Presented by Adele Waugaman, Harvard Humanitarian Initiative Fellow & Independent Consultant. October 15, 2012. Washington, D.C. "DC Design Week: Design for Disaster Relief," held in partnership with AIGA DC.
6. Case Study: Crowdsourcing
Geolocation Information
Using Crowdsourcing to Locate Hospitals on Maps
One of more difficult Facts: Examples Haiti operation Relief 2.0 Activities in the health
Fast problems during the of Disaster was determining the location of
facilities. OCHA turned to the volunteer and technical communities to map these facilities, asking Crisis
2010 Haiti Earthquake Response
Mappers and Sahana—examples of the Volunteer & Technical Communities—if they could crowdsource
the effort to geo-locate 105 health facilities thatLocate Hospitals The Maps to the crowd went
Using Crowdsourcing to had no location data. on request
One of more difficult problems during the Haiti operation was determining the location of health
out at 2:40AM on January 22. Approximately volunteer and technical communities to map these facilities,
facilities. OCHA turned to the 35
asking Crisis Mappers and Sahana—examples of the Volunteer & Technical Communities—if
hours later, the team working on the problem had located the de facto list of 102 of the 105 missing
they could data into the Sahana to geo-locate 105 health facilities verified each
hospitals, inputting all the crowdsource the effortdisaster management system. They had that had no location
facility by having an OpenStreetMap member went out at hospital or clinicJanuary 22. Approximately 35
data. The request to the crowd locate the 2:40AM on on high-resolution satellite
imagery at 15cm resolution and verify working on the problem located at the submitted coordinates. the 105
hours later, the team that health facility was had located the de facto list of 102 of
missing hospitals, data formats, data into the Sahana disaster management system. They
Sahana made the data available in openinputting all thewhich became one of the best resources for
had verified each facility by having an OpenStreetMap member locate the hospital or clinic on
health facility data for the next satellite imagery at 15cmunique individualsverify that health facility was
high-resolution month. More than 8,000 resolution and visited the site or pulled
locatedfrom the feed. Crowdsourcing transformedtheproject that normally would require several days into just
at the submitted coordinates. Sahana made a data available in open data
formats, which became one of the best resources for health facility data for the
over one day.
next month. More than 8,000 unique individuals visited the site or pulled
from the feed. Crowdsourcing transformed a project that normally would require
several days into just over one day.
Translating Creole through Mission 4636
Mission 4636 facilitated the collaboration that a team at Stanford established with
members of the Haitian diaspora to translate thousands of messages from Creole
into English. Haitian Creole has only 12 million speakers worldwide, 9
7. facilities. OCHA turned to the volunteer and technical communities to map these facilities,
asking Crisis Mappers and Sahana—examples of the Volunteer & Technical Communities—if
they could crowdsource the effort to geo-locate 105 health facilities that had no location
Case Study: Crowdsourcing
data. The request to the crowd went out at 2:40AM on January 22. Approximately 35
hours later, the team working on the problem had located the de facto list of 102 of the 105
Translations
missing hospitals, inputting all the data into the Sahana disaster management system. They
had verified each facility by having an OpenStreetMap member locate the hospital or clinic on
high-resolution satellite imagery at 15cm resolution and verify that health facility was
located at the submitted coordinates. Sahana made the data available in open data
Translating Creole through the 4636 Shortcode
formats, which became one of the best resources for health facility data for the
next month. More than 8,000 unique individuals visited the site or pulleda team at Stanford established with
Mission 4636 facilitated the collaboration that
from the feed. Crowdsourcing transformedHaitian diaspora towould require
members of the a project that normally translate thousands of messages from Creole
several days intointo English. Haitian Creole has only 12 million speakers worldwide, 9
just over one day.
million of which live in Haiti. In the first days of the disaster, Rob Munro (a
Translating Creole through Mission 4636
Mission 4636 facilitated the collaboration that a team Stanford)established with
computational linguist at at Stanford recruited Creole speakers from Facebook and other
members of the Haitian diaspora to sites, asking them to begin translating tweets, SMS messages, and
public web translate thousands of messages from Creole
into English. Haitian Facebook posts million speakers worldwide, 9
Creole has only 12 coming from Haiti.
million of which live in Haiti. In the first days of the disaster, Rob Munro (a computational linguist at Stanford)
recruited Creole speakers from Facebook and other public web sites, asking them to begin translating tweets, SMS
More than 130,000 text messages went through 4636 during the first month of the
messages, and Facebook posts coming from Haiti. More than 130,000 text messages went through 4636 during
response. At its peak, the effort had 1,200 Haitians translating thousands of
the first month of the response. At its peak, the effort had 1,200 Haitians translating thousands of
messages per day, usually within day,minutes of their arrival. By February, their arrival. By February, Rob began
messages per 4.5 usually within 4.5 minutes of Rob began a partnership with
Crowdflower, a social venture for microtasking that was working social venture another social venturethat partners
a partnership with Crowdflower, a with Samasource, for microtasking with was working
on the ground who could recruit Haiti citizens to perform translation services from Haiti. Theon the ground who could
with Samasource, another social venture with partners effort transitioned to this
professional translation platform soon thereafter.
recruit Haiti citizens to perform translation services from Haiti. The effort transitioned
to this professionalin Haiti Provides a Vital Link
Cell Connectivity translation platform soon thereafter.
According to a report by the US Institute of Peace, “approximately 85 percent of Haitian
households had access to mobile phones at the time of the earthquake, and although 70 percent
of the cell phone towers in Port-au-Prince had been destroyed in the disaster, they were quickly
repaired and mostly back online before the 4636 number was operational.” Haitians sent
hundreds of thousands of messages out via SMS to Twitter, Facebook, Ushahidi, the
8. Challenges with Linking Volunteer
Networks & Aid Agencies
• Protocol
• Capacity building (training, setting expectations)
• Standards (protection, privacy, data security)
• Push v pull
• Funding
• Sustainability (workflow transition)
• Relevance (engagement of local actors)
OCHA Humanitarian Graphics
11. Challenges with Direct Citizen
Engagement in Humanitarian Assistance
• Coordination
• Feedback
• Protocol
• Capacity building (training, setting expectations)
• Standards (protection, privacy, data security)
• Push v pull
• Funding
• Sustainability (workflow transition)
• Relevance (engagement of local actors)
WFP/John Wreford
WFP/John Wreford
Notas del editor
Technology critical to disaster preparedness & response in a range of ways I’ll be talking about how new info & communications technologies (ICTs) are influencing the way disaster preparedness and relief programs are designed.How many of you were in DC for the earthquake last August?Information is essential in emergencies – a public good, itself a vital form of aid ICTs, mobile in particular, a critical conduit for info delivery, esp in LMICs
Natural disasters around the world last year caused a record $380 billion in economic losses. That's more than twice the tally for 2010, and about $115 billion more than in the previous record year of 2005According to a report from Munich Re, a reinsurance group in Germany.Since 1980, number of severe floods has almost tripled, and storms have nearly doubled, in part, to the impact of climate change With increasing changes to natural environment like rising sea level and drought, there will be increasing confluencebetween natural disasters and political conflict fueled by competition over access to scarce resources
Traditionally disaster preparedness response approached from a top-down, command and control system (UN, gov, corps & CSOs, citizens)Ex: UN system: diff agencies, challenge of coordinationIncredibly complex systems, operating in emergency environments that are highly fluid, with high levels of complexityEx: Haiti, which I’ll talk more about in a moment, gov and UN building destroyed, virtual absence of data
Challenging this top-down, command & control model – rise of ICTs, esp mobile6.2B mobile (fastest-growing markets are LDCs)Mobile = web accessFacebook 1B active users/month, equiv 3rd largest country (china 1.3B, India 1.2B, USA 315M)Enabling networksof volunteersEmpowering communities, individuals- giving them agency to access information they need to make well-informed decisions
2009: New Tech. Early warning, preparedness, response, rebuilding – examples. 1-M: broadcast: 1-1: mobile voice SMS; M-M: social networks, mapping, crowd-sourcing. Then Haiti happens.2011: DR2.0 Social networks, geospatial imagery, crisis mapping, open data. Case study: Haiti What worked: Enormous international support, some V&TC efforts more successful than others. UshahidiTufts student-driven, mapped a lot of infoWhat didn’t work: Not a typical disaster, no activegov presence in early day. Ushahidi: but no plug-in with humanitarian system “responsibility to respond”?What do we need to do today to strengthen the future of resilience? Two reports address the question Dec 09: New technologies and innovative uses of existing technologies are improving crisis preparedness, response, and prevention. Yet barriers to bringing these ideas to scale remain, and pose potential risks and rewards.Shift from one-to-many info flows (e.g. TV, radio) to many-to-many info flows (social networks/social media)Info flows now two-way. Beneficiary communities now have a tool to convey their needs. Can overwhelm or inform a relief effort.March 11: Disaster Relief 2.0 report looked at what rise of real-time information environment (social networks, cloud-based computing, mobile web, etc) means for information-sharing in the aftermath of complex humanitarian emergenciesWe looked at the response to the Jan ‘10 Haiti quake: How was the humanitarian system (UN, NGOs, govs) able to work with web-based non-humanitarian actors like the crisis mapping community (VTCs)?What do large institutions need to do to adapt to changed environment; how to adopt new tools strategies, policies?Significant challenges: authentication/verification, privacy, accuracy/trust Recommended 5-part framework to establish coordination with VTC and HTC (neutral forum, innovation space, deployable field team, research & testing consortium, clear operational interface)
Used OCHA’s 3WsPulled data from publicly-available sources24 hour delay and other tactics to ensure any secure info pulled before made publicBy the time we went to print, already out of date. Libya crisis saw first systemic and open collaboration between HS and VTCsOCHA, UNOSAT and NetHope collaborating with the Volunteer Technical Community (VTC) specifically CrisisMappers, Crisis Commons, Open Street Map, and the Google Crisis Response Team Addressed several challenges:Verifying credibility of VTC contributors: the CrisisMappers Standby Task Force created. They mapped social media, news reports and official situation reports from within Libya and along the borders at the request of OCHA. Protecting privacy: The public version of this map does not include personal identifiers and does not include descriptions for the reports mapped. All information included on this map is derived from information that is already publicly available online.For security reasons, reports were placed on a 24 hour embargo before being made public.
Citizens are the true first responders, but large agencies aren’t tooled to engage directly with the public. Increasingly disaster affected populations have access to SMS and social media tools to relay their needs in near real-time. The convergence of broadband, multimedia and mobile makes it increasingly easy for these populations to be eyes and ears, gathering and reporting information about urgent needs. This ready access to communications tools coupled with the tremendous need to get and share information in emergencies is generating enormous quantities of new data. If decision makers wish to have access to (near) real-time assessments of complex emergencies, they need to figure out how to process information flows from thousands of individuals.