourcing for
and plans among disparate relief organizations. disaster relief.5 Crowdsourcing allows capable
Nevertheless, crowdsourcing applications based on crowds to participate in various tasks, from simply
social media applications such as Twitter and “validating” a piece of information or photo-
Ushahidi offer a powerful capability for collect- graph as worthwhile to complicated editing and
ing information from disaster scenes and visual- management, such as those found in virtual com-
izing data for relief decision making. This article munities that provide information—from Wiki-
briefl y describes the advantages and disadvan- pedia to Digg. This is a form of collective
tages of crowdsourcing applications applied to di- wisdom6 information sharing that strongly lever-
saster relief coordination. It also discusses several ages participatory social media services and tools.
10 1541-1672/11/$26.00 © 2011 IEEE Ieee InTeLLIGenT SySTemS
Published by the IEEE Computer Society
IS-26-03-Cpss.indd 10 24/05/11 1:03 PM
Report Documentation Page Form Approved
OMB No. 0704-0188
Public reporting burden for the collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources, gathering and
maintaining the data needed, and completing and reviewing the collection of information. Send comments regarding this burden estimate or any other aspect of this collection of information,
including suggestions for reducing this burden, to Washington Headquarters Services, Directorate for Information Operations and Reports, 1215 Jefferson Davis Highway, Suite 1204, Arlington
VA 22202-4302. Respondents should be aware that notwithstanding any other provision of law, no person shall be subject to a penalty for failing to comply with a collection of information if it

does not display a currently valid OMB control number.
1. REPORT DATE 3. DATES COVERED
2011 2. REPORT TYPE 00-00-2011 to 00-00-2011
4. TITLE AND SUBTITLE 5a. CONTRACT NUMBER
Harnessing the Crowdsourcing Power of Social Media for Disaster Relief
5b. GRANT NUMBER
5c. PROGRAM ELEMENT NUMBER
6. AUTHOR(S) 5d. PROJECT NUMBER
5e. TASK NUMBER
5f. WORK UNIT NUMBER
7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES) 8. PERFORMING ORGANIZATION
Arizona State University,Tempe,AZ,85281 REPORT NUMBER
9. SPONSORING/MONITORING AGENCY NAME(S) AND ADDRESS(ES) 10. SPONSOR/MONITOR’S ACRONYM(S)
11. SPONSOR/MONITOR’S REPORT
NUMBER(S)
12. DISTRIBUTION/AVAILABILITY STATEMENT
Approved for public release; distribution unlimited
13. SUPPLEMENTARY NOTES
14. ABSTRACT

15. SUBJECT TERMS
16. SECURITY CLASSIFICATION OF: 17. LIMITATION OF 18. NUMBER 19a. NAME OF
ABSTRACT OF PAGES RESPONSIBLE PERSON
a. REPORT b. ABSTRACT c. THIS PAGE Same as 5
unclassified unclassified unclassified Report (SAR)
Standard Form 298 (Rev. 8-98)
Prescribed by ANSI Std Z39-18
Requester
Organization
Organizations Coordination
The rapid growth of Response for the top events or
crowd sourcing applica- to cover the majority of
Interagency map
tions for disaster relief Request Crowdsourcing events.
profits from the develop- Third, providers can in-
ment of online social media clude geo-tag information
Public
that provides an open, con- for messages sent from
venient platform that can some platforms (such
collect data from various as Twitter) and devices
sources in a short time. (including handheld smart
Figure 1. Interagency map. The map works as an intermediary
As an example of an phones). Such crowd-
between the public and relief organizations. Requests are
attempt to adopt crowd- collected via social media crowdsourcing. Organizations can then sourced data can help
sourcing to help the relief take actions, share information, and coordinate with each other relief organizations ac-
community to enhance co- using the information on the map. curately locate specific
operation, Ushahidi (www. requests for help. Fur-
ushahidi.com) is an open source cri- of this free emergency number was thermore, visualizing this type of
sis map platform created in 2007 and spread through local and national data on a crisis map offers a common
deployed in locations such as Kenya,7 radio stations. disaster view and helps organiza-
Mexico, Afghanistan, and Haiti. It As of 25 January, the Haiti crisis tions intuitively ascertain the current
leverages Web 2.0 technologies to in- map had more than 2,500 incident status.
tegrate data from multiple sources— reports, with more reports being
phones, Web applications, email, and added every day. The large amount of Shortfalls of Crowdsourcing
social media sites such as Twitter and nearly real-time reports allows relief for Disaster Relief
Facebook—to provide an up-to-date, organizations to identify and respond Although crowdsourcing applica-
publicly available crisis map that is in to urgent cases in time. tions can provide accurate and timely
turn available to relief organizations. Second, crowdsourcing tools can information about a crisis, current
This platform uses crowdsourcing for collect data from emails, forms, crowdsourcing applications still fall
social activism and public account- tweets, and other unstructured meth- short in supporting disaster relief
ability to collectively contribute in- ods and then do rudimentary analysis efforts.10 Most importantly, current
formation, visualize incidents, and and summaries, such as by creating applications do not provide a com-
enable cooperation among various tag clouds, trends, and other filters. mon mechanism specifically designed
organizations. These can help partition the data for collaboration and coordination
into bins (such as most-frequently re- between disparate relief organiza-
Advantages of quested resources) and requests into tions. For example, microblogs and
Crowdsourcing predetermined, most-urgent catego- crisis maps do not provide a mecha-
for Disaster Relief ries (such as medical help, food, shel- nism for apportioning response re-
Compared to traditional relief meth- ter, or people trapped). Relief agen- sources, so multiple organizations
ods, leveraging crowdsourcing for cies can then concentrate on the might respond to an individual re-
disaster relief has three advantages. issues and events that are most quest at the same time.
First, crowdsourced data including important to the relief effort. A second shortcoming is that data
user requests and status reports are Figure 2 illustrates the food re- from crowdsourcing applications,
collected almost immediately after a quests on Ushahidi-Haiti, and Figure 3 while useful, do not always pro-
disaster using social media. Ushahidi- shows the most affected locations vide all the right information needed
Haiti was set up two hours after during the Japanese tsunami based for disaster relief efforts. The ac-
the 12 January earthquake by on the number of reports mapping curacy of the report’s geo-tag and
volunteers from Tufts University on Ushahidi’s crisis map. Using these content is not guaranteed, although
in Medford, Massachusetts.8 Soon maps, relief organizations can co- relief workers greatly need the abil-
after, organizations were able to bor- ordinate resource distribution and ity to automatically and accurately
row a short message service (SMS) make better decisions based on their locate crowdsourced data on the cri-
short code phone number (Mission analysis of crowdsourced data. Fall- sis map. That is, a geo-located tweet
4636) to send free SMS texts.9 News back plans can be further developed does not necessarily refer to the
may/june 2011 www.computer.org/intelligent 11
IS-26-03-Cpss.indd 11 24/05/11 1:03 PM
Figure 2. Food requests after the Haiti earthquake. The Ushahisi-Haiti crisis map helps organizations intuitively ascertain where
supplies are most needed.
geo-location point. Someone might groups, so publicizing details about help locate a report’s actual spatial
text the crisis map’s phone number relief efforts could endanger them. features.
to report something they saw earlier, In addition, research has shown the
possibly texting from a shelter about Challenges and Research need to leverage the information avail-
a bridge that collapsed 10 miles down Issues able from the response group. For ex-
the road. Furthermore, fraud reports Social media as a crowdsourcing ample, to supplement crowdsourcing
from malicious persons might appear mechanism provides aggregate situ- information, groupsourcing10 collects
as normal requests on crisis map. In ational awareness, important and disaster information such as a relief
addition, there are often duplicate new communications pathways, and situation summary, validated geo-
reports, and information essential some opportunities for assistance on tag information, and transportation
for relief coordination is not readily an individual level. However, to make conditions provided by a sanctioned
available or easily accessible, such as crowdsourcing a useful tool, we must group consisting of individuals with
lists of relief resources or communi- address several challenges to leverage disparate resources, goals, and capa-
cation procedures and relief organi- both the data and communications bilities working at the disaster scene.
zation contact information. capabilities, including sensemaking,
Lastly, current crowdsourcing ap- security, and coordination. To tackle Report Verification
plications do not have adequate secu- these problems, we can employ text To some degree, crowdsourcing can
rity features for relief organizations mining and social computing technol- automatically filter reports through
and relief operations. For example, ogies by managing social knowledge photos, videos, and comments from
crowdsourcing applications that are and modeling social behavior during other reports. Ushahidi, for exam-
publicly available for reporting are the disaster relief.11 ple, lets users verify a report by click-
also publicly available for viewing. ing a verification button. It leaves the
Although it is important to provide Geo-tag Determination verification problem to the crowds
information to the public, this can As we mentioned earlier, crowd- to get collective feedback. However,
create conflicts when decisions must sourced data might include inaccu- this strategy—an open self-adjusting
be made about where and when re- rate geo-tags, or might not include method, in technical terms—depends
lief resources are needed. In some them at all. Thus, using social min- on a large number of people validat-
circumstances, relief workers them- ing to consider both a reporter’s so- ing the reports. For Ushahidi maps,
selves might be targeted by nefarious cial and physical networks could however, generally only a few people
12 www.computer.org/intelligent Ieee InTeLLIGenT SySTemS
IS-26-03-Cpss.indd 12 24/05/11 1:03 PM
Figure 3. Ushahidi’s Japanese tsunami map. By visualizing the most affected locations, the map helps relief organizations
coordinate and prioritize their responses.
serve as verifiers, across the spectrum Because the volume and tempo of Scientific data on earthquakes,
of Ushahidi map cases. A fraudulent messages in a massive disaster will be floods, and other phenomena could
message could conceivably propagate so high, automated summaries and augment user data to populate mod-
itself quickly with good camouflage, analysis to make sense of hundreds, if els to predict the time and location
especially in a chaotic disaster envi- not thousands, of unstructured mes- of future requests and needs. Crowd-
ronment, unless better incentives for sages will be another greatly needed sourced data plus model results could
verification are present. capability. help first responders, governmental
As crisis maps and similar crowd- agencies, and NGOs get ahead of the
sourcing applications are developed, Spatial-Temporal Mining for Social demand curve and become more pro-
it will be necessary to develop trust Behavior Prediction active in deploying aid and rescue ca-
management systems that can recog- Currently, volunteers and victims pabilities. One possibility is to design
nize and report questionable postings on the ground report incidents and spatiotemporal classifiers for mod-
and flag them for more scrutiny and requests—that is, people who have a els that will assist in such demand
verification. phone or other communication de- forecasts.
vice. There are no applications for
Automated Report Summarization forecasting future requests or for fill- Scalability and Safety
Police departments and other first re- ing in the likely requests that would Scalability is a problem for many intel-
sponders will need summarized re- come from blacked-out areas if such ligent distributed systems because the
quest reports and the ability to drill people experience limited communi- increasing number of requests, agen-
down to individual tweets and texts. cation ability. cies, and agent behaviors significantly
may/june 2011 www.computer.org/intelligent 13
IS-26-03-Cpss.indd 13 24/05/11 1:03 PM
enlarges the dimension of the search be designed and leveraged for system 3. A. Bedrouni et al., Distributed Intelli-
space.3 Large-scale data-management evaluation and improvement. Creating gent Systems: A Coordination Perspec-
technology will be necessary to main- an automatic recommendation system tive, Springer, 2009.
tain a system’s stability. that helps improve resource distribu- 4. R. Goolsby, “Social Media as Crisis
Safety is another problem that tion coordination and transportation- Platform: The Future of Community
must be considered in a disaster relief route suggestions is another research Maps/Crisis Maps,” ACM Trans. Intel-
management plan, especially in the direction that can utilize real-time ligent Systems Technology, vol. 1, no.
design of a crowdsourcing system. data collection and relief history. 1, 2010, article no. 7.
While making data publically avail- There is no doubt that new coordi- 5. J. Howe, “The Rise of Crowdsourcing,”
able, such systems must protect the nation platform and crowdsourcing Wired, vol. 14, no. 6, 2006, pp. 1–4.
NGOs’ privacy and ensure the safety applications such as crowdsourced 6. J. Surowiecki, “The Wisdom of
of their workers. translation, filtering, and categoriza- Crowds,” Am. J. Physics, vol. 75, no. 2,
C tion will continue to become avail- 2007, p. 190.
rowdsourcing integrated with cri- able and will significantly contribute 7. P. Meier and K. Brodock, “Crisis
sis maps has been a powerful tool in to future disaster relief efforts. We Mapping Kenya’s Election Violence,”
humanitarian assistance and disaster are confident that great numbers iRevolution blog, 23 Oct. 2008; http://
relief. Future crowdsourcing applica- of lives will be saved as govern- irevolution.net/2008/10/23/mapping-
tions must provide capabilities to bet- ments and NGOs find better ways to kenyas-election-violence.
ter manage unstructured messages employ the integration of social 8. J. Heinzelman and C. Waters,
and enhance streaming data. En- media, crowdsourcing, and collabora- Crowdsourcing Crisis Information in
hancements will also include methods tive tools into disaster relief systems. Disaster-Affected Haiti, US Inst. of
of fusing information with data from Peace, 2010.
other sources, such as geological data References 9. CrowdFlower, “Mission 4636,” 2010;
about fault lines and reports of vari- 1. J. Morgan, “Twitter and Facebook www.mission4636.org.
ous earthquake aftershocks. Research- Users Respond to Haiti Crisis,” BBC 10. H. Gao et al., “Promoting Coordina-
ers are just beginning to consider trust News, 15 Jan. 2010; http://news.bbc. tion for Disaster Relief: From
management systems to help evaluate co.uk/2/hi/americas/8460791.stm. Crowdsourcing to Coordination,”
messages with respect to the huge 2. Digital Blogger, “Japan Tsunami Proc. Int’l Conf. Social Computing,
troves of messages being traded in and Earthquake Use of Twitter, Behavioral-Cultural Modeling, and
cyberspace and via cell phones. Facebook, Skype and Other Social Prediction, Springer-Verlag, 2011,
Furthermore, metrics that gauge the Networks,” 12 Mar. 2011; http:// pp. 197–204.
success of crowdsourcing and coordi- bubblecube.wordpress.com/2011/03/12/ 11. F.-Y. Wang et al., “Social Computing:
nation systems for disaster relief will japan-tsunami-earthquake. From Social Informatics to Social Intel-
ligence,” IEEE Intelligent Systems,
vol. 22, no. 2, 2007, pp. 79–83.
Huiji Gao is a PhD student in the Data
Mining and Machine Learning (DMML)
lab at Arizona State University. Contact him
at [email protected].
Our experts.
Your future. Geoffrey Barbier is a doctoral candi-
date in computer science and a student in
the Data Mining and Machine Learning
(DMML) lab at Arizona State University.
Contact him at [email protected].
Rebecca Goolsby is an anthropologist for
www.computer.org/byc
the US Office of Naval Research. Contact
her at [email protected].
14 www.computer.org/intelligent Ieee InTeLLIGenT SySTemS
IS-26-03-Cpss.indd 14 24/05/11 1:03 PM