Abstract. In disasters such as the earthquake in Haiti and the tsunami in Japan, people used social media to ask for help or report injuries. The popularity, effciency, and ease of use of social media has led to its pervasive use during the disaster. This creates a pool of timely reports about the disaster, injuries, and help requests. This offers an alternative opportunity for first responders and disaster relief organizations to collect information about the disaster, victims, and their needs. It also presents a challenge for these organizations to aggregate and process the requests from different social media. Given the sheer volume of requests, it is necessary to filter reports and select those of high priority for decision making. Little is known about how the two phases should be smoothly integrated. In this paper we report the use of social media during a simulated crisis and crisis response process, the ASU Crisis Response Game. Its main objective is to creat a training capability to understand how to use social media in crisis. We report lessons learned from this exercise that may benefit first responders and NGOs who use social media to manage relief efforts during the disaster.

1. Lessons Learned in Using Social Media for
Disaster Relief
- ASU Crisis Response Game
Mohammad Ali Abbasi, Shamanth Kumar, Huan Liu
Computer Science and Engineering, Arizona State University
Jose Augusto Andrade Filho
Department of Computer Science-ICMC, University of Sao Paulo
Data Mining and
Machine Learning Lab

2. Social Media, Disaster Relief, Game
• People used Social Media to send their report or requests
during disasters or emergencies
– Haiti earthquake, Hurricane Irene, Tsunami in Japan
• Twitter was one of the first sources of eyewitness
information during the Mumbai terror attacks in 2008
• Many HA/DR agencies are planning to use social media in
disaster recovery efforts
• To test Social Media for Disaster Relief effort in a
simulated controlled environment we designed a ASU
Crisis Response Game
Data Mining and
Machine Learning Lab
2

3. Social Media, Disaster Relief, Game
• Social Media in disasters or crises
– Haiti earthquake (2010)
– Hurricane Irene (2011)
– Tsunami in Japan (2011)
• We designed ASU Crisis Response Game to:
Create a training capability to understand
the effective use of social media in crisis
Data Mining and
Machine Learning Lab
3

4. Game Scenario
• A disaster or an emergency happened
• People use social media to report
damages or ask for help
• NGO
– Collect data from social media
– Filter and rank requests
– Create missions
– Send First responders to help victims
Data Mining and
Machine Learning Lab
4

5. ASU Crisis Response Game
Data Mining and
Machine Learning Lab
5

6. Data Mining and
Machine Learning Lab
6

7. ASU Crisis Response Game
• 75 Volunteers played for 4 hours
• 25 Teams of victims
• 8 Teams of First-responders
• 20 People in Command center
• 7 Different locations in ASU campus
Data Mining and
Machine Learning Lab
7

8. Game Exercise and Lessons Learned
• Collecting Tweets and Short Messages
– Two redundant systems
• Filtering system (Manual or Automatic?)
– Clustering, Spam detection, Ranking
• Educating people
– People don’t know how to use social media during crisis
• GeoLocation information
– Usually not included
• Language
– Text analysis tools usually can handle English
Data Mining and
Machine Learning Lab
8

9. Thanks!
Acknowledgments:
This research was sponsored in part by the Office of Naval Research
We would also like to thank
• Members of DMML Lab and all volunteers from ASU
• Our collaborators Catherine Graham from Humanity Road Inc.
• Mark Bradshaw and his team from NSWCDD, Office of Joint Staff
• Dr. Rebecca Goolsby from ONR
• Professor Kathleen Carley and her team from CMU
Data Mining and Mohammad-Ali Abbasi
Machine Learning Lab
ali2@asu.edu 9