Archive for the ‘Microtasking’ Category

The tech community’s response to Ebola

Monday, June 8th, 2015

I gave an interview this week about the response to the recent Ebola outbreak, talking about best practices for privacy, the media, and silicon valley’s response. Here’s a short transcription of the major questions:

What is your connection to the Ebola-hit regions of West Africa?

“There are several, as I’ve lived there, worked in Epidemic Tracking, and we currently support the UN’s health tracking in the region at Idibon.

Before I moved to the US I worked in environmental development in Sierra Leone, and for the United Nations High Commission for Refugees (UNHCR) in Liberia installing solar power. I installed some of the power systems in clinics treating Ebola patients, so I am familiar with the level of health services in the region.

Following that work, I helped found Energy for Opportunity, the largest installer of Solar Power in Sierra Leone. When I was last there, three years ago, the company was 100 people strong, with all but one person born in the region: it was a great example of transferring skills into areas where jobs and capacity is needed most.

Engineers from Energy for Opportunity installing power at Panguma hospital in Kenema, Sierra Leone.

Engineers from Energy for Opportunity installing power at Panguma hospital in Kenema, Sierra Leone in 2013.

And you have worked in epidemic-tracking, too?

“Yes, I have worked directly in epidemic tracking, with Metabiota in 2011. We were tracking an Ebola outbreak in Uganda several days ahead of large health organizations.

Currently, Idibon is helping track disease outbreaks in a number of countries. We have kept most of these private for data privacy reasons, but some of the work was recently covered in Nigeria, where we are partnering with the United Nations Children’s Fund (UNICEF) for SMS-based reporting.

What were the main takeaways from tracking Ebola in Uganda in 2011?

“The biggest takeaway was that it was possible to track the spread of epidemics very closely by bringing in reports from the ground. The second biggest was that we should not be publicizing this information. The ways that someone can be helped from a distance are, unfortunately, very limited, compared to how someone can be hurt locally in areas where victims are often blamed for contracting an infection.

No real-time reports about outbreaks should ever be reported on public reports, maps or other public-facing forums.”

Video: Robert Munro in April, 2014, “Crowdsourcing and Natural Language Processing for the Social Good”, at the Conference on Computational Social Science, hosted by the Stanford University.


At 25 minutes, Robert talks about tracking Ebola and the importance of privacy.

Were these insights relevant for Ebola in West Africa in the last year?

“Yes, very much!

Near the start of the outbreak in August last year, I estimated the death toll for people with treatable illnesses who were avoiding clinics for fear of Ebola.

My estimate was that other deaths would be 10 times greater than for the Ebola cases themselves. Sadly, this was confirmed last month at an Ebola Summit here in San Francisco by Madina Rahman of Sierra Leone’s Ministry of Health. The implications for this are chilling: if you launched an Ebola information campaign in West Africa without knowing how it would change behavior, then there was a 10 to 1 chance that you made the situation worse, by heightening fear and distrust.

The same is true for the international media who knowingly portrayed the dangers to people outside the region as being much greater than the really were: the media are responsible for 1000s of preventable deaths within West Africa.”

What was Tech Community’s response to the disaster, and how did they support Humanitarian Technology?

“First, there is no such thing as “Humanitarian Technology”. The problems that we face in humanitarians crises are problems that we face everyday: bursts of information, time-critical reports, the need to coordinate with multiple stakeholders, etc.

There is nothing that is required in addition to existing technology, so “Humanitarian Technology” tends to be a distraction from tools that people actually use. You cannot deploy technology in disasters alone, untested in each new situation: the technology needs to have been vetted for functionality and security well in advance by as much of the broader tech community as possible.

The biggest way that the Tech community responded was by ensuring that the infrastructure already existed: the cell-phones and towers that were already in use, the reliability of email clients, databases, spreadsheets and other building blocks of information sharing. These are the true humanitarian technologies: the pieces that work without you thinking about them when you need them most.

For specific examples? I can think of a few. I advised most major technology companies (Google, IBM, Facebook, etc) and governments both within and out of the region as they planned their response. The biggest failures were companies that wanted to try something big, like tracking every patient in the country. It would be a struggle to set that up in a short period even outside of a disaster. The biggest successes were by companies who understood that they needed to provide infrastructure, not innovation.

The standout for me was Facebook. Their not-for-profit arm diligently spoke to experts and decided that their greatest impact would be ensuring internet was available. This impressed me: rather than letting health organizations openly track the outbreak through messages on their platform, or some other integration with their core business, they realized that they would be violating privacy with no clear benefit. I remember my friends in Liberia were posting images/videos on Facebook of Ebola patients leaving hospitals to get food at markets. These were individual cases and fortunately the international media missed the first few weeks of these cases, when tension in Liberia was at the highest. If these videos had been found by the international media earlier, the hysteria and resulting unrest would have been devastating. The people I spoke to at Facebook rightly identified that they couldn’t evaluate the impact with enough certainty for anything other than basic internet.”

What have we learned from the response that will help for next time?

“The biggest takeaway is that we need to keep the media in check. Open reporting about the outbreak hindered efforts in many ways: airlines canceled flights (and resources) into the regions when it wasn’t warranted; the governments within the region took drastic actions to look proactive (curfews, military at borders, etc) when experts said these actions would have no effect; and people locally died by the 1000s as a result. I also read somewhere that more than 40 organizations launched sms ‘reporting’ services, when they had no capacity to respond, meaning that people who already knew who to call/text at a local clinic were instead reporting valuable information that went nowhere and helped no-one. Ultimately, we needed more local government transparency around their decision making, but the international media’s floodlight of hysterical attention drove the local decision makers underground, so we got the opposite of what helped: local officials keeping quiet and international media failing to understand local conditions, reporting only the worst.

The other takeaway is that basic infrastructure is more important than any tools specific to disaster response. Like I said earlier, the backbone of the response for tracking the outbreak was built on information technology that was already in place: cell-phones, standard databases/content management systems, basic email clients, etc. The majority of people in the region have access to a cellphone, so we are there in some areas of basic infrastructure, but less than 2% of people in the worst hit rural areas have access to grid electricity, so we are behind in other areas. The next step will be ensuring that future infrastructure is standard and robust. This will include proactively keeping out the NGOs who will come flooding in with software that is packaged as “humanitarian” but is ultimately sub-standard.”

Any final advice for Silicon Valley companies who want to help with epidemic tracking?

“Support more languages! There are 7,000 languages in the world. 90% of them are spoken in the small band of the tropics that the majority of our pathogens emerge from. When a new disease emerges, chances are that is one of the most linguistically diverse places on earth and it is spoken about for the first time in any one of 1,000s of different languages. Ensuring that speakers of all languages can interact with technology will let us understand and track outbreaks as early as possible. This works best when the broad language support is shipped as standard.”

Robert Munro
May 15, 2015