Monday, August 29, 2011

Starting from the end

Over the last two years we have had endless discussions about how crowd sourced information is going to change the way we do crisis information management. Some people go as far to say as the regular humanitarian information management is dead and that the time of crowd has come. But one thing that we have yet to show is that all this crowd sourced information actually provides the humanitarian response community with actionable information. We have a few anecdotes of individual reports being helpful, but no overall study of the effectiveness.

I have lately been talking to a number of colleagues from the humanitarian community and one of the best hint at how to solve this came from Lars Peter Nissen from ACAPS. He pointed out that when they are planning needs assessments they start by defining what decisions they want to try to affect by the needs assessment. Then they work their way backwards and design an assessment that helps provide the answers needed to make that decisions.

When deciding to do a crowd sourced project for a disaster or crisis response, we must do the same. We must first define what decisions we are trying to affect. Once we know what decisions we want to try to affect, we need to define what information we would use as the basis for making these decisions. Once we know what information we would use as basis, we should look at what is the best way to visualize that information to optimize the decision making. In the age of crowd sourcing we have focused a bit too much on the power of geospatial visualization, but often graphs, trends or tables can help us make a better decision.

Once we know what decisions we want to help facilitate and how we want to visualize them, then we can start thinking of how we can get data from the crowd and through data processing and data analysis turn that data into this information. This may lead us to ask the crowds for more controlled questions or for our media monitoring teams to monitor reports of certain data instead of trying to capture all the available data out there. We can then look at ways of either automatically process the data or use a mechanical turk to utilize a "crowd" to do that processing. Same applies to taking that processed data and analyzing it. This can either be automatic or done via a crowd of people.

So before the next major disaster happens and we activate the digital volunteers lets sit down and define the end product first and then work our way back. This way we can really ensure that all this digital volunteer effort is utilized to the max.


  1. I don't think that this approach is particularly reasonable, for a couple reasons.
    1. There is not some fixed end product that everyone is working towards -- and that's okay. Crowdsourced mapping isn't about solving a particular end goal -- in general, it's just an effort to make more information available more widely and more rapidly to the people who can use it.
    2. I'm not sure that I agree at all with your premise of what crowdsourced disaster recovery mapping is all about. When we were building maps to help respond to the earthquake in Haiti, we weren't trying to solve some specific goal, or create trends, or anything else like that: we were simply creating more accurate maps of the situation on the ground than had existed prior to that time. This was an extremely effective effort, connecting data from planes flown by aid organizations to websites used by crowdsourced workers to create OpenStreetMap data which was used on GPSes by workers in the field.

    Sure, there are some disasters that don't have the same need of information: Haiti was a special case in large part because it was such an *under*mapped territory to begin with. But I think that this post seems to ignore that type of effort -- and instead treat crowdsourced mapping as a needs assessment-style process.

    I think it's fair to say that crowdsourced mappers are not doing the same things as humanitarian aid needs assessments. But I don't think that's the primary value crowdsourced mapping provides in humanitarian crises. Creating more up to date and accurate maps using all available data sources and providing that data to classical needs assessment.

  2. As I have mentioned previously on my blog then the crowdsourcing community is often trying to serve two different purposes. One is to provide information to the people on the ground affected by the disaster. The other is to help those trying to respond.

    With all the effort that the crowdsourcing community puts into gathering data about the situation, it is essential that the data it is gathering actually makes some difference either for the people affected or for those responding. With either of these communities you need to start by thinking what decisions do they need to make and then work your way back. Otherwise you will overload the system with endless amounts of data that have no value at all.

    The OSM people did one of the most useful crowdsourcing effort in Haiti - creating maps - but if you think about it...why did they create help the responders make decisions on how to go somewhere...

    According to interviews I did with many of the responders in Haiti around 90-95% of the crowdsourced information was either not valuable or inaccurate/outdated. Hence the need to target it better to make it more valuable. Collecting data just for collecting data sake is not very efficient...

  3. This comment has been removed by the author.

  4. While I realize we're talking about first-world apples and developing world oranges here, Ushahidi has been perfect up here over the past week responding to Hurricane Irene. We've been able to make dozens of connections between responders and the affected population, with the blessing and participation of the civil authorities. By vacuuming up all the location-based info we can find/generate, both sides of the crisis coin have had the resources they need to act. Of course, I appreciate how fortunate we are here to have maintained a largely-intact telecommunications infrastructure throughout, but crowdsourcing has been an essential part of our response and now recovery: