« Continuous Partial Attention | Main | Socialtext Open » Notes on "open source disaster recovery"The peer-reviewed web journal First Monday has a paper on "Open source disaster recovery" that discusses the Katrina Peoplefinder project. As one of the participants in that project, I noted some differences in the article and my perception of the project. I sent a few notes to one of the authors... repreated here for the record. Italic items are quotes from the paper. "Ethan Zuckerman led the technical effort of assigning chunks of unstructured data to volunteers and on Saturday, 3 September, a wiki was set up for this purpose at katrinahelp.info." The actual sequence of events:
Project leaders were hesitant to form a relationship with the Red Cross, whose database was built with assistance from Microsoft. Jon Lebkowsky writes in the Smart Mobs blog, “Marty Kearns of Network Centric Advocacy encouraged the PeopleFinder project to throw its data to Red Cross and to push for the Red Cross site to be the single authoritative search for evacuees and other Katrina victims, and family and friends searching for them. Marty’s suggestion implied a difficult question: should the PeopleFinder project end?” (Lebkowsky, 2005). - We weren't really hesitant to work with Red Cross, in fact, I was actively pushing for it, and we all understood the importance of that relationship. The real question was how to get their attention, and whether they would take over. That was resolved by simply making the PFIF data availble to them (and to Yahoo and Google), One key figure — David Geilhufe — oversaw the PeopleFinder project and negotiated relationships with other organizations. - Initially, yes, but leadership became fractal pretty quickly. David couldn't be expected to stay on top of it all. Ultimately multiple leaders emerged, and were more or less effective. I recall one conference call where a whole new set of people were trying to assume leadership with no sense of the history and direction of the project. (Those who had been working on the project evidently didn't have time to make the call.) Also note that ShelterFinder, a significant project and arguably more successful in meeting its goals than PeopleFinder, emerged from within the PeopleFinder volunteer base with a different set of leaders and processes. Each of the organizers highlights his or her role within the project, sometimes to the exclusion of other key figures. The issue of receiving credit for work carried out also applies to PeopleFinder's tense alliance with the Red Cross, when volunteers worried that the project would come to an end as a result. Yet the alliance was made despite these worries because working with a better known organization would increase the project’s effectiveness, giving victims access to a central, well–known place to go for information and assistance. Again, I think this conclusion is based on an overstatement of the tension that I myself reported. And it was less that the project would end than that the work would stop before finished, and that some of the work wouldn't be used. A final word: I haven't seen any evidence that our data helped anyone find someone who was missing, and design decisions early in the process made the data less useful, in my opinion. For instance, those entering data couldn't make changes, so if they made an error, they had to re-enter the data (so we had many duplications). The assumption was that it was better to work fast and have duplications than to slow down to correct errors. My biggest concern was that there was no way to reconcile 'missing' reports with 'found' reports. You couldn't update records, period. jon posted this at 4:01 PM |
read weblogsky! latest posts: |
