1. Information system to assist survivors of disasters
Emergency information system designed to coordinate information flow between the
population and aid agencies during disasters
Nicolas di Tada Timothy Large
InSTEDD Thomson Reuters Foundation
Palo Alto, USA London, UK
nditada@instedd.org timothy.large@thomsonreuters.com
Abstract— This paper describes a free and open source • Reach out to affected population with useful and
information system designed to be deployed in emergencies actionable information in a timely manner.
caused by sudden onset natural disasters. The aim is to
streamline the communication flow and collaboration between III. WORKFLOW
media, aid workers and government agencies with the affected
population, to help the latter get verified, accurate and The overall goal of the designed flow is to streamline the
actionable information that will enable them to make decisions creation and maximize the availability of relevant and credible
and recover from the disaster. information to communities. This implies the creation and
distribution of such information. The flow improves when
The EIS system also provides means for affected population and people have not only direct access to information, but the
field workers to channel vital data back up into aid response. benefit also of credible intermediaries to help discover, gather,
This tool is part of a free information service run by Thomson compare, contextualize, and share information. [2]
Reuters Foundation to help survivors of natural disasters. It will
The flow (“Fig. 1”) can be divided in 4 stages:
serve the affected populations, local media and relief responders
by providing fast, practical and verified information in local 1. Collection of raw reports from the field.
languages through the best means available.
2. Filtering, processing and identification of important
Keywords-component; Information, disaster, response, information.
communication, collaboration, media, aid.
3. Editorial process including verifying the accuracy of
I. INTRODUCTION information, editing and translating the information into
local languages.
Underlying EIS is the assumption that people caught up in
disasters are not helpless victims and that life saving, 4. Inform relevant stakeholders.
actionable information is as important as blankets and
tarpaulin. As first responders, they need reliable information A. Subscription and Collection
to make decisions and minimize the impact. Citizens can report their location by sending an SMS to
the system with the name of the village they are. That simple
‘Information is a vital form of aid in itself […] Disaster- action registers them in the system to receive alerts targeted to
affected people need information as much as water, food, that village.
medicine or shelter. Information can save lives, livelihoods
and resources.’ [1] Every citizen, either registered or not, can report
information to the system through SMS.
II. OBJECTIVES
The system allows submission of raw reports through
In line with those assumptions, the design objectives have several channels:
been to:
• SMS: through mobile aggregators (like Clickatell)
• Collect reports originated from the very first layer of or plugged-in phones. People can send information to
affected population - survivors; combining that with a specific number, which then stores the message in
information received from the field including aid the system.
agencies, local media and government.
• Twitter: both streams from particular users or
• Provide fast and reliable means to filter, search and certain hash-tags can be imported.
detect important information in the pool of reports.
• Streamline the flow of information inside a
• Email: any number of email addresses can be
checked by the system and directed to specific
distributed editorial team including local translators. collection baskets or inboxes

2. Figure 1. Information workflow
• RSS: feeds can be checked periodically and new the most salient ones in these kinds of scenarios. These tools
articles will appear in the system. include search, tagging and automatic extraction of the
location the text refers to. Other features that can be added to
• Web: any user with permissions can post customize the experience and workflow include flagging
information using the web interface. documents, hiding them, commenting and relating them. The
1) Bounded and Unbounded Crowd-sourcing objective of these features is to aid in the process of
Through the use of a feature we called "Working groups" collaboratively selecting useful information to initiate the
and in combination with the "Baskets", both bounded and editing workflow.
unbounded crowd-sourcing processes can coexist in the C. Editing
stream.
A module called “Baskets” allows the configuration of a
Identified users can be assigned to "Working Groups" and custom workflow tailored to the specific team and situation.
the reports coming from those groups, can be configured to go The module works based on the idea of baskets. Each basket
to specific baskets. Therefore, some baskets can be monitored is a collection of pieces of information (items), that acts both
more often or by different people, depending on the working as inbox and outbox. Each user in the system is given
groups (bounded crowd) putting reports there. An unbounded permissions to read, write, move in or move out items from
basket could be even configured to be self-filtered by the one or more baskets.
same unbounded crowd and items with 10 or more people
marking them as "alert" or a specific tag could then be pulled Typically, a user would monitor one or more baskets based
by another basket, which is monitored by some agency. on his role. As new items appear, he will decide what to do,
act on the item, flag it as urgent, write a comment, create a
Using this combination of features a group of users can new alert or translate it to another language. He would then
collaborate around a pool or flow of information sifting useful put the new item or the edited one in another basket for
information, tagging (or correcting tags suggested by the someone else to do his part of the job.
system), classifying as urgent or alerts, adding geographic
information (or correcting the one suggested by the system) D. Informing
and several other types of metadata that modules provide. Once an alert is ready to be sent out to relevant recipients,
the system allows the user to target affected population by
B. Selection location through a map with references of clusters of citizens.
Through a number of different tools, users monitoring the The user sees a map with the different groups of citizens and
pool of raw reports can filter important information. The goal the number of citizens in each group. “Fig. 2”
in this stage is to try to deal with as many situation awareness
“demons” [3] as possible, being data overload and stressors The objective of targeting by location is not only to keep
the information relevant to everyone, but also to avoid that

3. Figure 2. Sending information to citizens
people in a desperate situation end up travelling to a distant A complete picture of the platform architecture and
location just because they received a message saying there is interactions can be seen in Figure 3.
food or blankets there.
The choice of creating a new tool was based on the
The alert can be sent through SMS and Email, and can following observations:
also be addressed to specific working groups, like aid
workers, local media or government agencies. 8. Existing free open source information aggregators were
based on more generic solutions like content
IV. TECHNOLOGY management systems. This creates the problem of a
platform too generic and abstract to provide a productive
The EIS system was built on top of RIFF (http://
instedd.org/evolve), an InSTEDD platform tool developed as and friendly environment to develop extensions. Riff is
a general collaboration environment for content creation, collaboration tool based around information aggregation
social metadata annotation, and automated analysis with and provides specific extension points that allow the
potential applicability in a wide range of areas. developer to focus only in the intended task or feature.
RIFF is a modular and extensible free open source 9. Machine learning classifiers are usually stand-alone ad-
platform built in C#. The platform tool was developed during hoc applications or libraries, without a context integrated
2008 and 2009 as an effort to provide an environment that with information aggregation and collective analytics.
could collect heterogeneous information sources and leverage
10. Many powerful analytic environments are either
the power of human expert collaboration with machine
learning and visualization aids. proprietary or commercial.
RIFF is extensible through several different means: EIS was designed and implemented as a set of extensions
to RIFF that allow receiving SMS messages, creating custom
5. Developers can add new types of information sources, workflows for information and targeting outgoing SMS
even custom or proprietary messages by location and working groups.
6. Modules can be developed to add new filters, provide The choice of SMS as a communication channel was based
additional functionality around classifying or tagging on:
items, extend the model of each item by adding new
fields or properties, create new visualizations and • Provides a two-way channel: citizens can report as
automatically extract information during the import well as receive information.
phase. Most of the basic functionality is developed • Allows for information to be targeted to citizens
through modules. based on their reported location and needs, while still
allowing for bulk sending.
7. Plugging the output of RIFF to another system. The tool
allows the extraction of tailored and filtered RSS feeds, • Allows for semi-structured reports to be parsed:
email subscriptions or push selected data to custom locations can be extracted, keywords can be identified
endpoints. and other entities like names or organizations can be

4. Figure 3. RIFF Architecture

5. extracted. team managed to process the information, select relevant
topics and going through a quick but reliable loop with local
A. Server Deployment translators could provide alerts with verified data.
EIS is running right now on an Amazon EC2 instance,
providing a very easy way for 3rd parties to quickly deploy A short summary of the pilot can be seen at: http://
and manage their own instances. www.alertnet.org/thefacts/reliefresources/125907780276.htm
V. RESULTS VI. CONCLUSION
As of the moment of presenting this paper, the system was We believe that we have built a tool that can handle
deployed with much success in Haiti: information collection, flow and reach out in a wide range of
deployment scenarios in disaster situations, providing the
• The population response was very high, getting support for channeling information up and down as needed
more than 5.000 subscribers in less than 10 days. among a heterogeneous ecosystem of stakeholders. The
Currently above 15.000. system will serve as the basis for further learning,
• More than 1.000 incoming SMS per day from the development and work on the subject.
survivors, including emergencies, needs or VII. ACKNOWLEDGMENTS
information requests to a total of more than 50.000
We would like to thank to all the people that worked in the
messages processed inside the system.
inception, design, development and testing of the system:
• Several agencies, under the UN umbrella, are both Monique Villa, Timothy Large, Astrid Zweynert, Joanne
receiving targeted information from citizens, and Tomkinson, Thin Lei Win (Thomson Reuters Foundation);
sending out messages regarding where to find food, Eric Rasmussen, Robert Kirkpatrick, Eduardo Jezierski, Luke
register missing people, get shelter and health Beckman (InSTEDD); Juan Wajnerman, Brian Cardiff, Martín
measures. Verzilli, Ary Borenszweig, Andres Taraciuk, Mariana Galvan,
Martin Scebba (Manas Technology Solutions).
• More than 20 users from different agencies and
groups. The help and feedback of the Palang Merah Indonesia
(Indonesian Red Cross) was essential to conduct the pilot and
Within hours of the earthquake, most of Haiti’s cell phone
improve the system.
towers were still operational and text messages were getting
though in most of the country but Port-Au-Prince. After 2 or A special thank to Willie Smits who hosted us and helped
3 days of work, the 2 main telephone companies could get us with the logistics in North Sulawesi.
most of the capital antennas back to work.
VIII. REFERENCES
The EIS team was deployed less than 48hs after the
earthquake and, in partnership with other organizations like 1. International Federation of Red Cross and Red Crescent
Ushahidi and FronlineSMS:Medic, quickly began (2005) World Disasters Report 2005: Focus on
negotiations with the telephone companies to establish a information in disasters, Kumarian Press, http://
shortcode for citizens to text-in and for the system to be able www.ifrc.org/publicat/wdr2005/index.asp
to send messages out. A diagram of the final ecosystem 2. The Knight Comission (2009) Informing Communities:
architecture of applications and organizations interacting Sustaining Democracy in the Digital Age, http://
around the shortcode and EIS can be seen in Figure 4. report.knightcomm.org/
More at: http://www.alertnet.org/db/blogs/ 3. Mica R. Endsley (2003) Designing for Situation
1564/2010/00/24-120746-1.htm Awareness: An approach to user-centered design, CRC
Before the Haiti deployment, a pilot of the system was 4. Ushahidi: Project 4636 revisited, http://
conducted in Tomohon, North Sulawesi Indonesia, in blog.ushahidi.com/index.php/2010/02/11/project-4636-
partnership with the Local Red Cross and the local Search and revisited-the-updated-info-graphic/
Rescue Team, has proved the validity of our assumptions.
Even if the incoming stream of reports from the field
contained (by design) a low signal to noise ratio, the editorial

6. Figure 4. Haiti deployment ecosystem of applications around the shortcode and EIS [4]