Potential, pitfalls of "big data" for humanitarians

Humanitarians and crisis-affected communities are increasingly exploiting new technologies to gather near real-time information to prepare for, prevent and handle disasters, even as analysts caution against overreliance on “big data”.

“Our opportunity today is to responsibly use the data to improve the aid systems to help [vulnerable communities] at every stage: from response and recovery through mitigation and preparedness for future disasters,” said Anoush Tatevossian, spokesperson for UN Global Pulse, a digital-tracking initiative founded in 2012 to research how aid workers can best exploit an ever expanding mass of data.

Big data - or the “traces of human action picked up by digital devices” according to the International Peace Institute (IPI), a non-profit peace and security think tank based in New York, crowdsourcing (soliciting contributions from the online community), crowdseeding (providing communities with mobile phones and credit to gather on-the-ground information from them) and social media (interacting in virtual communities such as Facebook and Twitter), have all emerged as increasingly key elements of the 21st century humanitarian response.

Modern information and communications technology (ICT) made it possible to find survivors buried under rubble in Haiti’s 2010 earthquake; more rapidly translate the UN Refugee Agency's English portal to Arabic to allow refugees and local responders easier access to information; and evaluate public sentiment during the Philippine’s 2012 Typhoon Pablo by categorizing over 20,000 social media messages in a database within 24 hours of the typhoon hitting.

Real-time information

Big data is currently being studied retroactively by the UN in what it calls “Global Pulse Labs” in Jakarta (Indonesia) and Kampala (Uganda).

The Indonesia lab created in 2012 found that tweets sent from 2010-2011 in Indonesia - a country where residents send and receive more tweets than any other worldwide - reflected the impact of rising food prices and inflation on the population, and served as a warning of the 2012 global food crisis, said Tatevossian, citing early research that showed a correlation between social media conversations on food-related topics and official inflation data.

This type of real-time social media monitoring could strengthen early warning systems, according to Global Pulse.

“We can approximate consumer price indexes for basic foodstuffs through keywords and the foods people discuss online,” Tatevossian explained.

With the labs’ research focus on past data, initial findings demonstrate the importance of using free digital information to inform policy decisions. “Decisions are often based on two-to-three-year-old statistics, while this ocean of data is produced for free all around us,” said Tatevossian.

There are more than six billion mobile phone subscribers globally, one billion Internet users in Asia alone, and one-third of the world’s estimated seven billion people has access to the Internet, according to the UN Office for the Coordination of Humanitarian Affairs (OCHA).

With more than one billion new cell phone subscriptions worldwide in the past three years, according to Patrick Meier of the Qatar Computing Research Institute (QCRI), and some 1.7 million billion bytes of data created every minute of every day, the so-called “digital divide” continues to narrow.

But governments, the UN, and humanitarian NGOs still need to learn how to use these new sources to make better decisions, according to Paul Currion, an IT and humanitarian coordination specialist and consultant for the International Council of Voluntary Agencies (ICVA) - a Geneva-based NGO network on humanitarian developments.

"The price of improved information coming at us from multiple angles [such as] communicating with disaster-affected communities, joint needs assessments, crisis mapping, and so on, is that we have less and less excuse for poor decision-making," Currion stated.

Manmade disaster

“For the prevention of violent crime the example of Latin America showed how horizontal citizen-to-citizen ICT initiatives are the most dynamic and promising,” said Francesco Mancini, the senior director of research at IPI.

In Brazil, Rio de Janeiro slum residents provide feedback on violence-prevention interventions through blogs; while in Mexico, the peace advocacy network Nuestra Aparente Rendición uses its website to facilitate communication between peace activists and raise awareness of new developments.

Unequal access to technology continues to bar social media from some populations, but Latin America has one of the fastest growing rates of Internet penetration and blogging, with online access growing 13-fold and cell phone subscriptions expanding 10-fold since 2003, according to IPI.

"New technologies might drive us further away from the human-to-human relationship that is so critical to the humanitarian endeavour"

Countries throughout Central America, where the national media often self-censor due to fear of retaliation by organized armed groups, citizen journalists have taken to social media and blogs to report violence.

With the highest under-24 homicide rate of any region in the world at 35 per 100,000 persons, according to IPI, and two-thirds of all Internet users under the age of 35, ICT can be used as a tool to contribute to boosting citizen safety in Latin America, say analysts.

In other parts of the world, social media has been used to incite violence through hate speech, such as in Myanmar’s Rakhine State in 2012 and 2013, but monitoring can also predict and mitigate the consequences of such violence, according to IPI and Global Pulse Lab.

“The rapid feedback and impact evaluation [with] real-time information about the behaviour of communities… allows for rapid, adaptive course correction… and achievement of results sooner,” said Tatevossian.

Natural disaster

The Pulse labs are still deciding just how useful big data is before trying out real-time monitoring. In 2013, Global Pulse is working on a project to determine how mobile phone data can provide early warning about the impact of drought on communities.

While the first responders to a crisis are almost always members of affected communities such as neighbours and local authorities, social media may motivate provincial and federal governments to mobilize resources and act on impending crises.

“New technologies spread information at almost real time among those affected by crisis, and to their local leaders who have a direct incentive to act… This is what I call the bottom-bottom [horizontal] approach,” said Mancini, contrasting it with a "vertical" flow to national and international decision-makers who are less accountable to local voters.

In the Asia Pacific, governments, NGOs, and communities have tapped into social media to create early warning systems for disasters.

The Filipino government used Facebook to warn people of imminent floods in December 2011 and the Digital Humanitarian Network (DHN) - a consortium of technical experts and digital volunteers convened by OCHA in December 2012 after the onset of Typhoon Pablo in the southern Philippines - has since been activated five times (Sudan, Philippines, Syria, Samoa, and the Democratic Republic of the Congo), facilitating the creation of crisis maps, improving the speed and accuracy of information to responders and reducing data collection costs, according to Cat Graham, a coordinator for DHN and vice president for Humanity Road.

"The humanitarian community needs to engage more... with the professional communities already working on that technology," said Currion.

DHN assists humanitarian groups to interpret data to map crises where tweets and Facebook posts are tagged to videos and photos and then located on a live map. It uses satellite technology to aid in search and rescue efforts, and identifies people who need help through Facebook posts.

In South Sudan, the team collected more than 15,000 pieces of data on displacement, returnees, security and other issues. In the DRC, it created maps and population statistics for more than 250 regions, and in the Philippines looked through more than 20,000 social media messages.

“The domain of the NGO is facing competition from peer-to-peer type grassroots relief efforts where these types of technologies are enabling smaller community-driven organizations to operate on the same level as the less agile bigger players,” said Michael Howden, a director of the Sahana Software Foundation, an “open-source” software company that allows users to download disaster-management software for no license fee.

Since launching in 2004 to aid Sri Lanka's recovery from an earthquake and tsunami that same year, Sahana's software has been deployed more than 12 times, such as by the Chilean Red Cross in 2012 to assist with detecting wildfires; helping community-based organizations organize Hurricane Sandy responses in New York City and New Jersey; registering charities responding to the 2011 earthquake and tsunami in Japan; and tracking more than 700 NGOs after Haiti's 2010 earthquake.

Data’s downside

However, digital data and new technology, including satellite technology and open-source software, is far from a panacea for humanitarian and development problems and can actually put people in harm’s way, others warn.

Aid organizations should not look to technology to solve their problems, but only for additional research and insight into populations in need, advised Global Pulse.

“Big data can be a buzz term and it cannot solve everything,” Howden said. “We shouldn't get caught up in the hype around buzz words and ensure that we apply appropriate technologies for the best solutions,” he added.

The “data deluge”, as termed by Global Pulse, can make it difficult to filter out relevant, life-saving facts.

For example, medical experts analysing flu trends through Google web searches from 2003 to 2008 found such searches (Google Flu Trends) predicted respiratory infections accurately, but not the influenza virus. People's perceptions of their flu-like symptoms did not always match medical diagnoses.

But such a deluge also contains valuable "data exhaust", noted Global Pulse about the trail of data that matters little to the holder of the data, but can be useful to others (mobile phone tower usage records to track displacement in Pakistan, for example).

Big data, mostly sourced from self-reporting, cannot be the only source of information to inform policy, caution analysts, who have pointed out the difficulty of verifying public perceptions reported en masse. While "digital straws" - comparing social media data with national media reports and triangulating with other tweets and posts - can help sift truth from falsehood, there is still a need to identify which situations will benefit from which kind of data and invest accordingly.

"There are barriers, blindspots, and biases in the data itself that will limit its utility in some situations," said Currion.

In addition, “there is a very real risk of overload of [staff and database management] systems that are not yet ripe to manage such a quantity of information,” said Jérémie Labbé, IPI’s senior policy analyst for humanitarian affairs. While the DHN’s free digital volunteers can contribute to data analysis and flow, the vast amount of available information may overwhelm NGO workers on the ground.

“The absorptive capacity of responders is pretty low. It’s not because they do not have an affinity to technology. It’s because they are really busy 98 percent of the time, and they are sleeping the other 2 percent,” said the UN Global Pulse's director Robert Kirkpatrick.

When virtual replaces human

Interacting virtually may also alienate humanitarian workers from communities they are trying to reach, warned Labbé.

“Just like an abusive use of social media can affect actual physical face-to-face contacts between people… new technologies might drive us further away from the human-to-human relationship that is so critical to the humanitarian endeavour.”

“It’s important not to lose the human connection in the cloud of data,” added Howden.

If service providers rely too much on new technology to communicate with disaster-affected communities, populations without access to, or understanding of  that technology, risk being overlooked.

“Unequal access [to technology] may mirror conflict cleavages, [and] problems with the representativeness of the data take on a whole new dimension,” reported IPI, which emphasized the need for all segments of the population to be included in any assessment to avoid information bias.

Not forgetting the human faces behind the current hype about big data, mapping, and tweeting is the only way to meet the humanitarian needs of populations, said Currion.

"Big data can help us to build the big picture, but we must never forget that behind every dataset, behind every map, behind every tweet, are people striving to live their lives with dignity in the face of great adversity. The fundamental question we must ask of any technology initiative is always: will this help those people?" he said.

Modern technologies in humanitarianism “are still in their infancy with much room for growth”, concluded Labbé. Or, as Global Pulse’s Tatevossian sees it: “Big data is the new research and development sandbox.”

dm/pt/cb