Big data is usually in the headlines for the wrong reasons – surveillance, exploitation of personal data for commercial or governmental ends, intrusion of privacy – but can also serve a valid and immensely exciting social purpose for development.
Kicking off a fascinating, packed and highly-interactive session, moderator Rohan Samarajiva, Founding Chair and CEO, LIRNEasia, set out this contradiction in perception of big data as a “competition of imaginations” between hype and pessimism, reminding us that big data is “of interest to all of us, as we are the creators of this data, the originators of this data”. Our mobile telephones, and by extension we ourselves, are permanently in communication with the nearest towers, sending out details of our whereabouts and activities in an ever-growing, highly personal call record.
This session aimed to “talk not about the imagination, but about what has been done”, exploring current and future trends in the use of big data for development.
Antonio Amendola, Executive Director International External Affairs, AT&T echoed this need to ground discussions in reality, explaining that the vast amounts of data generated by each of us each day, reaching the equivalent of 360,000 DVDs every single minute, have historically been seen as a liability rather than an opportunity, a by-product and a storage headache. Now that the power of data analytics is better understood in terms of customer relationships and efficient management in particular, “we can’t live without it.”
“It is huge, it is changing our lives, “ he continued, “but what is fascinating is that big data analytics starts by showing what happened in the past; if you study harder, it shows you why; and if you keep studying, you can even predict what might happen in the future or find out how to use it wisely and make that future experience better, whether by anticipating disease or diverting traffic or improving customer experience.”
Demonstrating just how that could work in practice, Linus Bengtsson, Co-founder of Flowminder, explained how his organization works with a number of operators around the world to map populations using satellite and mobile phone data in order to understand how infectious diseases spread or where best to target emergency relief. “The world is not stable. People don’t stay where they are, especially in times of drought or disaster, such as the recent earthquake in Haiti. Population mapping is fundamental to understand how people are distributed within an area and how they interact.”
Using anonymized call data records (CDRs) and distributed maps provides much more accurate data than government censuses conducted at long intervals, in particular in low- and middle-income countries. Bengttson shared a number of images showing variations in population movements due to seasonal factors such as summer holidays or outbreaks of epidemics such as cholera or Ebola. Following the phones proved far more accurate a proxy than estimates made by aid organizations at the time. Another big application is in the attempt to eliminate malaria, tracking how infected people move in and out of an area to understand the origin of specific cases, and steer resources and intervention more efficiently.
“Each person generates an average of 20 different events in the course of a day,” said Josh Blumenstock, Assistant Professor, University of Washington. “Using anonymized mobile phone data we can know approximately where someone is based on the nearest tower.” Patterns of communication change dramatically when a crisis happens in the real world, so that simply analysing call logs is enough to identify a violent event such as an earthquake. In areas where financial infrastructure is limited, mobile phone networks are used as a substitute to send support; a spike in mobile money transfer is often a much better early indicator than official data sources.
Satellite data is another important source of information, providing measurements on temperature rainfall or vegetation to map against known social and economic conditions. Extrapolating this country-wide produces a surprisingly accurate indicator of pockets of poverty or wealth. “Wealthy people use phones in a different way than poor people do,” said Blumenstock. “Poorer people miss calls more, the wealthy talk more, buy bigger chunks of minutes, make more international phone calls and have more central social networks. “ Plotting where poverty is located in the country enables targeted, real-time aid programmes and are an invaluable tool for the policy maker – or even for mobile operators seeking out elusive areas of the as yet unexploited moneyed minority.
In the context of developing countries where there are no or unreliable public censuses, mobile phone records also serve for socio-economic monitoring. CDR-based data is more accurate and picks up more detailed data for effective urban management such as waste or transport. Tracking how people move into and out of urban areas, which areas are more commercial or residential and how this might change over time has potentially huge implications for urban planning, explained Sriganesh Lokanathan, Lead Manager, BD4D Project.
But getting hold of the data sets in the first place is the major challenge. Everything hinges on the trust relationship between customer and operator; current market dynamics mean it is easy for a customer to move on, inhibiting operators from releasing data even in anonymized form and even for the wider social good. Making large chunks of government data open could lead to huge benefits for citizens and fairer allocation of resources, and might be more easily achieved – but for the ongoing issue of data security and personal privacy.
There is no single good answer, but several possible options, from anonymized data to encryption, using synthetic data to generate models or adding noise to distort the data time stamp. Informed consent and permission are very important, but the ensuring distortion must be worked around in statistical models. Finally, there are a range of third party technologies using open source used for managing internet data.
No one model can be adopted, especially as perspectives on privacy evolve over generations and vary across cultures. But if we spend too much time discussing how to work around privacy, government for development will evolve without us.
As Linus Bengttson objected: “These questions are important, but data from mobile phones is sent without encryption all around world, sold and given out to thousands of companies who have paid for it and use it – and they have thousands of employees.” A much larger potential breach of privacy – and a much less ethical use of the data, in particular in times of disaster.
The critical differentiator is the purpose of data analysis. And we should not forget how much our data is in the public domain without a higher purpose: from government censuses that specify ethnicity and religion to commercial, detailed household surveys or the ubiquitous flow of social media data.
Concluding the session, Rohan Samarajiva called on governments to pull together complementary data sets and get behind big data for development, with all the benefits it can bring at this early stage as: “This is terra incognita for us, we don’t know what we are exploring..and we need all stakeholders to play their part.”
Moderator
Dr Rohan Samarajiva, Founding Chair and CEO, LIRNEasia, Sri Lanka
Panellists
Mr Antonio Amendola, Executive Director International External Affairs, AT&T, Belgium
Mr Linus Bengtsson, Co-Founder, Flowminder, Sweden
Dr Josh Blumenstock, Assistant Professor, University of Washington, USA
Mr Sriganesh Lokanathan, Lead Manager, BD4D Project, Sri Lanka
Mr. Dean Bubley, Founder and Director, Disruptive Analysis, United Kingdom
Mr. Charles Matondane, Director, Africa, Kirusa Inc., Tanzania, United Rep. of
Mr. Carsten Rossbach, Partner, Roland Berger Strategy Consultants, Germany
Dr. Mohammad Shakouri, Chairman, WiMAX Forum, USA
Ms. Lise Tcheng, Vice-President and Global Head of Telecom Services, SAP, Ireland
