May 2015 SciDevNet; The idea that big data from mobile phones could have helped predict how Ebola spreads and so saved lives in West Africa, if only telecom companies had released it, seemed both powerful and out of sync at a gathering of Médicins Sans Frontières (MSF) staff last week.
Ken Cukier, data editor of The Economist, kicked off the first of the medical aid charity’s two scientific days on 7 May with a charismatic talk that made the case for big data as the inevitable future and, in the case of Ebola, the “gold dust” that could have reduced fatalities.
Thinking about those lessons, I couldn’t help but wonder:
Would people trust health workers if masses of personal data owned by a private company were released to the government? What about privacy and confidentiality risks? Who gets to select, evaluate and interpret the questions asked of the data?
Will those answers, full of the authoritativeness of numbers, take precedence over the more fuzzy, but no less valuable, messages that might come from working directly with communities?
It was difficult to judge the audience’s reaction – there was enthusiasm on Twitter but big data came up just once again in the course of the day. It seemed to capture the imagination, as ideas that push boundaries can do, but also seemed a world away from what MSF staff are grappling with.
Cukier is a big data believer and coauthor of an international bestseller on the topic. His talk made the case for the revolutionary potential of big data by likening its impact to that of the printing press. Taking us back to a time when a simple machine enabled mass production of articles and books, he argued that “more isn’t just more” – it leads to other changes, such as niche publishing products in the case of the printing press. “More is new, it’s better and it’s different,” he said, and more leads to a change in state: there are things we can do with big data that we can’t do with small data.
With an amusing touch of mischief, Cukier mistyped MSF founder’s name into Google Search, to show how the collection of more data is already making a difference by enabling the search engine to self-correct.
Ebola is an example of how big data could have helped MSF’s work, he argued. At the height of the crisis in West Africa, the problem was knowing where and how the virus was spreading. Data from mobile phones could have helped staff target interventions by pointing to where people were going, where they congregated and for how long.
A spreadsheet flashed on the screen, seductively suggesting the power that orderly rows and columns can have over a messy crisis. But the data wasn’t released, Cukier said, because phone operators are self-interested, regulations inept and politicians clueless.
I found the notion of a technical solution jarred with the operational complexities and arguments for social science intelligence voiced on our website, and elsewhere, by staff with first-hand experience of the crisis. Much of what responders struggled with were things such as people’s beliefs about the disease and the interventions, power dynamics and poor infrastructure.
Cukier touched on a couple of these concerns briefly and acknowledged the value of small data. But his argument would be more convincing if it came with a more generous nod to the fact that, just like Ebola response measures, data doesn’t exist in a vacuum.
Subsequent discussions of MSF’s mapping work spoke to this. How do you deal with the risk of open-access maps getting in the wrong hands, for example military groups, or having a perceived bias, the audience asked a panel presenting the CartONG, Missing Maps and EPIET projects. Replace spatial data with big data and the question is just as valid.
MSF staff also described the groundwork that goes into making sure maps represent reality accurately. Villages may have different names or more than one name; or they may be missing from existing maps altogether. MSF solves these problems by mobilising local people to collect and validate location data. I wondered whether the same process of ‘groundtruthing’ might be needed if big data predictions are to be mapped onto areas with missing or inconsistent features.
Creating maps in areas without electricity was another challenge. How representative are big data predictions when mobile phone databases miss people without a mobile phone or in places without consistent Wi-Fi signals?
There are other nagging questions and doubts about the power of big data. But they don’t cancel out the promise. And seeing MSF staff describe how they already innovate to tailor their systems to the needs of particular projects, I wondered what productive uses they might find for masses of mobile phone data. But it will be difficult to realise big data’s lifesaving potential without a serious look beyond spreadsheets.