Crowdsourcing and the crisis-affected community

The term ‘crowdsourcing’ has multiple meanings in the context of humanitarian work. As in the commercial sector it often means ‘microtasking’ but it is also frequently used in a non-standard way to mean ‘sourcing information from the crowd’, that is, citizen-reporting. In the context of this article, ‘crowdsourcing’ is used to mean the structuring of data by a parallel workforce: translation, categorizing, extracting location names, mapping, filtering/prioritizing reports, etc. Citizen reporting is simply called ‘reporting’.

An ‘international actor’ means any organization or individual that is not Haitian or based in Haiti. For the most part, ‘international actor’ simply means ‘non-Haitian’, but there were several engineers at organizations within Haiti who would have been foreign-born and not of Haitian descent.

‘Ushahidi Haiti’ went through several name changes, but was mostly known as ‘Ushahidi @ Tufts’ during the actual deployment, after Tufts University in Boston where most of the students were based (or the associated Fletcher School). When interviewing people for this report, a number of people expressed discontent at this part of the initiative being called ‘Ushahidi’, with claims that several other organizations committed more work. It is not the intent of this article to resolve branding disputes, but it is flagged here as an issue that seemed to important to many of the international actors. The exact reasons why ‘Ushahidi’ also ended up as the brand for the entire collaboration are complex and largely outside the scope of this article. However, in terms of motivation, it is worth pointing out that is was generally positive to have an identity that many volunteers, working alone, could point to and say ‘I am a part of that’. The larger aid agencies were not widely trusted in Haiti (for right or wrong) after a long history in the country, and the US Military were still remembered by many as a former occupying force, so an unknown organization from neither the USA or Europe was a reasonable brand. I have not observed any Haitian organizations complaining about attribution. For those international organizations that wished to claim more of Ushahidi’s role, it should be kept in mind that they are fighting for credit over, at most, 10 % of the broader Mission 4636 initiative, itself a relatively small part of the response efforts.

Jennifer herself joined the initiative through knowing Josh Nesbit, who brought many of the initial parties together, so in this case it was a volunteer found through direct contacts, not social media.

This is not a criticism of the few students who stayed on, in fact, the opposite – those who were still working at Tufts or as representatives in Haiti through the fourth week and beyond, like Denise Sewell and Jaroslav Valůch, were respected by international and Haitian relief workers alike. Digital Democracy’s support of KOFAVIV, a Haitian-run organization that supports victims of gender-based violence, is an example of a continuing positive transfer of technical and security knowledge into the country. Another volunteer, Francesca Garrett, chose to help with Mission 4636 and became one of our most valued coordinators.

In terms of abbreviations, the one for Port-au-Prince favored by relief workers, ‘PauP’ only occurs once in 1000s of mentions in the text messages, and even then only when referring to a hospital run by relief workers. It is clear why PauP was adopted by the international community—it is one of the shortest possible unambiguous abbreviations – but it also highlights the linguistic disconnect between the crisis-affected population and the international community.

Remittances already accounted for more than a quarter of Haiti’s GDP before the earthquake (Orozco 2006), which is typical of many countries that are most vulnerable to disasters. Economic support for diaspora (through work or otherwise) might be a useful application of distributed crowd-funding more generally.

After some consideration, I decided not to list the details of the responses to messages here. The focus on emergency response and the triumphant tone that has accompanied it in so many reports does not sit well against the knowledge that so many people could not be helped. For their part, the military did not publish the details of the response because of need to declassify their information first. They apologized for this, but no one wanted them to divert any attention from response efforts. They shared a dozen or so specific response cases at the time, and have talked at length at a higher level (CTIA-The Wireless Association: Wireless Helps Consumers in 2012). The declassification issue was only significant at the time in that it would have been a good internal motivator to share more cases with the volunteers first-hand. However, it was also clear that our counterparts at the US Coast Guard did not alway know the outcomes of every dispatch as information moved through the chain, and it is generally true of all relief agencies that information about outcomes is rarely passed back directly along the communication lines, so perhaps this is a property of information in crisis-response more generally.

The Thomson Reuters Foundation were, in fact, planning to send individual replies with the Haitian Red Cross (Tim Large, p.c.), but this effort never launched.

In a more volatile security situation or actual war, it would be ill-advised to deploy a crowdsourcing system like Mission 4636 at all, as it would be too hard to vet all volunteers and too difficult to prevent against deliberate misinformation advising people to share personally identifying details through potentially non-secure channels.

http://​www.​reuters.​com/​article/​2010/​01/​17/​us-haiti-info-idUSTRE60G164201​00117.

http://​blog.​ushahidi.​com/​index.​php/​2010/​01/​17/​the-4636-sms-shortcode-for-reporting-in-haiti/​.

http://​www.​mission4636.​org/​history.

The authors of the Disaster 2.0 report have independently expressed regret at not being able to present more about the role of the Haitians and there is certainly no intent to deceive. I do not separate myself from the criticism, either – in my earliest presentation I credited Ushahidi with building the initial platform, not realizing that Brian Herbert of Ushahidi had added the ticketing system to Tim Schwartz’s platform (both Schwartz and I assumed the other was from Ushahidi for the first couple of weeks by virtue of working on Ushahidi urls that were mapped by Herbert—no one misled anyone). Herbert was invaluable for these early few days, as was Mary Jane Marcus of InSTEDD who was one of the most important people helping coordinate translators at that time. So the NGOs do deserve credit for some of the core platform too—both were held in high regard by the people working on Mission 4636, both made time to be an extra pair of eyes when we later made the switch to CrowdFlower, and would have been welcome to remain longer term workers on the deployment. However, the role of these organizations is already well publicized. No one would dispute that there were hundreds of (mostly Haitian) individuals who worked longer and harder than anyone from either of their organizations, and this report is about them.

The term ‘independent’ does not mean the same thing as with a typical review at an engineering venue. The creators of the review primarily composed the report by allowing members of Ushahidi Haiti and some outsiders (including myself) to collaboratively write and rewrite the narrative through collaborative word-processing. Relative to peer-reviewed reporting it would probably be called something like ‘creating consensus narration.’ For example, Sharma’s early report (which he reports to have written while an intern at Ushahidi) lists Ushahidi partner Chrissy Martin as the ‘Ushahidi Haiti Evaluation Manager’ (Sharma 2010), although her name is omitted from the final ‘Independent’ report.

http://​blog.​ushahidi.​com/​index.​php/​2010/​02/​06/​ushahidi-how-we-are-doing/​.

To correct this narrative, the reason we used microtasking is much more mundane: it came from industry. I was first evaluating the possibility of using automated methods for processing the messages as I had already been researching with Medic Mobile (Munro and Manning 2010), but it was clear that automated methods would not be accurate enough. I had independently worked in both commercial and academic language processing through microtasking, in Silicon Valley start ups and at Stanford University (Munro et al. 2010; Munro and Tily 2011), and separate to the main Mission4636 collaboration I saw offers of help from 1000s of Haitians outside the country. So I instead pushed microtasking as a viable strategy to combine these. I had never heard of Ushahidi or their work in Kenya, so this version of the history is apocryphal (for 90 % of the crowdsourced information processing efforts, at least). However, as soon as Brian Herbet of Ushahidi joined Mission 4636, his first words were to support this strategy, so this wasn’t a competing approach, with many vital people contributing to the design and implementation. At launch, I was the ‘volunteer coordinator’ for Mission 4636, with several short steps to running the initiative more broadly. We lost our full-time commitment from the NGO-based engineers after the first week, meaning that I also took over engineering and platform. I evaluated several established crowdsourcing platforms before negotiating the transfer of hosting to CrowdFlower after an introduction to the CEO, Lukas Biewald, from a Stanford connection. The first meeting with CrowdFlower revealed that Samasource had signed an agreement to establish a work center in Haiti working with local NGO, FATEM, and within a few days we had formed an agreement to expedite the construction of their center so that they could take over the core work. I therefore also took over the role of liaison between Mission 4636 and our new main organizational partners, but Josh Nesbit remained as the connection between Mission 4636 and the State Department, who would come to fund the FATEM workers. For a high-stress crisis-response scenario, perhaps building on a more well-established information processing background is preferable.

The radio stations in Haiti are a clear exception to this. Broadcasting in Haitian Kreyol, they were only trying to address the crisis-affected population and not concerned with publicizing their engagement to the broader world.

http://​www.​mission4636.​org/​history/​.