Connecting Grassroots and Government for Disaster Response

policy series vol 3

by John Crowley Connecting Grassroots and Government for Disaster Response

by John Crowley, Public Policy Scholar Commons Lab, Wilson Center October 2013 CONNECTING GRASSROOTS AND GOVERNMENT FOR DISASTER RESPONSE

Commons Lab Science and Technology Innovation Program Woodrow Wilson International Center for Scholars One Woodrow Wilson Plaza 1300 Pennsylvania Avenue, N.W. Washington, DC 20004-3027

www.CommonsLab.wilsoncenter.org

Study Director: Lea Shanley Editors: Aaron Lovell and Lea Shanley Cover design: Diana Micheli

© 2013, The Woodrow Wilson International Center for Scholars, Washington, D.C. This work is licensed under Creative Commons Attribution-NonCommercial-Share Alike 3.0 Unported License: http://creativecommons.org/licenses/by-nc-sa/3.0/.

The reproduction of this work for commercial purposes is strictly prohibited. Users may copy and distribute this document in any medium noncommercially, provided that this Creative Commons notice is reproduced in all copies. Users may not use technical measures to obstruct or control the reading or further copying of the copies that they make or distribute. Nongovernmental users may not accept compensation of any manner in exchange for copies. This publication may be shared, copied, and distributed for noncommercial purposes provided that the following attribution is given:

Crowley, John. Connecting Grassroots and Government for Disaster Response. Washington, DC: Commons Lab of the Woodrow Wilson International Center for Scholars, 2013.

Available for download free of charge at http://www.wilsoncenter.org/publication-series/commons-lab

2 The Woodrow Wilson International Center for Scholars is the national, living U.S. memorial honoring President Woodrow Wilson. In providing an essential link between the worlds of ideas and public policy, the Center addresses current and emerging challenges confronting the United States and the world. The Center promotes policy-relevant research and dialogue to increase the understanding and enhance the capabilities and knowledge of leaders, citizens, and institutions worldwide. Created by an act of Congress in 1968, the Center is a nonpartisan institution headquartered in Washington, D.C.; it is supported by both public and private funds.

Conclusions or opinions expressed in Center publications and programs are those of the authors and speakers. They do not necessarily reflect the views of the Center staff, fellows, trustees, advisory groups, or any individuals or organizations that provide financial support to the Center.

The Center is the publisher of The Wilson Quarterly and the home of both the Woodrow Wilson Center Press and the dialogue television and radio program. For more information about the Center’s activities and publications, please visit us on the Web at http://www.wilsoncenter.org/.

Thomas Nides, Chairman of the Board Sander R. Gerber, Vice Chairman

Jane Harman, Director, President and CEO

Public Board Members: James H. Billington, Librarian of Congress John F. Kerry, Secretary, U.S. Department of State G. Wayne Clough, Secretary, Smithsonian Institution Arne Duncan, Secretary, U.S. Department of Education David Ferriero, Archivist of the United States Carole Watson, Acting Chairman, NEH Kathleen Sebelius, Secretary, U.S. Department of Health and Human Services Designated Appointee of the President from within the Federal Government: Fred P. Hochberg, Chairman and President, Export-Import Bank

Private Board Members: Timothy Broas; John T. Casteen, III; Charles E. Cobb, Jr.; Thelma Duggin; Carlos M. Gutierrez; Susan Hutchison; Barry S. Jackson

Wilson National Cabinet: Eddie and Sylvia Brown, Melva Bucksbaum and Raymond Learsy, Ambassadors Sue and Chuck Cobb, Lester Crown, Thelma Duggin, Judi Flom, Sander R. Gerber, Ambassador Joseph B. Gildenhorn and Alma Gildenhorn, Harman Family Foundation, Susan Hutchison, Frank F. Islam, Willem Kooyker, Linda B. and Tobia G. Mercuro, Dr. Alexander V. Mirtchev, Wayne Rogers, Leo Zickler The Science and Technology Innovation Program (STIP) analyzes the evolving implications of such emerging technologies as syn- thetic biology, nanotechnology, and geo-engineering. STIP’s research goes beyond laboratory science to explore new information and communication technologies, sensor networks, prediction markets, and serious games. The program provides critical yet nonpartisan research for the policymaking community and guides officials in the design of new governance frameworks. It gauges crucial public support for science and weighs the overall risks and benefits of technology for society at large.

The Commons Lab advances research and non-partisan policy analysis on emerging technologies that facilitate collaborative, science-based and citizen-driven decision-making. New tools like social media and crowdsourcing methods are empowering average people to monitor their environment, collec- tively generate actionable scientific data, and support disaster response. http://CommonsLab.WilsonCenter.org

Commons Lab Staff Lea Shanley, Director, Commons Lab Zachary Bastian, Early-Career Scholar, Commons Lab Ryan Burns, Research Assistant Joe Filvarof, Program Assistant Aaron Lovell, Writer/Editor

The Commons Lab is supported by the Alfred P. Sloan Foundation About the Author

John Crowley is a consultant at the Global Facility for Disaster Reduction and Recovery at The World Bank Group, where he leads partnership development around the Open Data for Resilience Initiative. He is also curator of humanitarian experiments at the Joint Interagency Field Explorations (a.k.a, “Camp Roberts Experiments”), a humanitarian technology accelerator run through partnership of the National Defense University’s Center for Technology and National Security Policy and Naval Postgraduate School. He is an affiliated researcher with the Harvard Humanitarian Initiative, where he was the lead author of the “Disaster Relief 2.0” report for OCHA and the UN Foundation. John holds an MPA from Harvard’s Kennedy School of Government, where he was the Robert C. Seamans, Jr. Fellow in Science, Technology, and Public Policy. He also holds an MA in History of Ideas and an MusB in Cello Performance and Music History & Literature from Boston University. He tweets at @ jcrowley. Acknowledgements

Several experts read this report in draft form and provided criticism and comments that improved the argument and clarified the text. Their candid comments pushed me to develop a final product that meets the highest standards for objectivity and evidence. I wish to thank the following individuals for their review of this report:

• Willow Brugh, Massachusetts Institute of Technology Media Lab | Center for Civic Media and Geeks without Bounds • Joshua Campbell, ABD, Humanitarian Information Unit, Department of State • Michael Gresalfi, Ph.D., Federal Emergency Management Agency • Graham Lampa, Public Diplomacy, Department of State • Heather Leson, Ushahidi • Desiree Matel-Anderson, Esq., Federal Emergency Management Agency • Patrick Meier, Ph.D., Qatar Computing Research Institute and International Conference of Crisis Mappers • Gisli Olafsson, Emergency Response Coordinator, NetHope • Shadrock Roberts, ABD, U.S. Agency for International Development • Andre Verity, Office for the Coordination of Humanitarian Affairs, United Nations • Jen Ziemke, Ph.D., John Carroll University and International Conference of Crisis Mappers

Although the reviewers provided many constructive comments and suggestions, they were not asked to endorse the conclusions or recommendations, nor did they see the final draft of the report before its release. Lea Shanley and Aaron Lovell of the Wilson Center were responsible for making certain that an independent examination of this report was carried out in accordance with institutional procedures and that all review comments were carefully considered.

I would like to thank the following individuals for offering me their wisdom, energy, expertise, and encouragement, and in many cases, for taking the time to grant me an interview or opportunity to clarify existing policies and protocols: David Applegate, Katie Baucom, Samuel Bendett, David Becker, Chad Blevins, Andrew Cedar, Kate Chapman, John Desmarais, Corina DuBois, Shoreh Elhami, Schuyler Erle, Kate Gage, Al Gembara, Erica Hagen, Taha Kass-Hout, Mikel Maron, Rob Munro, Robin Murphy, Amy Noreuil, Ryan Paterson, Eric Rasmussen, Lee Schwartz, John Scott, Mischa Shattuck, Galit Sorokin, Jeanette Sutton, Chris Vaughan, David Warner, Adele Waugaman, Linton Wells II, Benson Wilder, and Andrew Wiseman. In addition, I want to thank all the participants in the Fall 2012 Workshop on Connecting Grassroots and Government, the Federal Emergency Management Agency (FEMA) Innovation Team that deployed to the area damaged by Superstorm Sandy, and the teams at the National Defense University’s Center for Technology and National Security Policy and Naval Postgraduate School’s Field Experimentation Program.

I would also like to thank the Wilson Center’s Zach Bastian, who made detailed and careful edits to the penultimate draft, and Jason Kumar, who served as my research assistant over the summer of 2012.

A special thanks goes to Dave Rejeski, who provided the support and freedom to make this project possible.

And a profound thank you goes to Lea Shanley, who provided the impetus behind the report, pushed to bring me to Washington, D.C., and then drove the report to comple- tion with constant support and useful criticism. This paper would not exist without her vision, perseverance, and attention to detail.

I also owe personal thanks to those who opened their homes to me in several cities while I was in the process of writing this report, as well as to those who quietly opened the interface between grassroots and government in ways that most will never know or understand.

This publication was made possible by a grant from the Alfred P. Sloan Foundation. I am very grateful for the support of everyone who provided insight and advice around this dynamic area of policy, but neither they, the Wilson Center, nor the Sloan Foundation are responsible for the content, views, or data contained in this publication. This report exclusively represents the views of the author, who, of course, retains responsibility for the content, including all errors of fact and interpretation. Dedicated to the citizens of Newtown, Connecticut:

my hometown whose educators and first responders taught us a love of community, introduced us to the joys of public service, and filled our hearts with the courage to make great sacrifices to protect that which is most dear.

And to the citizens of my adopted home, Boston, Massachusetts:

who stand as one in times of great need.

Contents

Preface / 10

Introduction / 13

SECTION 1 / Defining the Context / 21

SECTION 2 / Defining the Grassroots: Who They Are and What They Do / 27 Defining Digital Humanitarian Organizations / 28 Organizational Design / 30 Data Quality: Why Trust DHO-Generated Data? / 36

SECTION 3 / Defining the Grassroots to Government Interface / 39 U.S. State Department/National Geospatial-Intelligence Agency / 41 Federal Emergency Management Agency/Civil Air Patrol / 42 U.S. Agency for International Development / 42

SECTION 4 / Grassroots to Government Framework / 45 Design / 46 Experimentation / 52 Activation / 55

Conclusion / 59

Appendix / 61

Endnotes / 63

9 Preface

Connecting Grassroots and Government agencies—is changing rapidly. Both the for Disaster Response is a testament to earthquake in Haiti and Hurricane Sandy the vision of the staffs at the Woodrow catalyzed efforts by federal agencies to Wilson International Center and the connect into the whole of community Alfred P. Sloan Foundation, who decided response. As a result, I am writing a to study the best ways to address an dynamic document: a text meant to be emerging problem in disaster response. edited like a wiki, rather than consulted After Hurricane Katrina, it is no lon- as a static brick that one scholar has ger acceptable to think of response added to a growing wall of knowledge. operations as the domain of governments alone. Communities have always With the Wilson Center, I will build a performed most of the actions on the wiki to edit the core toolkits and exam- ground. However, these efforts are ine policy issues. I will also drive the often poorly coordinated, both within development of a comprehensive map community activities and with govern- via workshops at the Wilson Center and ment-led operations. Communication other venues, such as the humanitarian technologies—from social media to crisis experiments at Camp Roberts1 under mapping—have improved coordination the Naval Postgraduate School and the at the citizen-to-citizen level for several National Defense University’s Center years. They are just beginning to affect for Technology and National Security the interface of grassroots efforts and Policy, the International Conference of government. This report examines the Crisis Mappers (ICCM), and the Open connections that forward-looking agen- Humanitarian Initiative under NetHope. cies are establishing, setting them out as exemplars of change. This report does not claim to be a new version of Disaster Relief 2.0 (the The interface between the two commu- so-called 2.1 version), but rather a 0.1 nities—grassroots organizations, such version of an interface between the as the Humanitarian OpenStreetMap grassroots and the government around Team (HOT) and the Standby Volunteer crises. Like most version 0.1 software, Task Force (SBTF), and U.S. federal its aim is to provide a framework for

10 CONNECTING GRASSROOTS AND GOVERNMENT FOR DISASTER RESPONSE

This report examines the connections that forward- looking agencies are establishing, setting them out as exemplars of change.

further work by many others. I hope that experts can expand my thinking and correct my errors early in the process. With luck, this paper will outline a framework that can shape these efforts, without locking agencies into a single approach, especially amid rapidly evolving legal and policy opinions about engaging with the grassroots. It is my hope that technologists, legal advocates, and senior policy makers will create a partnership to explore what can be done when collective intelligence comes together for disaster response.

John Crowley, 2012 Public Policy Scholar, Commons Lab, Wilson Center Cambridge, Massachusetts and Newtown, Connecticut

Commons Lab of the Science and Technology Innovation Program Woodrow Wilson International Center for Scholars Washington, D.C.

July 7, 2013

11 1212 Introduction

Leaders in disaster response are operations. Over the three weeks follow- finding it necessary to adapt to a ing the 2010 earthquake, 640 volunteer new reality. Although community mappers from OpenStreetMap traced actions have always been the core high-resolution satellite imagery released of the recovery process, collec- by Digital Globe and GeoEye (see figure tive action from the grassroots 1) They made 1.2 million edits to the map has changed response opera- in less time than it would have taken an tions in ways that few would have agency to solicit bids from vendors. This predicted. Using new tools that in- effort built a free and open atlas of roads terconnect over expanding mobile and critical infrastructure that is among networks, citizens can exchange the most detailed digital maps in the information via maps and social world. The U.N. used these data to build media, then mobilize thousands of its maps, and the United States soon people to collect, analyze, and act followed (albeit with some caveats). In on that information. Sometimes, addition, SMS-based initiatives1 collected, community-sourced intelligence translated, and mapped (geo-located) may be fresher and more accurate more than 90,000 text messages from than the information given to the Haitian citizens and made them available responders who provide aid. in various formats for responders and urban search and rescue teams. They also Federal agencies have noticed this broadcast more than 600,000 messages newfound capacity. Over the past two back to the affected population.2 years, a series of events caused several agencies to explore ways to engage citizen-driven efforts and fashion the citizens into new partners in agency response operations. This work began in Haiti, when—for the first time—crowdsourced datasets became the foundation on which the United Nations (U.N.) and U.S. federal agencies planned their response

13 Commons Lab | PolicY series | VOL 3

After the Deepwater Horizon Oil Spill in the spill. In 36 sorties over 7 weeks, vol- May 2010, federal agencies missed an unteers collected 11,000 photographs. opportunity. Most imagery was avail- By reference, experts can use 15cm able only at coarse resolutions (1m to imagery to identify individual species of 250m). The Public Laboratory for Open plants and animals. To the knowledge of Technology and Science (PLOTS) orga- PLOTS, none of its data were integrated nized volunteers to take 3cm resolution into the federal response operation.* balloon images of beaches affected by

Figure 2: Balloon imagery from southern Chandeleur Islands, Louisiana, collected by the Public Laboratory for Open Technology and Science and LA Bucket Brigade after the Deepwater Horizon Oil Spill, on May 9, 2010. Used under CC-BY 3.0 license. Available at http://publiclab.org/notes/ gonzoearth/5-23-2011/chandeleur-islands-louisiana.

* Jeffrey Yoo Warren, Grassroots Mapping: tools for participatory and activist cartography, MS Thesis (Massachusetts Institute of Technology, 2010): 69-74, available at http://unterbahn.com/thesis/. See also, Public Laboratory for Open Technology and Science archive page on the efforts during the Deepwater Horizon Oil Spill, http://publiclab.org/wiki/gulf-coast.

14 CONNECTING GRASSROOTS AND GOVERNMENT FOR DISASTER RESPONSE

In 2012, when Superstorm Sandy hit over mobile phones. Others engage the East Coast of the United States, experts to develop and deploy tools from 6,717 volunteers analyzed more than citizen scientists and technologists for 35,535 photographs, completing more difficult challenges in disaster response. than half of that work in 48 hours (see Each type has shown great promise figure 3). The analysis combined with for augmenting the capacity of federal several other geospatial advances, which agencies. together accelerated FEMA’s response operation by several days over previous For all the potential power that would methods. In a time when budget cuts are flow from strengthening these new accompanying increased expectations forms of mutual aid, creating an interface from American citizens around disas- between grassroots and government has ter response capabilities, this surge in raised numerous challenges. After study- capacity is difficult for federal agencies ing efforts to connect the grassroots and to ignore. government at the Departments of State, Defense, Health and Human Services, In reaction to the challenges of these Homeland Security, and Interior, the operations, several grassroots technol- major issues can be distilled to two ogy initiatives transformed into stable questions. First, under what authority can organizations that now provide services government personnel create workflows to governments during crises. Some with grassroots communities that apply analyze imagery and build maps. Some techniques like crowdsourcing, partici- develop modular structures to translate, patory mapping, and other approaches categorize, and summarize social media that build information resources from and reports from the general populace the bottom up (i.e., socially constructed

15 Commons Lab | PolicY series | VOL 3

knowledge)? Second, what controls tentially) exceed their appropriations? must be put in place within any such What happens when a federal em- workflow to ensure that the process ployee has contracting authority and itself stays within the law and the result- agrees to open-ended indemnification ing data can be reliably delivered and clauses that have become common trusted? within the terms of service for most social media (Anti-Deficiency Act)? Based on research funded by the Commons Lab at the Woodrow Wilson • How should these processes deal International Center for Scholars, the with personally identifiable information answer to the question on authority (Privacy Act) and prevent disclosure is straightforward. No law prohibits of private information (Nondisclosure government engagement with grass- Act)? roots technology organizations during disasters. Several policies encourage • How can government personnel avoid agencies to pursue community-based creating undue burdens on citizens activities—termed “whole of community” (Paperwork Reduction Act)? Can engagements—that increase national they work with OMB to seek formal resilience. For example, the U.S. Agency exceptions to the law to cope with the for International Development/ Office of exigencies of disasters and national Transition Initiatives actively fosters these crises? activities, including participatory mapping. Similarly, FEMA catalyzes these • How should agencies ensure that data approaches to major disasters under integrated adhere to the standards for Presidential Policy Directive 8. quality, objectivity, utility, and integrity that citizens expect from government The laws and policies studied for this (Information Quality Act)? report (outlined below) govern only the second question: how government Research indicates that several U.S. agencies engage with grassroots com- federal agencies have found answers to munities. The question breaks down into most of these questions and developed several practical areas: pilot projects establishing experimental interfaces with grassroots technol- • How should agencies establish work- ogy organizations. Although each pilot flows with volunteers that create no responded to a unique problem and expectation of payment and no overlap emerged in the context of the agency’s with duties already performed by gov- interpretation of law and policy, a pattern ernment personnel (Anti-Deficiency is emerging. Although the develop- Act)? ment of this grassroots to government interface happened iteratively—without • Do the terms of service of underlying a master plan pursued via linear steps— technical platforms expose agencies agency efforts generally unfolded in four to financial liabilities that could (po- phases:

16 CONNECTING GRASSROOTS AND GOVERNMENT FOR DISASTER RESPONSE

1. Design: A process to scope the legal 4. Learning/Evaluation: A process for and policy questions, define tasks, auditing the quality of data submitted and determine why, by whom, and from the crowd and integrating les- how the tasks are to be done. This sons learned. After activation, partners established the goals of the pilot, discovered that parts of their design clarified the roles of the agency and needed alterations. They also found their partners, built a worknet, includ- that they had the data to prove the Ding legal and technical advisors, and Lmerits of their approach to skepti- created the holding environment for cal audiences and began advocating exploring legal, policy, and technology for the proposed design as a policy challenges. change.

2. Experimentation: A process for The Design-Experiment-Activate-Learn the iterative exploration of options (DEAL) framework is the lens through that inform the design, workflow, which this report will analyze three ex- policy/legal framework, and choice amples of crowdsourcing: of technologies. Agencies created a shared laboratory space, which gave 1. The U.S. Department of State them and their partners a zone where Humanitarian Information Unit’s pilot Ethey tried multiple approaches and workflow with HOT and the National failed—sometimes repeatedly—in a Geospatial-Intelligence Agency to cata- safe environment. lyze remote mapping activities like those that took place during the first weeks of 3. Activation: The execution of an initial the Haiti earthquake operation3 plan, including a plan for collecting metrics on system performance. 2. An interface FEMA established To learn how to apply experimental with the Civil Air Patrol and HOT to approaches in the real world (or provide damage assessment from simulation), the partners implemented photographs collected by aircraft after their designs and gathered data about disasters4 their performance. In this way, they were able to learn and measure their 3. The process that USAID created with results without expectations of fully the SBTF and GISCorps to ‘munge’5 operational capability. Agencies were geographic data on loan guarantees in Aalso able to refine (or reconceive) of developing nations. their requirements in a pre-acquisition environment, before various obliga- Through this report, I aim to explore and tions under the Federal Acquisition provide approaches to the questions Regulations might have forced federal that commonly emerge when building an program managers to limit the amount interface between grassroots to govern- of learning and alterations to the ment. This report will also dive into the design that can happen. specific legal, policy, and technology questions that each team had to answer.

17 Commons Lab | PolicY series | VOL 3

For all the potential power that would flow from strengthening these new forms of mutual aid, creating an interface between grassroots and government has raised numerous challenges...

Report Overview research. Stakeholders were selected from three categories: Section 1 begins with context: what is crowdsourcing, and why is it useful? 1. Digital humanitarian leaders Why is it hard to mix crowdsourced data with authoritative government data? 2. Federal agency experts in crowd- Section 2 explores how grassroots enti- sourcing and policy around their ties emerged and how they function. It application to emergency operations. also explores problems with engaging the crowd. Section 3 focuses on three 3. Outside experts short cases. Section 4 examines the DEAL framework by which agencies suc- Interviewers met with stakeholders in cessfully established crowd engagement. person or talked with them via Skype or telephone for 30 to 120 minutes. A An online site will provide a space for a set of eight research questions (see growing repository of case studies. It will Appendix 1) formed the core of each be linkable from: http://CommonsLab. interview. In some cases, researchers wilsoncenter.org. convened several stakeholders from a given agency or agency partnership for Methodology a larger group conversation. In cases where the interviewees agreed, inter- The research team used a mixed views were recorded, and researchers methods approach to this study. Based then took notes from the audio. In all on requests from federal agencies and cases, the researchers took detailed the Wilson Center’s Commons Lab, notes. we identified key stakeholders across a range of digital humanitarian organiza- Research on the grassroots to govern- tions and federal agencies that work in ment interface also occurred through disaster or crisis response and devel- participatory observation at two venues. oped a set of questions to form the One such investigation took place core points of discussion for qualitative

18 CONNECTING GRASSROOTS AND GOVERNMENT FOR DISASTER RESPONSE

through an existing Joint Interagency Field Experimentation program, which is held each quarter by the Naval Postgraduate School and the National Defense University at Camp Roberts (Paso Robles, CA). This research focused on the intersection of policy, legal, and technical challenges around the collection, analysis, and application of information gathered from both unbounded and bounded crowds, where membership is limited to trained volunteers. The second occurred when the FEMA Innovation Team invited a small team of experts—including the author—to the response operation in New York to explore ways to connect federal agencies to grassroots operations, such as OccupySandy. In both cases, the author worked with the grassroots communities and the government to bring hidden issues to the surface and build frameworks to address them. The results of this research informed the research in this paper.

The research team developed this report over a period of eight months. In the process, federal agencies adapted their approaches to incorporate ever greater integration with digital humanitarian organizations, necessitating a round of edits and approvals of the final text with federal agencies so that this document (completed in July 2013) reflects current practice.

19 20 Defining the Context 1

The interconnectedness of so- During recent disasters, this social cieties is transforming disaster network has behaved like a central response. Billions of people have nervous system. Signals traversed wide direct access to cellular phones areas, sometimes giving early warning and communications networks. to those beyond the immediate impact Billions of networked devices zone. This happened with the 2011 participate in this ecosystem as Virginia earthquake. Just as light from an automated intelligent agents. exploding firework precedes the boom, Via face-to-face relationships a Twitter wave of memes7 about the with other community members, earthquake preceded the buckling of millions have indirect access to the earth’s surface. Citizens in New York messages. With this increase in and Boston knew an earthquake was scale, the cost of sending a mes- happening seconds before they felt any sage is approaching zero.6 shaking, simply by glancing at the surge of messages (see figure 4).8

Figure 4. XKCD Seismic Waves. http://xkcd.com/723/. Used under the Creative Commons 2.5 Non-Commercial-Attribution License. See http://xkcd.com/license.html.

21 Commons Lab | PolicY series | VOL 3

This phenomenon of citizens as sen- responses, and learn about the disaster. sors becomes more complicated once This phenomenon occurred at the time of response operations begin. When the Japan earthquake, Hurricane Sandy, organizations mobilize their personnel, all and (during editing of this paper) the parties endeavor to make sense of the Boston Marathon bombing. data available to determine who needs what where and who is doing what Such spikes in information velocity are where. The grassroots and govern- growing as the baseline rate of infor- ment efforts to make sense of Hurricane mation flows also increases. Federal Sandy generated more than 20 million agencies already live in an informa- tweets, several terabytes of satellite and tion superstorm. If citizens’ capacity aircraft imagery, and an incalculable to publish needs via multiple channels number of emails, SMS/text messages, increases, the familiar approach of and documents. For federal agencies, linearly scaling staff to process increase the task of aggregating, filtering, and information flows will no longer be suf- interpreting such information streams is ficient. Our capacity to process and overwhelming. As communities partici- contextualize information must scale12 pate more in response operations, the exponentially. interpretation problem is going to get bigger. Much bigger. Federal agencies in the United States are not ready for the volume and Study after study shows that the stock of increasing velocity of this flow, nor are human information is expanding expo- international humanitarian institutions. nentially.9 As Eric Schmidt of Google Collecting, processing, and analyzing noted in 2010, we now produce as information requires staff to monitor much data in 48 hours as did all of hundreds of channels, many of which human civilization from the beginning are compartmentalized by law, policy, or of written history until 2003.10 However technology into silos that may not permit fast this data creation may seem, this cross-correlation because of Privacy ‘information velocity’ is also accelerat- Act concerns or where the fusion13 of ing. The doubling rate for today’s level information may cause political conse- of data production is already less than quences. The management of such high three years, and that rate is increasing; information flows is beyond the capacity in other words, the rate of information of federal agencies—even those with production in 2016 will be more than reserve staff, such as FEMA and the double what it is today, and the numbers Department of State. in 2019 will be more than double those in 2016.11 For disaster response agen- Setting aside traditional methods,14 there cies, these numbers should ring alarm have been two innovative approaches to bells. Disasters cause spikes of data addressing this challenge: production that may be orders of magnitude above baseline levels; the rate of 1. Teaching computers to make information flow spikes as families track sense of the ways in which humans loved ones, coordinate household-level express ideas in crisis (augmenting

22 CONNECTING GRASSROOTS AND GOVERNMENT FOR DISASTER RESPONSE

the capacity of humans to discern nications or time to “close the loop” with patterns) the crowd via social media. The problem 2. Teaching humans how to work in then gets pushed into a space where new organizational designs that two or more bureaucratic hierachies enable thousands of individuals to must determine who is right. Such a con- collaborate on aggregating, pro- flict caused search and rescue teams in cessing, and analyzing large streams Haiti to clear the same building multiple of data15 times.

The first approach—machine intel- For federal agencies, another problem ligence—uses technologies where has emerged around big data: aggregat- algorithms help to discern signals from ing data about American citizens on a noise. Large U.S. agencies, includ- government-managed platform—one that ing the Department of Defense and correlates a citizen’s personal expres- the intelligence community, as well as sions with other conversations—raises academic institutions, have used this troubling privacy issues. Further, machine “big data” strategy. Their projects ingest learning does not provide an opportunity millions of messages from social media to delegate the interpretation of data to and text message services, and then citizens, nor does it allow citizens in situ extract memes, such as aid requests or a chance to offer mutual aid without the changes to the perception of a situation information flows first passing through (also called sentiment analysis).16 This many layers of government. As imple- process often uses a “human in the loop” mented by federal agencies, big data to correct and teach the machine how to efforts tend to keep the information discern patterns that are not semantically flow inside government and its contrac- obvious. For example, several university tors, inaccessible to citizens.17 Thus, teams collected social media during the second approach to information Hurricane Sandy and identified patterns processing—collective intelligence—is of communication between federal agen- taking on special currency. cies and the affected population; they then divided those dialogues into memes A collective intelligence enables volun- that summarized the general sentiment. teers and paid crowds to work on open platforms to aggregate, process, and Unfortunately, humans make this ap- analyze large pools of information (e.g., proach complicated. Language is Wikipedia). Their methods are divided contextual. We are apt to retweet mes- primarily into two categories. They either sages about people buried under rubble separate a large stock of information long after the building has been cleared. into smaller pieces that individuals can This causes a conflict between the analyze (the origin of “crowdsourcing”), big data analysts, who assert that the or they aggregate a stream of data into crowd is certain people are still buried larger stocks of information (sometimes there, and field personnel at the location, into a market or a wiki). In some cases, who know that all bodies have been they use a combination of both strate- recovered but lack the Internet commu- gies.18 The OpenStreetMap community’s

23 Commons Lab | PolicY series | VOL 3

that mobilize large user bases to provide aid. Examples of the former include Crisis Commons, HOT, and the SBTF. Examples of the latter include Google MapMaker.

During recent disasters, these organizations have mobilized collective intelligences to coordinate and channel the energy of an interconnected society. They are providing ‘information as aid’ to responders and the affected population.20 This new capacity—and the legal Figure 5: Humanitarian OpenStreetMap and policy challenges it raises—cata- Team Tasking Server, showing division of an area into grid squares that can be checked lyzed this report. out (red) and those that have been audited and cleared (green). © 2012 Humanitarian Here is the rub: federal agencies have OpenStreetMap Team. Used under CC-BY-SA an easier time working with corporations by OpenStreetMap. Source: task.hotosm.org. License is embedded in the image itself. that engage in collective intelligence than with communities that emerge from mapping of refugee camps in East the grassroots. Like agencies, corpora- Africa illustrates the fusion strategy. tions have careful internal processes to Approximately 40 individuals reviewed vet data before publication. Agencies a grid of fresh satellite imagery, dividing also have contracting mechanisms to the magnitude of the stock into smaller ensure data quality and place liabilities areas (see figure 5). Then, they used the onto the corporate contractor. However, OpenStreetMap software to add roads corporations have weak incentives and structures to each grid square, to work collaboratively; Google and creating a digital map of the camps Facebook compete both during an around Somalia by adding a multitude emergency and in between events. of small edits together to create a whole picture.19 During crises, grassroots organizations have significant advantages. They at- Collective intelligences do not emerge tract individuals with direct connections ex nihilo, without dependency on the with the affected populations. Because existing resources; they require an grassroots organizations can mobilize organizational structure that provides and train eager volunteers, they can coordination, determines methods and scale and process information at faster tools, and includes a mechanism for rates than most large corporations.21 For recruiting and mobilizing individuals. In those federal agencies, this speed is crisis response, these structures take important. Agencies must quickly define two forms: (1) communities of volunteers the scope of their response to a sud- that associate and often incorporate for den onset emergency and then decide support and (2) corporate structures how many zeroes will be assigned to

24 CONNECTING GRASSROOTS AND GOVERNMENT FOR DISASTER RESPONSE

During recent disasters, these organizations have mobilized collective intelligences to coordinate and channel the energy of an interconnected society. This new capacity—and the legal and policy challenges it raises—catalyzed this report.

their operations In this use case, rough how these groups handle data aggrega- answers on a fast timeline are often bet- tion within the federal bureaucracy. ter than exact figures that are delivered after the crisis has moved into another The next two chapters provide an initial stage. Federal agencies are not always overview of the answers uncovered from optimized to provide rough estimates a summer of research. They are inher- in hours or minutes. The rise of collec- ently incomplete and therefore placed tive intelligences outside of government into the public discourse for editing and makes it possible to quickly process expansion by the community. The follow- information and mobilize citizens to ing report is intended to be placed into provide mutual aid (with caveats added a wiki, so that practitioners can amend about potential risks of verification of the (and probably emend) the analysis and data). framework. We begin with a definition of the grassroots and how they work in For all the power that grassroots Section 2, followed by an examination organizations have demonstrated, their of three cases at the interface between interface to U.S. federal agencies is still these groups and federal agencies in underdeveloped. These groups have Section 3. experienced growing pains—some from lack of resources during the recent economic downturn, which occurred right at the time in their maturation when they were beginning to scale their operations.

Federal officials interviewed for this report raised many questions about how these groups work, what they do, and why they do it. The officials want to know how these groups make decisions and how a federal agency can invest in this capacity. They are curious about

25 26 Defining the Grassroots: Who They Are and 2 What They Do

During our research, federal of- filled by employees being paid by their ficials frequently asked, “Who respective employers to participate in a are these volunteer technical collective intelligence. Similar roles have communities, and why are they long existed in the field of open-source trying to provide services during software. These individuals take on disasters? Why should we trust leadership roles for a variety of reasons. the data?” This paper sketches Some test their tools and practices in out preliminary answers and offers crisis response, which is one of most basic frameworks on which these unforgiving environments in the world communities and future research- for privacy, security, and austere com- ers can build. munications (and therefore an ideal site for spurring innovation). Others engage The first step toward this end is to to advance efforts at corporate social abandon two terms that have come into responsibility or other forms of giving. common use: “volunteer technical com- Many leaders of volunteer organizations munity” and its variant, “volunteer and have become paid professionals who technical community.” Both have been earn part or all of their income running used to describe the grassroots commu- incorporated grassroots organizations. nities that emerged from crisis mapping activities. They appeared in 2010 after The individual volunteers who participate the Haiti crisis, when the movement was in these efforts are not all technologists, inchoate and appeared to be composed nor are the grassroots organizations of volunteers using technologies inside purely technical or limited to conducting online communities. It is now better un- their work via online/virtual communities. derstood that both terms have changed Most participants are, in fact, experts or were inaccurate to begin with. in other fields who happen to be adept at using crisis mapping tools, many Important leadership roles— of which are simple, familiar tools that “supernodes”22 in a large network of are applied within a set of increasingly contributors—are not always filled standardized practices (such as SMS/ by volunteers, but instead are often text messages and social media). These

27 Commons Lab | PolicY series | VOL 3

grassroots organizations foster practical Defining Digital approaches that focus on relationship Humanitarian building, information analysis and fu- Organizations sion, rather than software development. Several organizations now send staff The DHOs studied harness collective directly into affected communities to intelligence to provide information as aid mentor members on collecting, analyzing, before, during, and after a crisis. Each and curating community data. DHO applies a mix of social technologies using an evolving set of practices Last, these groups may mobilize com- shared by members. These practices munities, but they are now generally often involve the use of open-source incorporated entities with different software, open data, and open inter- expectations around their behavior, faces. All the DHOs have adopted or are including governance and reporting re- adopting ethical standards in codes of quirements. They are more appropriately conduct. Our working definition is: called organizations. A DHO is a grassroots organization that In late 2011, an association of these mobilizes a large number of individuals communities adopted the term digital hu- that share a set of open tools, practices, manitarians, because their tools tend to and ethical standards to create a collec- be based on ideas that emerged during tive intelligence for providing information the digital revolution.23 Today, most orga- as aid. nizations work under this moniker and an umbrella called the Digital Humanitarian An Abbreviated History of DHOs Network (DHN). However, the term digi- Most DHOs evolved from networks of tal humanitarian is also imperfect. Entities volunteers that collaborated to solve like HOT are not purely digital; they use complex problems during previous emer- a mix of handheld-and paper-based gencies. The first two—MapAction and tools when they work from the field with GISCorps—mobilized seasoned geo- affected communities. They are also not spatial professionals to provide mapping necessarily fully subscribed to humani- and analytical services to institutions tarian principles of impartiality, neutrality, learning how to leverage geographic in- and independence. These organizations formation system (GIS) tools in the early are focused on working with collective 2000s. The evolution of these networks intelligence to bring together affected defined the organizational design of communities, diasporas, and experts. subsequent DHOs. However imperfect the term, digital humanitarian organizations (DHOs) is used Based on the successes of these first here to reflect current practice. DHOs and parallel work in disciplines such as crime mapping, Patrick Meier and Jen Ziemke convened a group of practitioners, academics, and corporate representatives to explore the

28 CONNECTING GRASSROOTS AND GOVERNMENT FOR DISASTER RESPONSE

The Crisis Mappers network was the forum where international responders coordinated requirements with volunteer efforts of hundreds of organizations.

concept of crisis mapping. In October Force for Haiti, UN OCHA; and a myriad 2009, they hosted the first International of open-technology communities, includ- Conference of Crisis Mappers, which ing OpenStreetMap and Ushahidi.25 brought together geospatial experts from the United Nations, the private sector, In the following months, Crisis Mappers and the public sector, including multiple responded to a string of major emergen- governments.24 This meeting came at a cies: Chile in February–March 2010, the pivotal moment. The relationships and Deepwater Horizon oil spill in April–May ideas created a buzz throughout the late 2010, the Pakistan floods in July 2010, fall of 2009. When the Haiti earthquake and the Haiti cholera outbreak in Oct hit on January 12, 2010, these relation- 2010. As a result of the intensity and ships became a game-changing part tempo of these activations, DHOs affili- of the response operation—a dynamic ated with Crisis Mappers found that they that is explored in the report, Disaster could not function solely as an informal Response 2.0. worknet. Individuals had insufficient support from their organizations to keep When dozens of international agencies self-activating. Some overstepped their were searching for images and maps mandates or job duties, and their ongo- of Haiti, the Crisis Mappers discussion ing, non-emergency projects suffered list quickly catalyzed into a worknet—an because of the time that they committed organizational design that pools key to disaster response. Others required experts across institutional boundar- a business model to continue work ies. This network was the forum where on crisis mapping. During this period, international responders coordinated several DHOs began building formal or- requirements with volunteer efforts ganizational structures that could receive of hundreds of organizations. These funding, sign contracts, and limit their organizations included: companies like members’ liabilities.26 Digital Globe, GeoEye, Google, and ESRI; formal response organizations Today, the organizational structures of like the U.S. State Department, the U.S. DHOs reflect this evolving dynamic. Department of Defense’s Joint Task Some are communities that are co-

29 Commons Lab | PolicY series | VOL 3

ordinated by corporations (Google This paper examines each of these four MapMaker) or larger formal associations elements so agency officials understand (e.g., GISCorps is under the Urban and the basic structure of DHOs’ work, their Rural Information System Association). decision making processes, and their Others are fully incorporated indepen- values. All are essential for building dent nonprofit organizations (HOT, trust and creating an interface between Humanity Road, Crisis Commons) or UK DHOs and government. charitable organizations (MapAction). Some remain unincorporated voluntary Communities associations (SBTFs). Groups in this last class of organization retain the strongest Mobilizing collective intelligence requires link to the idea from which they emerged: an organizational design that harnesses grassroots networks.27 Although they low-cost communications to enable 30 retain a great deal of flexibility with as an coordinated action. Leaders must moti- informal association, they incur additional vate experts to devote time (leadership), risks for volunteers. As Edward Robson then mobilize them to use their skills ap- noted in 2012, the actions of one ‘bad propriately and efficiently (management). apple’ may expose volunteers to tort When a federal agency interfaces with a liabilities.28 DHO, DHO leadership and management need to be channeled to meet a given end, rather than treating it as a con- Organizational Design tracted service. As a result, interviewees emphasized the need for a statement DHOs have four elements that enable of work to establish clear expectations them to perform their work: between organizations and to give DHO leaders guidelines that they can provide 1. Communities. They have the reputa- their members. tion and skills to mobilize individuals around an approach to building collec- A growing number of DHOs have come tive intelligence. to occupy a special niche in an ecosystem. They each have a different design 2. Technologies. They use a suite of by which they coordinate their internal hardware and software tools to facili- activities. The following short descrip- tate their work. tions of the major players are derived from their own descriptions: 3. Practices. They apply their tools within best practices that provide an Humanity Road has a mission “to organizational design for delivering educate the public before, during and information as aid. after disasters on how to survive, sustain and reunite with loved ones.”31 Humanity 4. Ethics. They ensure participants in Road activates hours after a disaster, collective intelligence work within providing immediate information triage shared values. and categorization for later analysis. It has also deployed field teams to provide direct aid.

30 CONNECTING GRASSROOTS AND GOVERNMENT FOR DISASTER RESPONSE

Box 1. Conflation of Technologies, Channels, and DHOs

Research revealed an important misun- of community, technology, and best derstanding in many federal agencies: practices that DHOs have created. DHOs are often confused with the However, without grassroots mobiliza- technologies that they deploy. Some tion of collective intelligence, important DHOs are called by the name of the elements are missing from the pack- tools they use instead of by their orga- ages some companies are selling. nizational names. However, the Standby Implementations of crowdsourcing as Volunteer Task Force is not Ushahidi. a tool will lack the best practices by The Humanitarian OpenStreetMap which participants coordinate, build Team is not OpenStreetMap. These are a shared understanding, and estab- four separate organizations.29 Large lish the ethical frameworks that keep system integrators have perpetuated the values of thousands of individuals this confusion: it is far easier to sell a aligned. Without these, technologies silver-bullet technology to the govern- are only empty shells. ment than to build the combination

The Standby Volunteer Task Force and economic development.33 HOT (SBTF) endeavors “to provide live map- trains communities how to map (so- ping support to humanitarian, human called community mapping) while also right, and media organizations.”32 It organizing a worldwide community of offers a range of services organized by mappers to trace satellite imagery during specialty and managed by seasoned emergencies (remote mapping). volunteer coordinators. It has an activation protocol that can deploy in the initial GISCorps is one of the oldest DHOs, hours of a disaster and lasts up to 3 launched in 2003. It provides “short weeks into the response operation. term volunteer GIS services to underserved communities worldwide both in The Humanitarian OpenStreetMap post disaster, humanitarian relief, capac- Team (HOT) aims to coordinate the ity building,” and other services. It is an creation, production, and distribution association of approximately 2,600 GIS of free mapping resources to sup- professionals who donate their exper- port humanitarian relief efforts in many tise to the betterment of underserved places around the world. It applies “the communities. principles of open source and open data sharing towards humanitarian response

31 Commons Lab | PolicY series | VOL 3

MapAction is another of the oldest is an open opportunity for another re- DHOs in disaster response, with its searcher to dive more deeply into these first deployment in 2004. It is the only communities. nongovernmental organization (NGO) that can deploy teams of GIS experts any- Technologies where in the world in a matter of hours. MapAction is the official mapping support A growing number of platforms enable element of the United Nations OCHA. It collective intelligences. Companies or is a U.K.-based charity. nonprofit organizations generally support the development of the software, Crisis Mappers is an “international but it must be emphasized that these community of experts, practitioners, organizations support the underlying policymakers, technologists, researchers, software—not deployments for disaster journalists, scholars, hackers, and skilled response operations. Although many volunteers engaged at the intersection such organizations have donated their between humanitarian crises, technology, time for special cases, large agencies crowd-sourcing, and crisis mapping.”34 assume that disaster response is their Crisis Mappers provided the forum by mission. It is not. Most exist to build the which many other DHOs, agencies, ecosystem around the software that universities, NGOs, and corporations drives social value, a revenue model coordinated the exchange of imagery, (profit or nonprofit), or both. The plat- maps, and operational data during recent forms in common use are here divided emergencies. It is a network more than a into four categories: imagery, mapping, DHO, but a critically important coordinat- citizen reporting, and data aggregation ing mechanism. and analysis.

Several newer DHOs have emerged. The Imagery. Many volunteer organiza- Public Laboratory for Open Technology tions have access to satellite and aerial and Science is developing remote imagery, and they have developed tools sensing as a citizen science. DataKind to collect imagery using various aerial provides experts in databases and data and ground-based platforms. Given the manipulation. Geeks without Bounds challenge of making sense of so many leverages experts in convening and pixels, groups have also developed incubating technology interventions. tools to crowdsource analysis among Translators without Borders mobilizes thousands of volunteers. These include experts in translation. And Statisticians MapMill from the Public Laboratory for without Borders recruits experts in the Open Technology and Science (PLOTS application of statistical analysis to or Public Lab) and OpenAerialMap. humanitarian problems. These organizations are worth further exploration by U.S. Mapping. The development of maps agencies. They are omitted from this re- has traditionally been done by closed port no for lack of merit, but only because groups of surveyors and cartographers. there are no specific case studies around With advancements in global positioning their work with U.S. federal agencies. This system (GPS) technologies, it is now

32 CONNECTING GRASSROOTS AND GOVERNMENT FOR DISASTER RESPONSE

possible to use standard GPS units (and DHOs Are Not Social Media even some smart phones, albeit with less accuracy) to trace roads and map It is critical to separate the channels of major points of interest. There are cur- social media, the techniques of crowd- rently two platforms that dominate this sourcing, the platforms that enable those market: OpenStreetMap and Google techniques, the organizational structures MapMaker. around DHOs, and the actions that they perform. Citizen Reporting. When citizens act as sensors, they generally submit text The term social media refers to several of reports via text messages from their the various channels through which citi- cellular phones or tweets via Twitter. zens share reports from a crisis. These Increasingly, they are using applications include short text messages (Twitter, built for field data collection by open- Facebook), photos (Instagram/Facebook, source communities and corporations Flickr), videos (YouTube, Vimeo), as well (both for-profit and nonprofit). There as fusions of all three (Facebook). Social are a growing number of platforms in media can be bounded or unbounded. this area, but the most well-known are Distribution can be limited to a social Ushahidi, FrontlineSMS, and Twitter, network that the user has defined, or Fulcrum/PushPin, FormHub, Kobo distribution can be public. In the hands Toolbox, and the Open Data Kit. These of a skilled organizer, social media can technologies are the ones most often become a set of interlinked channels. conflated with the communities who A quote from an activist in Cairo from deploy them. In specific, the Standby the Arab Spring provides insight: “We Task Force is not Ushahidi; they are two use Facebook to schedule the protests, entirely different entities. However, the Twitter to coordinate, and YouTube to tell 35 communities of technologists who build the world.” the software work in close collaboration with communities that mobilize a collec- That said, social media represent only a tive intelligence. DHOs are important subset of the total channels being used users, though not the only users. in collective intelligence. Others include wikis, which allow any authorized user Data Aggregation and Analysis. to edit a common knowledgebase, as Collating data from multiple sources well as OpenStreetMap and Google into a common platform for analysis MapMaker, which are essentially wikis is fundamental to move sense making for maps. In addition, traditional Web 2.0 from a solitary activity on a desktop interactive tools and the emerging set to a collective one on the Internet. of Web 3.0 (Semantic Web) services Several applications facilitate this type like Web Mapping Service (WMS), of work, such as GeoIQ, ArcGIS Online, Resource Description Framework (RDF), Google Crisis Map, Data.gov, Amazon and other languages aim to create an Mechanical Turk, MapStory, Sahana, ecosystem of humans and automata Crowdflower, and Idabon. interacting on each other’s behalf.

33 Commons Lab | PolicY series | VOL 3

Practices location). The U.S. Geological Survey (USGS) Did You Feel It? program, Separate from the channels over which where citizens submit information information flows, crowdsourcing is the about earthquakes is one example.37 term more accurately used to describe a Another derives from the Louisiana specific set of techniques within the realm Bucket Brigade, which partnered with of collective intelligence. Many of such the MIT-affiliated Public Laboratory for techniques are detailed in the Handbook Open Technology and Science to take of Collective Intelligence from the MIT ultra-high resolution aerial photographs 36 Center for Collective Intelligence. The of the beaches around the Deepwater following are the most relevant to crisis Horizon oil spill. They not only docu- mapping: mented the initial damage, but also took periodic photographs to show the 1. Crowdsourcing: Dividing larger tasks ongoing issues and absorption of oil into smaller ones, each of which can be into the ecosystem. performed in a relatively short period of time. 2. Data Aggregation (crowdsourcing, markets, wikis): In many cases, 2. Wiki Development: Building collective information about a disaster is spread understanding through aggregation and across multiple data sources: Twitter, collation of knowledge. Facebook, Crowdmap, Google+, Flickr, YouTube, and (in several coun- 3. Markets: Aggregating information tries) proprietary or local websites. on complex problems (like pricing or Of note to federal agencies, many logistics needs) through the exchange local sites may perform their work in of data between many individuals. In this languages with non-Roman charac- context, markets need not set pricing or ter sets or low-density languages. lead to sales of items. Mobilizing volunteers to aggregate and categorize these data is a time-con- 4. Open Source: Developing common suming task that may require special understandings or code bases by the language skills. For example, the SBTF collective action of many, as with open- aggregated and categorized social source software. media reports from Libya on behalf of several U.N. agencies, allowing those These techniques are then applied to agencies to see the emerging needs actions performed by communities, which on the ground. tend to fall into four areas: 3. Data Processing (crowdsourcing): 1. Data Collection (markets and wikis): Once data have been aggregated, Citizens and their devices act as sen- they often must be transformed into sors, where submitted data may range other formats or languages so that from simple text reports to eye-level someone can make sense out of them. photos of a flood taken with a smart- Sometimes this work is done through phone (and “geotagged” with the

34 CONNECTING GRASSROOTS AND GOVERNMENT FOR DISASTER RESPONSE

traditional crowdsourcing. An organi- Ethics zation breaks a large task into small tasks that can be done by individu- DHOs exist as a form of mutual aid. als, and citizens perform those tasks, Their assistance to affected communi- re-aggregating them into a processed ties often takes the form of assistance to dataset. The Mission 4636 Initiative, the organizations with a formal mandate for example, translated Haitian Kreyol to respond to the emergency, from local text messages into English. Also, HOT fire departments to federal agencies and traced satellite imagery and turned international organizations. This focus those pixels into polygons on the map on using information to save and sustain of Haiti. lives has exposed both opportunities and risks.39 For example, humanitarians have long known that family reunification 4. Analysis (wikis, open source): Turning is difficult. Although collecting informa- processed data into information that tion about refugees can enable them to can be used to drive decision-making locate separated family members, making often requires analytical skills that that database fully open can expose the extend far beyond those of the typical vulnerable to further risks.40 This concern volunteer. These functions have in- is especially acute in areas where geno- cluded the fusion of datasets that map cide or blood feuds may have motivated social media activity over both time and a family to take the risk of migrating from place, showing the progression of a one place to another. As DHOs enter this fire line against movement of persons. area, they bring more powerful tools than This type of analysis may include qual- humanitarians have traditionally used. ity assurance work on the processed They also have mobilized a base of vol- data of other volunteers. For example, unteers from many backgrounds, most of GISCorps deployed experienced GIS whom have never been in the field. DHO professionals (average 8+ years of leaders have recruited humanitarians who experience) to analyze the accuracy provide an ethical mindset and build train- of work performed by the SBTF in ing programs that enable volunteers to processing the geodata of a USAID understand how opening information may dataset.38 also open risks. Recruiting individuals to participate in a collective intelligence and channeling Most DHOs combine specific train- them to activities that are appropriate for ing with codes of conduct to ensure their skills and interests are themselves participants are working with common competencies. Most DHOs have devel- techniques and values. These programs oped a set of recruiting processes and vary widely in rigor, but ensure that any- workflows around this challenge, usually one who helps build collective intelligence in the form of an activation protocol. As has basic knowledge of the technologies federal agencies learn how to engage and practices. They also provide forums with DHOs, they have much to learn from where mentors can correct issues and these evolving protocols. answer questions, both in real time via chat and asynchronously via discussion

35 Commons Lab | PolicY series | VOL 3

In closed systems, such as government, the fear is being in error. Making a mistake can be costly, can generate penalties, and sometimes end a career—or a life. In an open, collective intelligence—where knowledge is a social construction—the fear is burying error.

forums.41 These mentors sometimes form It is far easier to mandate data quality into teams to review participants’ work. controls for machine learning or big data This review may take place either before initiatives than for collective intelligence data are released to partners or once the projects. Agencies can issue legally data are in a wiki-like forum, such as the binding contracts for the development of OpenStreetMap database. big data tools, which can be structured around intellectual property agreements Data Quality: Why Trust that lock the methodologies, code, and DHO-Generated Data? often the data into tightly scoped work groups. Teams can be made responsible Federal agencies are legally obligated to for harnessing various inputs, process- provide data that are accurate, reliable, ing and analyzing the data, and delivering and useful. They must take steps to a complete product that answers either ensure the integrity of information. They pre-established questions in a workflow must also prevent the release of data or identifies emerging threats that need that breach the privacy or security of further analysis. As a result, the soft- citizens or organizations, violate nondis- ware and data often remain locked in closure agreements, or endanger national the system; the method and questions security. When agencies consider the use being asked may be open (but are often of DHOs either to process existing data closed). or to collect new data, they must ensure data quality throughout the process. They In contrast, collective intelligence requires must release data to DHOs in a way a far greater openness around software, that protects the privacy of citizens and data, and methodology. To harness the prevents the disclosure of protected infor- capacities of a network, large numbers mation. Likewise, federal agencies must of individuals must be trusted to view or know that the data received from DHOs edit subsets of the total dataset. In some are accurate, reliable, and useful both for cases, the release of the complete data- government decision-making and release set is highly desirable or even mandatory. to the public. An initiative for collective intelligence works best when it has transparent processes enabled by open-source soft-

36 CONNECTING GRASSROOTS AND GOVERNMENT FOR DISASTER RESPONSE

Box 2. Three Definitions in Open Technology

Open data are “a piece of data or technological community as a response content is open if anyone is free to use, to proprietary software owned by reuse, and redistribute it — subject corporations.”44 only, at most, to the requirement to at- tribute and/or share-alike.”43 Open standards/formats for data provide a free and openly available Open-source software is a piece of specification for “storing digital data, software whose “source code is avail- usually maintained by a standards able to the general public for use and/ organization, which can therefore be or modification from its original design. used and implemented by anyone. Open source code is typically cre- For example, an open format can be ated as a collaborative effort in which implementable by both proprietary and programmers improve upon the code free and open source software, using and share the changes within the com- the typical software licenses used by munity. Open source sprouted in the each.”45

ware to manipulate open data that are Early academic studies of DHOs exchanged using open standards.42 indicate a high degree of accuracy in the data that they have processed. Open systems establishes a very OpenStreetMap data can be very different view of the construction of close to professional cartography.46 knowledge than closed system. In Crowdsourced geotagging of large closed systems, such as government, datasets can be more accurate than the fear is being in error. Making a automated processing of the same mistake can be costly, can gener- information.47 The challenge of creating ate penalties, and sometimes end a an interface between the grassroots career—or a life. In an open, collective and government is more than integrating intelligence—where knowledge is a new techniques into federal workflows. social construction—the fear is burying It is about building trust around new error. There is a fear that errors will be methods of generating knowledge in the hidden in secrecy and that collective open. This trust is one of the core chal- efforts will perpetuate or even amplify lenges of developing open government. those errors. This social approach to constructing knowledge is every bit as Several federal agencies have been powerful as the authoritative method: it developing processes to build this trust. has generated Wikipedia as well as the The next section explores the framework major software that drives the Internet by which they have been performing this (including Linux and Apache). work.

37 38 Defining the Grassroots to Government Interface 3

The path can be unclear for and health threats. The 2.0 version federal agencies that wish to har- brings together a community of public ness DHOs for disaster response. health officials and other experts to The first pilot projects have only analyze real-time health data, with some begun to map the legal, policy, pulled from social media. and technology issues involved with integrating crowdsourced Although these tools do connect an information sources into official, agency to citizen sensors and convene authoritative datasets. That said, a community of trusted experts around several U.S. federal agencies have their data, creating an interface to DHOs been involved with collecting and requires a special level of policy review verifying data collected from citi- that neither a pure sensor network (e.g., zens in crisis for many years. USGS’s Did You Feel It? See Box 3) nor an interagency community of practice The best-known example may be the (e.g., BioSense) has yet required. They USGS Did You Feel It? service, which are therefore the focus of the cases in was established in 1999 to augment the this report. USGS’s understanding of the human impact of earthquakes. Over the past There are three known cases in the U.S. 14 years, more than two million citizens government that established an inter- have submitted data about the effects face to grassroots technology networks of earthquakes in the United States, as within the crisis mapping community: well as in other countries (see fig- two specific to disaster and one with a ure, pg 15). In 2010, the Centers for development focus, but with obvious ap- Disease Control and Prevention (CDC) plications to crisis response: implemented BioSense to enable the community of local, state, and federal 1. The U.S. Department of State public health officials and practitioners Humanitarian Information Unit’s pilot to identify emerging disease outbreaks workflow with HOT and the National

39 Commons Lab | TRANSFORMINGPolicY series |EARTHQUAKE VOL 3 DETECTION AND SCIENCE THROUGH CITIZEN SEISMOLOGY

Box 3. Citizen Responses to Did You Feel It? On August 23, 2011, approximately 148,000 individuals used DYFI? to describe their experience of a magnitude 5.8 earthquake affecting the Washington, D.C., area.a Additionally, because high-magnitude earthquakes are fairly rare along the East Coast, only a handful of nearby seismometers recorded the main shock of this earthquake. Thus, much of the preliminary data about the extent and level of shaking of this earthquake came from DYFI? Because similar sensor conditions exist in most of the country outside of California, DYFI? continues to offer invaluable benefits.

a. U.S. Geological Survey, “One Year Anniversary: Magnitude 5.8 Virginia Earthquake,” USGS website, 2012, accessed September 4, 2012, http://www.usgs.gov/blogs/ features/usgs_top_story/one-year-anniversary-magnitude-5-8-virginia-earthquake/.

40 CONNECTING GRASSROOTS AND GOVERNMENT FOR DISASTER RESPONSE

Geospatial-Intelligence Agency to • Developing software to ensure that catalyze remote mapping activities those who participate in remote map- like those that took place during the ping can perform five linked tasks: (1) first weeks of the Haiti earthquake users can consent to the terms of the operation48 intellectual property agreement; (2) users can check out a small section of 2. An interface FEMA established imagery for work; (3) users can check with the Civil Air Patrol and HOT to their completed work back into the provide damage assessment from system; (4) editors can check/correct photographs collected by aircraft after initial work of remote mappers; and disasters49 (5) the HIU can ingest the DHO data for inclusion in its analytical products. 3. The process that USAID created with the SBTF and GISCorps to ‘munge’50 The project iteratively uncovered geographic data on loan guarantees in issues through experiments held developing nations51 under a program called Research and Experimentation for Local and International First Responders (RELIEF) U.S. State Department/ with the National Defense University and National Geospatial- Naval Postgraduate School. This pro- Intelligence Agency gram provided the holding environment where lawyers, line managers, analysts, In late 2010, the U.S. State Department and DHOs co-designed solutions Humanitarian Information Unit (HIU) and iteratively. the National Geospatial-Intelligence Agency (NGA) explored ways in which By using the quarterly field exploration the U.S. government could use the com- cycle of RELIEF to gradually explore mercial satellite imagery that it purchases techonology and policy issues, agencies to catalyze remote mapping activities by built trust with DHO leaders and created entities like HOT. This initiative required solutions to all three challenges. During finding solutions to several legal, policy, its first activation for the famine in the and technology challenges: Horn of Africa, 40 mappers traced imagery from 10 refugee camps in Kenya • Seeing if and how intellectual property and Ethiopia, creating a dense map licenses with government vendors with three days of activation. The State could apply to the creation of derived Department officially released informa- works by DHOs. tion on the process under the moniker Imagery to the Crowd.52 • Developing a workflow for requests from DHOs so the HIU would not be overwhelmed and so NGA would not violate its status as a supporting federal agency.

41 Commons Lab | PolicY series | VOL 3

Federal Emergency first 48 hours. These initial damage as- Management Agency/Civil sessments gave FEMA’s leadership a Air Patrol rapid understanding of the areas most affected by the storm. They realized that this capability could accelerate the FEMA had a problem similar to that of the deployment of federal assets to those State Department, but could not use sat- sites.53 ellite imagery to solve it. Within the first few hours of a new emergency, FEMA’s leadership must make rapid decisions U.S. Agency for about what federal resources should be International deployed. In general, satellite imagery Development is not available fast enough, especially if there is cloud cover over the disaster The Development Credit Authority (DCA) area. In contrast, the Civil Air Patrol has of USAID provides more than $200 550 aircraft in all 50 states and has million loan guarantees to entrepreneurs. developed the ability to photograph cri- This program is a success, with only sis-affected areas quickly and to collect about $9 million in defaults paid by U.S. imagery from under the clouds. However, funds (more than offset by $10 million neither FEMA nor the Civil Air Patrol had in income from bank fees). However, a method to quickly analyze thousands of the program had only country-level images and assess the level of damage geographic data on the 117,000 bor- over a wide area. So FEMA and the Civil rowers that benefitted from this source Air Patrol partnered with HOT to build a of capital; USAID could not tell where method of crowdsourcing this analysis. in a country these loans were flowing. As a result, the agency could not tell if Working through the RELIEF experi- guarantees were being concentrated mentation process, Kate Chapman and in urban areas or if adjacent regions in Schuyler Erle from the Humanitarian different countries had aligned strategies OpenStreetMap Team developed a basic across USAID missions. The DCA also workflow and modified open-source wanted to release the data to show the software called MapMill. Originally devel- impact of their work, which would require oped by the Public Laboratory for Open releasing data about each individual Technology and Science to crowdsource loan, including the geographic data that analysis of imagery collected from low- USAID missions had collected about cost balloons, this new fork of MapMill these borrowers. for the Civil Air Patrol provided a simple interface that FEMA, the Civil Air Patrol, In the spring of 2012, the DCA worked and HOT could deploy in a disaster. with the USAID GeoCenter in an effort to determine how to make best use of this During its first activation in Hurricane extant geographic data. An analysis of Sandy, more than 6,717 volunteers the database showed that about 40,000 processed 35,535 images in a few days; of the records had only the country listed much of the work was completed in the for the borrower’s address and, thus,

42 CONNECTING GRASSROOTS AND GOVERNMENT FOR DISASTER RESPONSE

borrowers could be mapped only at the the entire queue in only 16 hours. A country level. The remainder had ad- subsequent review found that the qual- ditional data that might enable the DCA ity of the resulting dataset had 69,038 to map the borrowers at the subcountry records at 64 percent accuracy. In (i.e., Admin1) level or below. Further comparison, the crowdsourced process analysis indicated that the process would refined an additional 9,616 records at take about 15 minutes per record to 85 percent accuracy.54 USAID showed address—a time commitment of approxi- that volunteers are willing to cleanse mately 9 person-years. The process by government records at low or no cost. which USAID addressed this challenge combined the big data approach with crowdsourcing.

USAID worked with the Department of Defense to develop a process to automate the extraction of geographic data from a single column of data into additional layers of specificity. The set of tools that they built used natural language processing to match geographic data with the National Geospatial- Intelligence Agency’s GeoNames dataset. This technique enabled USAID to process an additional 66,917 records, leaving a sizable chunk of the data (about 9,600 records) to be processed by humans. Because the amount of time needed to process this many records was still not within any USAID budget of time or resources, the team turned to the SBTF and GISCorps to develop a crowdsourced process.

Over three months, a team of USAID geographers, lawyers, and managers worked with the SBTF and GISCorps to develop a workflow on data.gov—a process that required adding new data-editing tools to this government open-data platform with the vendor, Socrata. The plan was to divide about 300 volunteers into shifts over three days to process the remaining 9,600 records. In practice, 145 volunteers completed

43 44 Grassroots to Government Framework 4

In each of these three cases, 1. People: From the grassroots and the government staff engaged in government, who should be part of the informed risk taking, As a result, discussion to build a solution to the the pathway was built by walking problem? After the solution is built, it.55 Analysis after their work indi- who needs to be authorized to work at cates that the main areas of effort the interface of grassroots and govern- occurred in four stages: Design, ment, and what are they authorized to Experimentation, Activation, do? How does one establish a neutral Learning/Evaluation (DEAL). space between the grassroots and the government that can act as an • Design: A process to scope the legal incubator? Who needs to be included and policy questions and defining in this worknet, and who can build tasks, why, by whom, and how. the worknet between grassroots and government? • Experimentation: A process for the iterative exploration of options 2. Processes: What processes are that inform the design, workflow, authorized for work at the intersection policy/legal framework, and choice of of grassroots and government? What technologies. laws and policies govern these processes? When is socially constructed • Activation: The execution of an initial knowledge admissible into authorita- plan with a plan for collecting metrics tive datasets, and how should it be on system performance. sourced and labeled? What policy issues emerge around key parts of the • Learning/Evaluation: A process for federal contracting interface and re- auditing the quality of data submitted quirements of the U.S. government for from the crowd and integrating les- the storage and release of data, includ- sons learned. ing privacy and information quality?

During each stage, agencies had to deter- 3. Protocols: What standards and mine answers to three groups of questions: technologies are in use by the parties

45 Commons Lab | PolicY series | VOL 3

Federal staff have to be empowered to go to the events where digital humanitarians develop their tools, practices, and ethics.

to the grassroots and government sight, several interviewees emphasized interface? What open standards are that the primary goal of this period was available? How can the platforms be building trust between the agency and altered so that that any agreed upon the participating DHOs. To this end, it processes can be used by authorized was imperative to pursue two activities persons? How does one bound the in parallel. system, and when (and how) should it be opened? 1. Federal staff had to be empowered to A set of common issues and approaches go to the events where DHOs develop is emerging from insights garnered in all their tools, practices, and ethics. This cases that we researched. The DEAL freedom helped federal staff connect framework aggregates the experiences with DHOs and enabled DHOs to of individuals who have worked across build trust with U.S. agencies.56 the cycle of designing a project, experi- menting with one or more prototypes, 2. Once the agency and the DHO had deploying a pilot, and then evaluating the built sufficient trust to try working a entire process. Instead of narrating a his- problem together, the whole team tory, the framework outlines an approach needed to define who needed to be that an agency might take based on the involved, what the interface between lessons learned. the grassroots and the government was supposed to do, how it would Design work, and when this interface would be needed (i.e., under what circum- In the design phase, the appropriate stances the DHO would be activated). minds from the grassroots and the government need to be convened, often with People a trusted facilitator. As in any project, the team needed to decide upon the goal for In each case studied, the first part the project and establish the framework of the design phase was to build the by which the team would work. In hind- worknet that connected all participating

46 CONNECTING GRASSROOTS AND GOVERNMENT FOR DISASTER RESPONSE

organizations. This group may become they be the managers of the process quite large, but not all those in the loop in the long term. needed to be active in all aspects of the design. Interviewees emphasized • Legal Advocate(s). Both the that it was best to keep the team small champion and project managers so that work could be done efficiently, encountered areas of policy and law with experts consulted and informed that were never designed for gov- periodically. ernment integration with grassroots technology communities. Interviewees The roster generally included the follow- emphasized that at many points in the ing types of individuals: process, they needed and found a legal advocate who helped them find a • Government Champion. Typically, way to accomplish what needed to be the team had a senior leader, such as done within the law. This legal advo- an member of the Senior Executive cate was critical to the success of the Service (SES), with the authority project. Some argued that, other than and legitimacy to take informed risks. the champion, the legal advocate was This person had the political clout the most important person on the team to protect crowdsourcing initiatives and remarked that is was especially from potential adversaries within the helpful when the legal advocate and bureaucracy and the interagency the champion worked closely together. process. One of the greatest fears during the project was that the opposite of a • Project Manager(s). One or two legal advocate might appear: a lawyer federal staff or contractors who were who creates delays and tells innova- the action officers assigned to the tors what cannot be done. crowdsourcing project, often directly reporting to the champion, were • Technology Advisor(s). It was generally on the team. When there important to include individuals with were no project managers, the pro- knowledge of the data standards in cess moved far more slowly; federal use by the agency, along with the in- staff recommended that someone be formation quality criteria and platforms. named project manager in the future. These were not people whose jobs On the fastest moving process— were to keep the servers running; they USAID’s crowdsourcing work with the were instead information strategists DCA—the project managers did the who could contact specific experts for work on their personal time, which was particular technical questions. not ideal either. The recommenda- tion was to keep stable leadership so • Subject Matter Experts (SMEs). that the project managers would be Each agency needed advice from the the ones charged with developing the operational staff, their pool of vendors interface and running it through the and contractors, and academics. Learning phase. It was desirable that Although some of these interactions

47 Commons Lab | PolicY series | VOL 3

were brief, these individuals were gagement on the project. Some brought into the worknet and kept organizations are decentralized and abreast of progress. Examples cellular; they may also have roles that included remote sensing experts, can be occupied by a rotating stock contracting officers who understood of individuals. It may, therefore, be specific data licensing agreements important to know when one is inviting around the data that the DHOs were a role rather than an individual. The handling, and vendors whose govern- agencies should also ask that these ment platforms would be involved in DHO leaders remain stable through- the initiative. Vendors often viewed the out the project. In return, agencies project as a positive opportunity to should ensure stability in the agency explore the addition of new features to staff assigned to the core team. In this their products. case, trust building requires advice from those who work with such orga- • Facilitator(s). Sometimes the nature nizations (e.g., Occupy). of the agency (e.g., the National Geospatial-Intelligence Agency) • DHO Technical Advisor(s). or of the project itself required an Technology platforms used by DHOs intermediary to assist with parts of can be at the cutting edge of the field. the process. This was the case with In fact, some of the DHO experts are design and experimentation, when defining the cutting edge. Therefore, agencies needed an individual who DHO experts should be included in understood how to apply crowdsourc- the dialogue, both as architects and ing techniques and had personal visionaries. A technologist of this relationships with DHOs. The facilita- caliber can make seemingly difficult tor advised the project managers on problems disappear with a few hours whom to include in the worknet while of hacking (this has occurred at Camp focusing the team on the need to Roberts numerous times). They are preserve the inclusive values fostered often called alphageeks. within the DHO community. For experiments that required the coordination • DHO SMEs. From volunteer coordi- of federal assets and DHO personnel, nators to field staff, DHOs often have the facilitator found it helpful to have a depth of expertise that needs to be an assistant who could plan events, engaged in the design of a collabora- schedule meetings, and coordinate tion with a federal agency. Sometimes logistics. the issues are not known until they are explored as part of the design • DHO Leadership. Federal agencies process. should understand the formal and informal authority structures of DHOs Processes sufficiently well to allow a given DHO to determine who should participate Once the worknet was organized, the in the various phases of their en- team answered four major groups of questions that laid the groundwork for

48 CONNECTING GRASSROOTS AND GOVERNMENT FOR DISASTER RESPONSE

the legal, policy, and technology chal- the federal government purchased the lenges that the team confronted later imagery from the private sector. during the experimentation phase. 3. What legal issues need to be 1. What is the goal of the work? overcome? The legal issues ad- Those interviewed emphasized that dressed four laws, in addition to the clarifying the goals helped them to intellectual property requirements define the legal and policy questions of the process and agency specific that needed to be explored and to concerns: design a project that would serve the use cases of a specific set of Anti-Deficiency Act: At USAID, audiences. In two of the case studies, project managers needed to ensure the goals iterated: they started in one that crowdsourcing work did not place and evolved to another. This replace work already performed by shift demonstrated the importance of federal staff. Volunteers had to agree being flexible when integrating les- that their work did not imply that they sons learned into the design process had any employment relationship in real time. It also showed that some with the U.S. government, that they freedom is needed for federal agen- would receive no compensation for cies when exploring a new workflow. their efforts, and that they waived any and all claims against the U.S. 2. What tools exist to accomplish government around the services they this task? Agencies should conduct were providing.57 To do so, volun- a review of alternatives to ensure teers checked a box on a web form that a crowdsourcing project is an on data.gov. appropriate approach for the intended goals. As was discovered with the After the Office of Management USAID initiative, crowdsourcing ef- and Budget (OMB) memo of forts might best be done by modifying April 4, 2013 on the potential existing channels, like data.gov. In Anti-Deficiency Act implications of contrast, some federal platforms have certain social media applications, a limitations that make it far easier to new legal review is required.58 The extend the tools managed by DHOs. memo states that under a limited set Instead of creating a special (ex- of circumstances, the indemnifica- pensive) federal GIS platform in the tion clause in the terms of service of Department of State/Humanitarian social media platforms could lead an Information Unit case, for example, the agency to incur a potential financial State Department asked HOT to make liability that could (theoretically) simple changes to the existing HOT exceed the agency’s appropriation. Tasking Manager that would support Thus, if a federal agency works with an important federal requirement: that a DHO community (like the SBTF or all volunteers agree (via click-through) HOT) to process either citizen-gen- to the NEXTVIEW license under which erated data or social media during disasters, as opposed to contracting

49 Commons Lab | PolicY series | VOL 3

directly with the underlying software exemptions; it is the best place for technology platform like Ushahidi or agencies to contact to learn how OpenStreetMap, a question arises: disasters change the requirements Would it be the DHO community under the Paperwork Reduction Act. that agrees to the terms of service of the underlying technology platform Information Quality Act. in the course of their work with a Agencies must ensure that informa- federal agency or the federal agency tion released from DHO projects itself? Would this avenue or a struc- adheres to the agency’s standards ture like the Digital Humanitarian under the Information Quality Network for the United Nations Act.60 In the examples, the agency make it possible to avoid potential conducted an audit of the data Anti-Deficiency Act liabilities for before release or, initially, prioritized federal agencies? speed and utility over accuracy. Establishing basic metrics and data Privacy/Nondisclosure. Data collection procedures, as well as a sources that might trigger the process for monitoring and evaluat- Privacy Act require special han- ing the project, helped meet the dling. Sometimes certain personally requirements under this act. identifiable information must be removed before a grassroots organi- Intellectual Property Licensing. zation can be allowed to handle the The project managers had to ensure data. The legal advisor and project that data released or collected had managers from the U.S. govern- appropriate intellectual property ment must determine when removal rights. To link the effort to open of data is necessary, balancing government, they focused on making privacy concerns with efficiency and the data as open as possible. This effectiveness. openness was a core value shared with DHOs and one of the motivat- Paperwork Reduction Act. ing factors for DHOs to partner Although none of the three cases with federal agencies. It also gave explored the obligations under agencies a success under the Open the Paperwork Reduction Act to Government Initiative. measure the information collection burden, there is a small chance 4. What political concerns need to that a disaster might trigger the be addressed? It is imperative to Paperwork Reduction Act. During confront the internal politics at the the Superstorm Sandy response, participating agencies. Discussions FEMA received orders to first cut with or about DHOs sometimes be- red tape and then “take a blowtorch came quite heated and did not always to the red tape.”59 The OMB Office yield clear answers. The champion, for Information and Regulatory project managers, and facilitator must Affairs is the point of contact keep the group moving even in the within the federal government for face of uncertainty. Interviewees noted

50 CONNECTING GRASSROOTS AND GOVERNMENT FOR DISASTER RESPONSE

This openness was a core value shared with DHOs and one of the motivating factors for DHOs to partner with federal agencies.

that it was better to enter experimen- shaped to conform to project goals, tation with unanswered questions than legal channels, and political realities. to get stuck in a series of conference However, it was necessary to map out calls on hypothetical risks. the way in which these tools integrated with federal enterprise systems. The real 5. How do DHOs develop trust with issues were often discovered only during federal agencies, some of which the experimentation phase, but in each operate in ways quite foreign case, the process of mapping out how to the volunteers? In addition, an data would flow in and out of the agency intangible, but important, process started in the design process. This work under any crowdsourcing initiative is incorporated two principles (open data convincing internal skeptics that this and open-source technology) and an alternative model can create data emerging method (open workflows). that meet government standards for timeliness, quality, objectivity, util- Open Data Standards. Engagement ity, and integrity. Understanding their with the grassroots usually generates an criticisms allowed supporters to plan economic public good, where “individu- data collections to mollify skeptics als cannot be effectively excluded from or convince them that the use of use (non-exclusivity) and where use by DHOs is a valid method of generat- one individual does not reduce avail- ing authoritative data. One important ability to others (non-rivalry).”61 A good concern raised was the sustainability example is Wikipedia itself, where an of DHOs. The plan for the pilot project open license on the content ensures must include instructions for hand- non-exclusive use and where global ing off the work to the DHOs within a availability on the Web ensures non- sustainable framework. rivalrous use. Because each individual contribution to a public good adds value Protocols to the good, each contribution can be used and augmented by many actors to The open technology used by DHOs bootstrap efforts at the community level. tends to be very malleable: it can be

51 Commons Lab | PolicY series | VOL 3

In contrast, some government data stan- Experimentation dards are controlled by an ecosystem Experimentation offers a way to iterate of vendors, whose platforms may not upoon explorations that are only begun support open data standards (though in the design phase. One critical aspect the new U.S. Open Data Policy should of experimentation was to expose legal change this situation). Establishing open advocates, operational staff, and skep- data standards—especially by incorpo- tics to the challenges inherent in the rating those from international standard implementation of the idea. In this way, bodies or associations—can create a they were able to offer criticisms and forcing function to help establish data insights that technologists and DHOs and transparency as a public good. could address before the tools were Agencies should strongly consider using deployed. this lever as part of engagement with the grassroots; it may be a non-negotiable A key lesson learned from previous item with many DHOs. Given the new projects was that it is important to fail U,S. Open Data Policy, the development early and often (within an environment of criteria will take on additional impor- of shared good will). Failure in this tance and urgency.62 context is an opportunity for both sides to discover stumbling blocks, misunder- Open-Source Technology Platforms. standings, or incompatibility before an Open-source software allows every- activation. This kind of space is rare in one to see the methods by which data a federal bureaucracy, where failure is are collected, analyzed, and visualized. usually penalized. However, it is impera- Opening these methodologies to all is tive for the champion to establish that an important aspect not only of many the purpose of experimentation is to fail DHO activities, but also of the interface early and often, and requires making a between grassroots and government.63 series of “new mistakes.” It is best to discover and fix these problems before Open Workflows. Designing work- an operation forces the activation of a flows between multiple organizations DHO for an actual crisis. This is the time requires transparency. In the process when the team discovers, in Edison’s fa- of tracing the data flows, the entire mous phrase, 9,000 ways to not make a team—those who function in the legal, lightbulb.64 Experimentation was the key technical, policy, and operations areas— time to build trust between individuals, encounter issues about which others bridging the government and grassroots. may have only tangential awareness. The more open the process of development, People the more likely that design will raise issues that can be explored and solved The experimentation phase involved during experimentation. a small group of innovators who can bridge the various challenges. These innovators included the facilitator, project manager(s), legal advocate, technology

52 CONNECTING GRASSROOTS AND GOVERNMENT FOR DISASTER RESPONSE

advisors, SMEs, and DHO representa- and resilient. Although it may not be the tives. For two of the cases, the inclusion criterion for government applications in of lawyers in field experiments gave the non-emergency contexts, amid the chaos DHOs and government technology advi- of disasters, it is essential to have simple, sors the opportunity to take questions lightweight, minimal processes. out of the hypothetical and demonstrate the use of specific technologies to Some experiments can be performed in solve specific problems. This specific- an office or in virtual space. Interviewees, ity enabled the legal advocates to use however, indicated that face-to-face case law and policy against a particular interactions in a safe space—removed set of problems. The author person- from the agency office—are an effective ally witnessed lawyers co-designing and worthwhile expense. One issue that workflows with technologists from both emerged during the budget crises was DHOs and federal agencies, breaking a strong desire to measure the effec- through weeks of hypothetical questions tiveness of experimentation (managers in a matter of hours. asking, “What are we getting for our money?”). The tendency to impose Processes formal structures on experimentation so that specific goals are achieved or the Experimentation is an iterative process, experiment “fails.” While this approach is not a linear one. The worknet may enter understandable, the experience of Camp this phase and find that multiple itera- Roberts/RELIEF experiments points to tions are necessary to work through the effectiveness of semi-structured issues before the project is ready to be experimentation in a remote location— activated during an emergency. This where the design of the work can be learning loop is a feature, not a bug. It is altered midstream to meet emerging an opportunity for all parties to explore understandings, and where failure is a new ways of working—the core of inno- significant finding that provides feed- vation. The results should be measured back and enables partners to seek new not only by the project’s success, but pathways. also by the number of minds that were opened to new ways of approaching Experimentation usually created reports crisis information sharing challenges. about what each side learned during each event. That said, the most important Minimal, Lightweight Workflows. deliverables tended to be the ideas that DHOs are often intensely focused worked their way from bar napkins to for- on keeping things simple. They must: mal documents and projects. One such they have to work with a wide range of sketch appears in figure 7 with digital volunteers, sometimes across lan- annotations over the original markings. guages. Simplicity is a design feature for government, as it forces a focus on the essential instead of on the creation of expensive tools to cover all possible scenarios. For disasters, simple is fast

53 Commons Lab | PolicY series | VOL 3

Statement of Work. Interviewees process data, project teams were able were clear: creating a clear statement to move from general agreements into of work is a necessity for success. specific courses of action. Over time, They emphasized that the statement the teams added more complexity. In of work must be flexible. It cannot be turn, these iterative statements of work a boilerplate document or inflexible enabled the DHOs to build their tools to contract. Rather, it must be a malleable, federal requirements. adaptable statement of expectations created collaboratively and iteratively. Its Volunteer Management Plan. By generation (and redefinition over time) co-designing the workflow, the DHOs allows agencies and DHOs to negoti- and agency representatives were able to ate (and renegotiate) their approaches, develop plans to mobilize, motivate, co- make their technologies interoperable, ordinate, and manage volunteers around establish data standards, and create the experimentation. The co-design pro- evaluation metrics and processes. cess aided in the development of a plan to coordinate and manage the activation Interviewees also recommended that of volunteers during an emergency. The projects should start small. By discuss- specificity of these plans enabled DHOs ing how to build a minimally viable to recruit SMEs, who led teams around example of how a collective intelligence specific issues raised by the federal would use a specified set of tools to government, including data validation,

54 CONNECTING GRASSROOTS AND GOVERNMENT FOR DISASTER RESPONSE

the handling of personally identifiable the costs incurred by transferring information, and the licensing of intel- data over a satellite terminal to a field lectual property. The process of keeping team. Technical interoperability is often volunteers motivated under an increas- achieved only by failing repeatedly as ing tempo and duration of deployments teams discover that underlying as- will require additional study. sumptions about networks and other infrastructure are incorrect. Protocols • Testing User Experience (UX). Experimentation will lead to exploration The human factors in crowdsourcing of the interaction of government and can make or break the initiative. The DHO technologies. This phase explores UX design expectations of volunteers the quirks of making open-source is quite high. To keep volunteers software, open data, and open methods engaged, each task should be simple work together. This review of protocols and easy to perform, which requires should include: good design. Interviewees pointed to a problem: the UX design of government • Testing Open Data Standards. web applications tends to be quite The specification and testing of open poor. As a result, work with DHOs data standards might expose dif- creates a disruptive effect among fed- ferences in implementation or the eral vendors and staff, pushing them need for adaptors between different to think more deeply about the user open data standards. For example, experience. This effect can be used as the OpenStreetMap, Google, and a positive lever for change. ESRI data models are each compli- ant with open standards, but each has a slightly different way to draw Activation a polygon. As a result, it is necessary to design adaptors to handle The activation phase gives agencies polygons as data move between and DHOs an opportunity to show the OpenStreetMap, Google, and ESRI results of all previous planning around software. legal, policy, and technical issues. In two of the cases, the projects were activated • Testing the Flow of Data across only after the teams had tested the tools Technology Platforms. The data repeatedly and resolved major issues. flow is generally easy to describe on In the case of the use of MapMill during a whiteboard. It may all be structured Superstorm Sandy, however, the code using open standards, and pathways saw its first use during an actual emer- between government and grassroots gency. Although the workflow around systems may be clear. However, it is MapMill worked very well (and the code only by experimentation that the team remained stable), this was a matter of will discover the file limit placed on a luck and credit to the skill of the develop- specific server by government infor- ers at the Public Laboratory for Open mation technology staff or discover Technology and HOT, not prior plan-

55 Commons Lab | PolicY series | VOL 3

ning. Managers at federal agencies will standards for getting approval and generally not allow such a leap of faith. sustainment. To our knowledge, no col- Given the scale of Sandy, the dire need lective intelligence initiative has reached for imagery analysis, and the fact that the level of becoming a standing policy. internal independent analysis was going All are still “pilots.”As a result, they are to occur in parallel, the cost–benefit still free to innovate and adapt during the ratio pointed strongly toward the deploy- emergency—a privilege (and opportu- ment of the new tool. With MapMill, nity) rarely afforded to mature enterprise FEMA demonstrated that it is possible systems. to manage the risks of using information processed from the crowd. Choose non-critical missions. For many agencies, engaging in a new form of People knowledge creation in a mission-critical activity is frightening. For the initial An activation generally includes the activation, it is often best to choose people who have been identified in the topics, regions, or datasets that allow workflow designs that the team built the worknet to explore and learn without in the previous two phases. Because triggering anti-body reactions within an innovative deployment will attract the federal bureaucracy. This approach attention when it works well, federal (often called a shadow operation) gives project managers need to ensure that the pilot a chance to prove its worth in they include those who are necessary an environment that permits the team to without overburdening the worknet. It take risks and make changes on the fly. is prudent to keep the activation simple and minimal, confining the list of those in Protocols leadership posts to people who are essential to the activation. Crowds can be The technologies and data standards somewhat broader, but should adhere should have been established in the to the policies that the team designed, design phase and then honed during unless circumstances (e.g., Superstorm the experimentation phase. They must Sandy) call for wider participation. remain malleable during the activation phase. Although science requires hold- Processes ing variables constant in order to learn, the mission during an emergency is not For most federal agencies, the establish- to remain loyal to the scientific process; ment of a standard operating procedure it is to preserve lives, health, and prop- or new policy around crowdsourcing erty. The worknet must remain adaptable is a complicated process that engages and may elect to adapt its protocols to a wide range of actors. Some of these meet unexpected situations or emerging participants may not yet be ready for needs. In field discourse, this is some- a deep commitment to a new form of times called “semper Gumby—always knowledge creation. As a result, most flexible.” If the DHO pilot occurs during activations of DHOs are character- an emergency, in no case should the ized as “pilots,” which must meet lower plan for collecting performance metrics

56 CONNECTING GRASSROOTS AND GOVERNMENT FOR DISASTER RESPONSE

get in the way of operations or provision framework to draw specific individuals of aid to affected communities. into official internal government reviews. If nondisclosure agreements are Learning required, they should be in place before activation. Qualitative and quantitative evaluation of the project is important for adapting the design to the actual requirements of Protocols activations. Open data and open-source projects are just beginning to establish frameworks People for monitoring and evaluation. To date, there are no off-the-shelf frameworks for The process of evaluating and learning evaluating the performance of DHOs. from a pilot activation should include The DHO community expects that all not only those in the worknet from the learning will be published in an open design, experimentation, and activation forum, such as a wiki, where collective phases, but also SMEs who can assist intelligence can add questions and layers in the evaluation of collective intelli- of meaning to the findings. No govern- gences. Such SMEs are found within the ment project has yet used this format. DHO realm, as well as in academia and However, this resource should be a core commercial crowdsourcing companies. principle of open government. Parties Many people specialize in the analysis must share what partnerships learn, and of DHO operations using a range of future work should publish as much as methodologies. Most experts mix qualita- is feasible. In this way, other agencies tive research with quantitative analysis and DHOs will not need to rediscover of the data that a project creates during lessons learned. an emergency (e.g., mixed methods approaches).

Processes

Open Evaluations. Evaluation and learning should be in an open forum, so government and grassroots can learn from each other. Internal government reviews are often marked with the caveats For Official Use Only (FOUO) or Sensitive But Unclassified (SBU). While it may be necessary to address some internal arguments over data quality and ensure that certain internal assessments remain confidential, this approach slows down collective learning. Openness should be written into the statement of work with the DHO so that there is a

57 58 Conclusion 5

“When Thomas Kuhn defined outlining a framework for this connection. paradigm change in The Structure The task is now on the community—the of Scientific Revolutions, he agency champions and DHO leaders—to described a state where a tra- fill out this initial skeleton. ditional framework and several experimental approaches existed The key to the successful use of a col- in parallel—a period when the ex- lective intelligence will be generating planatory power of the old system trust in the knowledge it creates—not wanes while some inchoate new just inside the government, but within system explores and codifies the the populations that may be affected by methods that are strong enough to future disasters. When FEMA or USAID begin replacing the old ones.” 65 uses citizen-generated knowledge to make decisions around saving and I wrote those words two years ago, for sustaining life, citizens must trust that the the conclusion of a World Bank report data used to generate those decisions about the performance of volunteer were the best available at the time. The technology communities in Haiti. Despite process of deciding when to use collec- these technological determinism that tive intelligence to augment traditional can be found in the discourse of DHOs, mechanisms of sense making will medi- DHOs will never fully replace traditional ate how this trust is built. institutions for disaster response. That said, they are accelerating sense making Collective intelligence is a form of lead- during emergency operations—a role ership: it requires asking our citizens to that will become ever more critical as participate in responses as a ‘whole of information flows scale at an exponen- nation’ activity. Despite the many reports tial rate. Building an interface between of apathy and disengagement with grassroots and government will be a domestic and international operations, challenge. However, agencies have al- throughout my experiences. I have seen ready begun to confront the legal, policy, that this sense of service is alive and and technology issues that DHOs have well. That said, we need leaders inside of raised. This report is only the first step in government to harness collective action

59 Commons Lab | PolicY series | VOL 3

to speed our ability to determine at least two questions: who needs what where and who is doing what where. Call it the 21st-century bucket brigade: the information brigade.

I wish to state—emphatically— that this document is only an initial version of an interface—in technical lingo, an ‘alpha’ version. I encourage those with expertise and experience to fill out the framework, emend its errors, and amend its scope. Like patches in an open- source software project, your edits will be merged into the main body of the work through an editing process. Not all will make the cut, but the objective is not to integrate every edit; rather, it is to create a cohe- sive, relevant, and accurate framework that others can use and build upon—and come to trust.

60 CONNECTING GRASSROOTS AND GOVERNMENT FOR DISASTER RESPONSE

Appendix

Research Questions

The following 8 questions formed the core of research interviews:

1. How has your agency approached the 7. How is your agency working with the issue of engaging citizens in the col- decision making processes of grass- lection or analysis of US government roots organizations? What challenges data? Could you give some examples and lessons emerged? or case studies? 8. How is your agency tracking or 2. What policy or legal challenges measuring progress on these early emerged as your agency explored efforts? working with grassroots organizations like the Standby Volunteer Task Force, Humanitarian

3. What approaches did the agency choose towards those policy or legal challenges? How do these solutions work?

4. What decisions can be made using data quality?

5. What privacy issues have been raised by working with citizen-generated or citizen-analyzed data?

6. What procurement issues had to be overcome?

61 62 Endnotes

1 Short message services (SMS) is also how-to-make-crowdsourcing-disaster- called text messaging or texting. SMS relief-work-better. See also Jennifer enables a mobile phone subscriber to Chan, John Crowley, Shoreh Elhami, send text messages approximately 140 Schuyler Erle, Robert Munro, and Tyler characters in length, sometimes with Schnoebelen, Aerial Damage Assess- attached photographs, video, or other ment Following Hurricane Sandy (Idibon, media. GIS Corps, Harvard Humanitarian Initia- tive and Humanitarian OpenStreetMap, 2 See Robert Munro, “Crowdsourcing and 2013), available at http://idibon.com/ the Crisis-Affected Community: Lessons wp-content/uploads/2013/05/Sandy_ Learned and Looking forward from After_Action_Report.pdf. Mission 4636,” Journal of Information Retrieval 16, no. 2 (2013):210–266. 5 Data munging is “loosely the process See also “InSTEDD’s response in of manually converting or mapping Haiti,” ICT for Peacebuilding (ICT- data from one ‘raw’ form into another 4Peace), at http://ict4peace.wordpress. format that allows for more convenient com/2010/05/11/instedds-response- consumption of the data with the help of in-haiti/ (last accessed June 1,2013). semi-automated tools.” See “Data Wran- gling,” Wikipedia, available at http:// 3 K. Walker, “Disaster Response Efforts en.wikipedia.org/wiki/Data_wrangling. Highlight Value of Relationships, Non- traditional Partners,” Pathfinder 10, no. 4 6 That said, the household costs of (2012):18–20. See also Humanitarian communication as a percentage of Information Unit, “Imagery to the Crowd,” household budgets is a new load on the Department of State website, available vulnerable. at https://hiu.state.gov/ittc/ittc.aspx. 7 A meme is an idea or concept that 4 Jennifer Chan, “How to Make spreads from human to human using Crowdsourcing Disaster Relief mechanisms analogous to the spread Work Better,” U.S. News & World of information in genes. See Francis Report (November 23, 2012), avail- Heylighen, Evolution of Memes on able at http://www.usnews.com/ the Network: From Chain-Letters to opinion/articles/2012/11/23/ the Global Brain, available at http:// pespmc1.vub.ac.be/Papers/Memesis.

63 Commons Lab | PolicY series | VOL 3

html. See also, Oxford English Dic- who are often professional knowledge tionary, meme, available at http:// managers. That kind of performance oxforddictionaries.com/us/definition/ metric appears to be absent in industry. american_english/meme. There is too much to know.

8 XKCD, http://xkcd.com/723/. Used 15 A third approach—a hybrid of the two— with permission under the Creative is now in experimentation, but has never Commons 2.5 Non-Commercial-Attri- been applied to an emergency. bution License. See http://xkcd.com/ 16 Good examples include the social license.html. media tools used in the American Red 9 See How Much Information? 2003. Cross Disaster Operation Center, as well Available at http://www2.sims. as the Information Volume and Velocity berkeley.edu/research/projects/ joint capability technology demonstra- how-much-info-2003/. tion from the Department of Defense.

10 Eric Schmidt, Techonomy 2010 Confer- 17 There are other ways to run a big data ence remarks. See http://techcrunch. project—ways in which citizens are com/2010/08/04/schmidt-data/. integrated into the analysis. Patrick Meier is building such a system at the 11 Ibid. See also What’s the Big Qatar Computing Research Institute Data? at http://whatsthebigdata. (personal email, May 23, 2013). The com/2012/06/06/a-very-short-history- Edward Snowden National Security of-big-data/. Agency scandal in May 2013 revealed 12 The verb scale here refers to the scal- additional complications around the use ability of a system—its ability to grow of big data without citizen voice in the to match the demands placed upon it oversight of a program apparently run by without needing to change the underly- a private contractor (in this case, Booz ing architecture. See Wikipedia, http:// Allen Hamilton). en.wikipedia.org/wiki/Scalability. 18 See Boethius, The Consolation of 13 Information fusion is the “combina- Philosophy. This sixth-century text es- tion of information into a new set of tablished the seven liberal arts, including information towards reducing un- arithmetic, which was then concerned certainty.” Several federal agencies with the aggregation of small multitudes operate information fusion centers. See into larger structures (induction), and Wikipedia, http://en.wikipedia.org/wiki/ geometry, which was then concerned Information_integration. with ratios and the division of larger con-

14 If a traditional approach (such as cepts into small structures (deduction). a SharePoint) portal can withstand This division is one of the earliest codi- exponential rates of increase to total fications of the pursuit of knowledge information resources in a problem through the concept of magnitude and domain while scaling knowledge multitude—core concepts at the root of management staff linearly, it will need the development of natural science in to show year over year exponential the 12th century and, from there, of the increases to productivity of its curators, Scientific Revolution.

64 CONNECTING GRASSROOTS AND GOVERNMENT FOR DISASTER RESPONSE

19 For more information on Open- technical communities (personal email, StreetMap and refugee camps, see May 19, 2013). Humanitarian Information Unit, “Imagery 24 See International Conference of Crisis to the Crowd,” Department of State Mappers, available at http://crisismap- website, https://hiu.state.gov/ittc/ittc. pers.net/. aspx. 25 OpenStreetMap (http://openstreetmap. 20 One nongovernmental organization org) is a community of more than 1 now has a campaign called “Informa- million volunteers who are building a tion is Aid.” See Internews’ work with on free and open map of the world, using a Humanitarian Communications, available model analogous to Wikipedia. Ushahidi at http://www.internews.org/global- (http://ushahidi.org) is a platform for issues/humanitarian-information, as well citizen reporting based primarily on mes- as the now closed Information as Aid sages submitted by SMS/text services project, available at http://infoasaid.org. via mobile/cellular phones. 21 Small crowdsourcing companies that 26 See Edward S. Robson, Responding fuse the two concepts, such as Idabon to Liability: Evaluating and Reduc- and Crowdflower, which have their own ing Tort Liability for Digital Volunteers models for scaling quickly, are possible (Washington, DC: Commons Lab of the exceptions. Woodrow Wilson International Center 22 In social network theory, supernodes for Scholars, 2012), available at http:// have an extraordinary number of con- www.wilsoncenter.org/publication/ nections within (and often between) responding-to-liability-evaluating-and-re- networks of individuals. They function as ducing-tort-liability-for-digital-volunteers. influencers and amplifiers. In collective The surprising finding is that Good intelligences, their ability to be at the Samaritan laws in the United States do intersection of so many communica- not protect volunteers engaged in crisis tion pathways enable them to have the mapping activities within the United best understanding of who is doing States; in fact, current law may open what and therefore they often become them to liabilities in multiple jurisdictions. coordinators—and thereby leaders—in 27 An emerging class of DHOs now self-organizing networks. See Malcolm requires study: the FlashNGO. First seen Gladwell, The Tipping Point: How Little in Haiti with the Mission 4636 Initia- Things Can Make a Big Difference. Little, tive (discussed later in this report), a Brown and Company (2000). FlashNGO emerges to serve a specific 23 Andrej Verity, Office for the Coordina- purpose during a single response operation of Humanitarian Affairs, indicates tion and then disbands. This structure that the term digital humanitarian dates was not specifically examined by Mr. from the November 2011 meeting of a Robson’s report, but the liabilities an un- Community of Interest around the ap- incorporated association appear to hold plication of the SBTF model to provide true for it. The operation during Hur- an interface between the United Nations ricane Sandy saw the rise of numerous and a growing number of volunteer FlashNGOs, so many that they are now

65 Commons Lab | PolicY series | VOL 3

on FEMA’s radar as an emerging policy fairs.com/articles/67038/clay-shirky/ issue. At the time of this report, creat- the-political-power-of-social-media. ing a workflow with these FlashNGOs 36 Handbook of Collective Intelligence, remains an ad hoc art form. Given the available at http://scripts.mit.edu/~cci/ importance of these FlashNGOs to the HCI/index.php?title=Main_Page. The delivery of aid during Hurricane Sandy, it version on SocialText from 2008 is the may be necessary to perform additional version referenced. study on this issue in a domestic legal context. 37 Jason C. Young, David J. Wald, Paul S. Earle, and Lea A. Shanley, Trans- 28 Robson, Responding to Liability. forming Earthquake Detection and 29 Ushahidi is a U.S.-based 501(c)3 non- Science Through Citizen Seismology profit organization that supports the (Washington, DC: Commons Lab of the development of a software platform for Woodrow Wilson International Center citizen reporting. The OpenStreetMap for Scholars, 2013), available at http:// Foundation is a United Kingdom–based www.wilsoncenter.org/publication/ charity that supports the development transforming-earthquake-detection-and- of the software and community behind science-through-citizen-seismology. OpenStreetMap.org. 38 Shadrock Roberts, Stephanie Grosser, 30 See Thomas Malone, The Future of and D. Ben Swartley, Crowdsourcing to Work: How the New Order of Business Geocode Development Credit Authority Will Shape Your Organization, Your Man- Data: A Case Study (Washington, DC: agement Style, and Your Life. Harvard USAID, 2013), 17. Hereafter cited as Business Press Books, 2004. USAID Crowdsourcing Study.

31 “Humanity Road’s Page,” Digital Hu- 39 Louise Searle and Phoebe manitarian Network website at http:// Wynn-Pope, Crisis Mapping, Humani- digitalhumanitarians.com/profile/Hu- tarian Principles and the Application manityRoad, last accessed January 8, of Protection Standards: A Dialogue 2012. between Crisis Mappers and Op- erational Humanitarian Agencies, 32 “Standby Task Force’s Page,” Digital available at https://docs.google.com/ Humanitarian Network website at ument/d/1N25XfSXHcR5ZmkEegSW- http://digitalhumanitarians.com/profile/ 1lao0i8r3GKyqpsgKCafU7A/edit. Cf. StandbyVolunteerTaskForce. also Emmanuel Letouzé,; Patrick Meier, 33 Humanitarian OpenStreetMap Team’s and Patrick Vinck, New Oil & Old Fires: Page,” at http://hot.openstreetmap.org/. Reflections on Big Data for Conflict Prevention 34 “International Conference of Crisis (New York: International Mappers’s Page” at http://crisismap- Peace Institute, 2013), available at pers.net/. http://irevolution.net/2013/04/10/ big-data-conflict-prevention/, as well as 35 See Clay Shirky, ‘The Political Power George Chamales, Towards Trustwor- of Social Media,’ Foreign Affairs (Jan/ thy Social Media and Crowdsourcing Feb 2011), at http://www.foreignaf- (Washington, DC: Commons Lab of the

66 CONNECTING GRASSROOTS AND GOVERNMENT FOR DISASTER RESPONSE

Woodrow Wilson International Center 46 Morechai Haklay, “How good is for Scholars, 2013), available at http:// volunteered geographical information? www.wilsoncenter.org/publication/ A comparative study of OpenStreetMap towards-trustworthy-social-media-and- and Ordnance Survey datasets,” Environ- crowdsourcing. ment and Planning B: Planning and Design 37, no.4 (2010):682–703, avail- 40 Joel R. Reidenberg, Robert Gellman, able at http:www.envplan.com/abstract. Jamela Debelak, Adam Elewa, and cgi?id=b35097. See also Trias Aditya, Nancy Liu, Privacy and Missing Persons Nurrohmat Widjajanti, Dany Laksono, after Natural Disasters (Washington, S.T., Rindi Kurniati, S.T., Maratun Solihah, DC, and New York: Commons Lab S.T., Siti Purwanti, S.T., Burhan Sidqi, of the Woodrow Wilson International S.T. (Dompu), Adisty Paramitasari, S.T. Center for Scholars and Fordham Law (Yogyakarta), Mulya Sina Diputra (Yog- School, 2013), available at http:// yakarta), Evaluation of OpenStreetMap www.wilsoncenter.org/publication/ Data in Indonesia, A Final Report (2012). privacy-and-missing-persons-after- natural-disasters. 47 Roberts, Grosser, Swartley, USAID Crowdsourcing Case Study. 41 Federal officials often lack access to these real-time exchanges because 48 K. Walker, “Disaster Response Efforts chat is blocked at the federal firewalls. Highlight Value of Relationships, Non- This creates liabilities for crisis response traditional Partners,” Pathfinder 10, no. 4 operations, as federal officials face de- (2012):18–20. See also Humanitarian lays in correcting the actions of DHOs Information Unit, “Imagery to the Crowd,” and federal staff need to go offsite to Department of State website, available work with DHOs, separating them from at https://hiu.state.gov/ittc/ittc.aspx. internal resources. 49 Jennifer Chan, “How to Make 42 The Whitehouse /Developers blog on Crowdsourcing Disaster Relief open data and open source software, Work Better,” U.S. News & World available at http://www.whitehouse.gov/ Report (November 23, 2012), avail- developers. able at http://www.usnews.com/ opinion/articles/2012/11/23/ 43 Open Definition, available at http:// how-to-make-crowdsourcing-disaster- opendefinition.org/. relief-work-better. See also Jennifer 44 Open Source, Wikipedia, available Chan, John Crowley, Shoreh Elhami, at http://en.wikipedia.org/wiki/ Schuyler Erle, Robert Munro, and Tyler Open_source. Schnoebelen, Aerial Damage Assess- ment Following Hurricane Sandy (Idibon, 45 Open Format, Wikipedia, available GIS Corps, Harvard Humanitarian Initia- at http://en.wikipedia.org/wiki/ tive and Humanitarian OpenStreetMap, Open_format and Open Standard, avail- 2013), available at http://idibon.com/ able at http://en.wikipedia.org/wiki/ wp-content/uploads/2013/05/Sandy_ Open_standard. After_Action_Report.pdf.

67 Commons Lab | PolicY series | VOL 3

50 Data munging is “loosely the process 60 The White House, Agency Informa- of manually converting or mapping tion Quality Guidelines, available at data from one ‘raw’ form into another http://www.whitehouse.gov/omb/ format that allows for more convenient inforeg_agency_info_quality_links. consumption of the data with the help of 61 Wikipedia, http://en.wikipedia.org/wiki/ semi-automated tools.” See “Data Wran- Public_good, last accessed June 23, gling,” Wikipedia, available at http:// 2013. en.wikipedia.org/wiki/Data_wrangling. 62 Executive Order on the U.S. Open 51 USAID Crowdsourcing Case Study, Data Policy, available at http://www. cited above. whitehouse.gov/blog/2013/05/09/ 52 A brief can be found at https://hiu.state. landmark-steps-liberate-open-data. gov/ittc/ittc.aspx. 63 Larry Lessig, Code: And Other Laws of 53 FEMA Deputy Administrator Rich- Cyberspace, Basic Books (2000). ard Serino and Federal Coordinating 64 Larry Page, CEO of Google, calls for Officer Mike Hall released these such a safe space for experimentation videos about the MapMill process: in his speech at I/O 2013, available at http://www.fema.gov/medialibrary/ http://techcrunch.com/2013/05/15/ media_records/10370 and http:// larry-page-wants-earth-to-have-a-mad- www.fema.gov/medialibrary/ scientist-land/. media_records/10369. 65 John Crowley, Volunteer Technical 54 USAID Crowdsourcing Study, p. 23. Communities: Open Development, World 55 Antonio Machado, “Proverbios y can- Bank and GFDRR (2010): 15. tares XXIX” [Proverbs and Songs 29], Campos de Castilla (1912); trans. Betty Jean Craige in Selected Poems of Anto- nio Machado (Louisiana State University Press, 1979).

56 Many individuals in DHOs are not U.S. nationals.

57 USAID Crowdsourcing Study, p. 8

58 OMB Memorandum, Antideficiency Act Implications of Certain Online Terms of Service Agreements, April 4, 2013,available at http://m.whitehouse.gov/sites/default/files/omb/ memoranda/2013/m-13-10.pdf, last accessed June 1, 2013.

59 FEMA Deputy Administrator Richard Serino, speech at Harvard University, May 13, 2013.