Much of what is discussed as “big data” does not include the poor, because smartphone penetration is still low, social media are not used by all classes and datafied records are rare in developing countries. Therefore, the session focused on research that has been/is being done on pseudonymized mobile network big data in developing countries. Instead the usual “battle of imaginations” which posits the optimistic scenarios that tend toward hype against the pessimistic scenarios that imagine all sorts of bad things that could happen, we began with reality. What had been actually done on the ground in countries as different as Namibia, Afghanistan and Sri Lanka were presented by data scientists who knew the ins and outs of data cleaning, pseudonymization, and what software needs to be used to analyze petabytes of data at a time.
The active audience raised a range of questions. Collective privacy, how one works with real-time data, and how one balances public and private incentives were discussed. The panel highlighted the importance of allowing freedom to experiment and innovate at this exciting early stage. It was pointed out in response to government statisticians that the best results would come when traditional data were combined with the new mobile network data that were by-products of the operation of the networks. It was concluded that collaboration among government authorities, private firms that had possession of the data and researchers who had the tools to extract the best lessons for public purposes would yield the best outcomes.
The moderator’s summary of key points from Big Data for Development session, 1100-1230 hrs, 9 Dec 2014