Crisis Mapping Haiti: Some Final Reflections

For various reasons, the Ushahidi-Haiti Project continues to be cited in numerous reports, book chapters, media articles and conferences more than two years after the devastating earthquake struck the city of Port-au-Prince resulting in massive casualties and loss of life. For many, the ad hoc initiative remains a major milestone in the field of crisis mapping. The project demonstrated the power of online volunteer networks and highlighted the potential of new technologies for humanitarian response. Some cite the initiative as the inspiration for their own projects and humanitarian efforts. For others, particularly critics, the project continues to be a major obsession. They strive to identify every possible mistake that volunteers made instead of offering constructive criticism and better practices. In short, they are absolutely masters at smart-talk. Dealing with smart-talkers is time-consuming and unpleasant. They talk rather than do and tend to provoke an endless debate of back-and-forth arguments, often verging on personal harassment. So here are some final reflections. There is absolutely nothing to hide. Indeed, many of us continue to learn a lot from the Ushahidi-Haiti Project. For example, we collected some sensitive information during the project. My friend Josh Nesbit secured a free SMS short-code from Digicel Haiti within 48 hours of the earthquake. The purpose of this short code, 4636, was to crowdsource information on urgent needs from the disaster-affected population and to map these text messages on the public map. Of course, this presented some data protection and privacy concerns. This explains why we consulted a Professor at The Fletcher School of Law and Diplomacy who is an expert on Haiti and has a law background. We specifically asked him whether we could add the content of these text messages to the public map, particularly if some of these contained personal identifiers. Personal identifiers are often key pieces of information because they can help render messages more actionable for responders. We posed our question via email and the Professor's reply, dated January 18, 2010, was as follows: "It seems quite clear to me that if you are able to obtain their numbers and they are sending you this information, consent is implied."  Still, we decided to seek a second and third opinion just in case. Another knowledgeable colleague at Fletcher responded with the following on the same day (and also via email): "If people are texting you, with the intent of getting aid or reaching out to someone, then consent would be implied. They know what you are doing, right?" The 4636 short code had been broadcast via local community radio and Diaspora radio, explaining what the service was for (more on this below). This was the principal way that the project was disseminated in Haiti. (Incidentally, when we were told that there was implied consent to publicly share the reports in full, I shared this news right away with all volunteers on the team who, like myself, appreciated the guidance. For example, Rob Munro who did an outstanding job coordinating the volunteer translation efforts for the incoming text messages replied via email with: "Thanks Patrick - that's reassuring! I can go with whatever was decided.") Now, the advice provided by the contacts at Fletcher was not official or legal advice, but rather personal advice. Seeing that the situation was "really pretty low risk," to quote the Fletcher Professor, and that time was of the essence, we decided to map some of the most urgent life and death text messages. We also had members of the Haitian Diaspora in Boston helping us with the mapping effort who knew would directly benefit from having this data open since they were particularly active in the massive online response efforts. Indeed, the response coordinated by the Diaspora is one of the biggest untold stories of this entire effort. Haitians in Port-au-Prince were also able to leverage the map precisely because the data was open. Indeed, Carel Pedre, one of the most well known radio and television personalities in Haiti was able to use the mapped SMS's to communicate directly with disaster affected communities. In any case, the possible harm that could come to local populations was minimal. Do No Harm is a standard principle for anyone operating in a humanitarian crisis or recovery context. The principle is important because it recognizes there is a risk to any intervention. And what is critical in deciding whether a certain course of action is ethical is to consider whether it could potentially cause harm to the local population. Even if there had been some Haitians who did not want their messages published, we all felt that the possible harm that could result was indeed "really pretty low risk." Note that when my colleague Josh Nesbit secured the SMS short-code for us, it was specifically for this Ushahidi-Haiti Crisis Map, and not for any other parallel initiative. Josh's tweet below, published about 20 hours after the earthquake, is clear proof of this. As it turns out, someone following Josh's twitter feed from Cameroon saw the tweet and put him in touch with a colleague in Haiti who worked for Digicel, the main telecommunications company there. In other words, the 4636 project was never meant to be a side-initiative or parallel to the Ushahidi-Haiti efforts: the short-code was set up specifically for the Crisis Map that my colleagues and I had launched a few hours after the earthquake. With the short code secured, my Ushahidi colleague Brian Herbert literally pulled an all-nighter to customize the web-based interface below. The purpose of this second platform was to enable Haitian-Creole speaking volunteers to translate and geo-locate text messages sent to Digicel's 4636 short-code. In effect, the platform was designed to enable the crowdsourced translation and geo-location of incoming text messages. Again, this entire effort was geared towards mapping the needs of disaster-affected communities in Haiti as part of the Ushahidi-Haiti Project. The whole point of securing the short-code was to add the content of text messages to the public map. As this was happening, some partners in Haiti helped us get the word out about the short code via local community radio. The agreed-to messaging was "Text 4636 with your most urgent needs and location." They spoke with dozens of local radio stations to explain the messaging and overall purpose of the project. Some 1,100 volunteers reportedly logged on to Ushahidi's translation platform and apparently translated tens of thousands of text messages. But less than 2% of these—only the most urgent text messages—were made public on the crisis map of Haiti. Since a number of these SMS's still required more precise geo-location before they could be added to the map, we worked closely with many members of the Haitian Diaspora in Boston who obviously knew their country far better than any of us did. A number of our Haitian friends actually joined us (often for long hours on end) in our make-shift "Situation Room" at The Fletcher School (picture below). Incidentally, the woman pictured below, Sabina, was our dedicated Diaspora liaison, given her previous work in Haiti and her complete fluency in Creole. Sabina and I spent many (many) hours on Haitian Diaspora Radio and Television to explain the purpose of the Ushahidi Crisis Map and the dedicated short-code that had been set up for the map. We answered dozens and dozens of questions from callers who were looking for more information on 4636. Since we (together with our partners in Haiti) explained the purpose of the map so publicly, one humanitarian colleague noted that "it does not make sense to suggest that users of 4636 did not know their messages would be made public." In other words, consent was implicit in sending the message. On a related note, it is worth keeping in mind that at no point did any of the Haitian journalists or callers we spoke to suggest that the map or SMS's be made private. Everyone, including the Haitian volunteers helping us at The Fletcher School clearly saw and understood the immediate value that the map provided. So did the US Coast Guard and US Marine Corps who made active use of the "real-time" content displayed on the map. On January 19th, just a week after the earthquake, someone from the US Coast Guard emailed us with the following question: "I am compiling reports from Haiti for the US Coast Guard and Joint Task Force Command Center. Is there someone I can speak with about how better to use the information in Ushahidi?" Several days later, we set up a dedicate Skype chat with the Coast Guard to fast-forward the most urgent (and actionable) text messages that were being added to the live Haiti Crisis Map. On January 22nd, a few days after we began mapping urgent life and death text messages, the US Marine Corps got in touch with us via email: "I am with the US Marine Corps. I am stateside assisting the 22 MEU [Marine Expeditionary Units] coming off the USS Bataan [on the Haitian Coast]. We want to use your data to bring aid to the people Haiti right now. The USMC is focusing on Leogane, Grand Goave, and Petit Goave. Is there a way to import your data into Google Earth or GIS? We want to make this work for the people of Haiti...please let me know ASAP" Needless to say, we replied ASAP. On the same day, FEMA's Administrator Craig Fugate,  shared the following on his Twitter feed: Five days later, the same contact from the US Marine Corps shared the following by email (which I am making public in full for the first time): "I can not overemphasize to you what the work of the USHAHIDI/Haiti has provided. It is saving lives every day. I wish I had time to document to you every example, but there are too many and our operation is moving too fast.  Here is one from the 22 MEU: 'We had data on an area outside of Grand Goave needing help. Today, we sent an assessment team out there to validate their needs and everything checked out. While the team was out there, they found two old women and a young girl with serious injuries from the earthquake; one of the women had critical respiratory issues. They were evacuated.' Your site saved these people's lives. I say with confidence that there are 100s of these kind of stories. The Marine Corps is using your project every second of the day to get aid and assistance to the people that need it most. We did have a tech barrier that we had to surmount. The Marines downrange have Google Earth and your site does not work on the ship for them. So, I had Georgia Tech create a bridge from your site to Google Earth. But it is YOUR data and YOUR work that is putting aid and assistance directly on the target and saving lives. Our big gap right now is locating NGOs and where they are working. Your site is helping with that. Keep up the good work!! You are making the biggest difference of anything I have seen out there in the open source world." While the experts we consulted at Fletcher opined that there was implied consent vis-a-vis making the SMS's public, and while this information reportedly helped save hundreds of lives, there's absolutely no doubt that many of us would handle this kind of information differently in the future. Since Haiti was in so many ways unprecedented, with everything happening incredibly fast and with no established protocols on data protection to guide this specific type of operation, a number of experts I have since consulted have noted that we were indeed largely on "unchartered territory," with the added pressure of time. In many ways, unfortunately, we still are. This is precisely why I wrote my previous blog post on "Crowdsourcing, Crisis Mapping and Data Protection." Incidentally, I called for an "SMS Code of Conduct" in this blog post just two months after the Haiti earthquake, and did so again in this co-authored and peer-reviewed publication. The good news is that a major humanitarian organization is now taking the lead to articulate and propose appropriate data protection protocols to deal with this type of situation and related challenges vis-a-vis SMS and social media. Indeed, the ICRC and a consortium of NGOs are busy updating their official data protection protocols given the fact that disaster-affected communities are increasingly digital and thus the source of "Big Data" for disaster response. We're especially pleased to be actively collaborating with the ICRC in this process. But what exactly would we do differently vis-a-vis Haiti? No one who was involved in the Ushahidi-Haiti efforts, as far as we know, has actually taken the time to pro-actively and constructively propose some forward-looking suggestions to this effect vis-a-vis data protection. I did try to catalyze that SMS Code of Conduct just two months after the earthquake, but if you know of any other related initiatives, please do leave a comment below. So here are some preliminary, pro-active ideas. One might be to ask those Haitian volunteers who translated the incoming text messages wether or not the resulting translations should be shared publicly on a live map. I'm not sure this makes much sense from the point of view of "consent" since they are not the ones who created the original content. That said, the IOM Data Protection Principles, which I blogged about here, do mention the notion of consent by proxy: "oral or written consent provided on behalf of data subjects by relatives, authorized community members or close associates in exceptional circumstances." I imagine there were some relatives and possibly close associates who were translating text messages from family members and friends. This may also have been the case for the < 2% of text messages that ended up being mapped. So perhaps the 1,100 volunteers should have been asked to make a note if an SMS they processed came from a family member and whether they consented to making the content public along with any personal identifiers. This type of feature could have very easily been added to the web-based translation platform that Brian Herbert developed. Unfortunately, no one (as far as I know) proposed this feature at the time, and I myself was not personally involved or responsible for managing the volunteer SMS translation efforts for the Ushahidi-Haiti Project.

Many Haitians created public we-need-help signs, often including personal identifiers

Incidentally, I did recently consult an experienced humanitarian colleague about this "consent by proxy" idea in the context of Haiti. Her reply: "Would we really want to go down this path? I think consent by proxy through translators who may or may not happen to be your relatives is not worth considering. This is not really scalable, nor predictable." Nor, I imagine, is it automatically verifiable. Another suggestion would be to try and obtain explicit consent directly from the sender of the original text message. This would have been technically feasible since the Ushahidi-Haiti platform allowed volunteers at The Fletcher School to text back the original sender of an SMS for further information on their stated need. The SMS cost on our end would not have been a big issue—a few hundred dollars at most since we mapped fewer than 2% of the text messages obtained from Digicel's 4636 short-code. The cost on the originating end would also have been zero since the short-code was free to text to within Haiti. (That said, there were real concerns about preserving the battery life of those texting from Haiti). Some may argue (and indeed have) that texting every Haitian back for consent may not have been practical since we were already operating at way-beyond capacity and hardly getting any sleep for days on end. Indeed, the two main volunteers at Fletcher who were tasked with the job of filtering through thousands of text messages and mapping the most urgent ones were quickly facing a backlog. Perhaps this feature could be automated in the future. And perhaps the messaging that goes out should ask people to include "No" in their SMS replies if they do not want their text messages made public. Something like: "Text 4636 with your urgent needs/location but add NO if you do not want your SMS made public." This would perhaps avoid the need for a second round of text messages and thus obviate the need for processing more data. Of course, an entirely different course of action would have been to keep all text messages private by pushing them to a password-protected map available to the US Coast Guard and Marine Corps only. (In fact, we did something along those lines for the Libya Crisis Map last year vis-a-vis UN and other humanitarian organizations). But again, I was not working on the SMS side of this initiative during Haiti, and as mentioned above, we had already consulted an expert on this issue who opined that there was indeed implied consent. If that had not been the case, then there's absolutely no doubt that we would have looked for an alternative solution, like a password-protected site and/or opening up a dedicated Skype chat with the Marine Corps  like we had with the Coast Guard. In sum, there was general agreement that consent was implied in this extreme-but-low-risk situation and that we might do more good—and possibly help save more lives—if the data was kept open. In this case, our assumptions turned out to be correct. In the future, however, I believe that some (or all) of the suggestions I proposed above need to become standard operating procedures starting now. In conclusion, whilst we need to learn from the past, we also need to be forward thinking and action-oriented. There have been several serious crisis mapping deployments since Haiti, and each of these provide equally important lessons. While some volunteers were only involved in one major crisis mapping project (i.e., Haiti), that was over two years ago now, and many of us have since been directly engaged in several major back-to-back crisis mapping projects. Today, there is a lot more to crisis mapping than the deployment in Haiti—and indeed more questions than answers. Moreover, the 2010 earthquake was an outlier in terms of disasters and humanitarian response. This is yet another reason why we need to broaden our hands-on experience and draw on lessons from crisis mapping deployments beyond January 12, 2010 and not get caught in endless debates with smart-talk masters. Onwards.