A goat story
Touching down on the dusty ground in Dollo Ado, Ethiopia — by all accounts a desert — Rebeca Moreno Jimenez was eminently aware that her in-depth knowledge of killing computer bugs would not help her much with the flies and mosquitoes that inhabited the air. Instead, she would have to call upon her humanity while on the ground.
Moreno Jimenez, UNHCR Innovation Service’s resident Data Scientist, was there on a two-pronged mission. First, she was there to support Project Jetson, an Artificial Intelligence project that predicts in advance the number of displaced people that Dollo Ado camp would receive from Somalia. She was in Ethiopia on the hunt for variables that could be fed into some algorithms. Second, in doing so, she also hopes to start a more meaningful conversation about how the humanitarian sector views data science.
“It’s all about context,” she says. “Silicon Valley data scientists are far away from the reality and the ethical problems data [present] in an unequal world. Trying to explain the complexity of this world with a few models and some datasets are too limited in [the humanitarian] sector. [The humanitarian sector] is about humans, and sometimes irrational and rational human choices. ”
She says that visiting Ethiopia was critical in identifying the best variables to feed into the predictive analytics project. Naturally, these variables would have to relate to the sought-after metric of the number of displaced individuals, which is why Moreno Jimenez felt it was so essential to be on the ground, conducting qualitative interviews with the people themselves, to understand the variables that they personally found important when considering whether or not to migrate, and therefore the variables that should be more important for the artificial intelligence program.
She recounts her interaction with some of the people on the ground with awe and surprise, describing the questions she asked refugees, and how they responded. Dollo Ado is a rocky, copper-colored plot of land, spotted haphazardly with tents of different shapes and colors under a piercing blue sky painted white with clouds by the smothering humidity. Women walk among the loose stones with slow precision in flowing robes adorned with intricate patterns dyed in bright jewel tones. Leashed goats roam with their owners, squealing all at once, sounding like a corridor of rusty door hinges.
Amongst all this, Moreno Jimenez started her questions, as anyone would, with the basics — the numbers of people leaving their homes, the routes they were taking, the modes of transportation, and many other questions — but, like many radically course-changing moments in history, it wasn’t until she got to the topic of goats that she reached a watershed.
The people she was interviewing mentioned that, before heading on their journey, they had to sell their goats, since the delicate animals wouldn’t survive the arduous trek. She says that they explained to her that goats were the equivalent of assets to the pastoralists. “It’s considered like a bank savings account,” she says, so, selling the goats was a liquidation of their assets.
When she asked the community how many goats they sold before leaving Somalia, they answered that they had sold all of them. Surprised, she thought there may have been a misunderstanding, or perhaps a mistranslation. “Can you repeat the question?” she asked the interpreter, only to receive the same answer. Every last goat? She was shocked. With goats such a valued commodity, it seemed like selling your fleet of Italian supercars, simply because you were moving homes.
“They don’t survive,” she learned, “because [goats are] very delicate, compared to camels, maybe, or cows… It hit us that this commodity was what, in data science, we call a ‘nesting behavior.’ A nesting behavior is not the predictor, nor is it the actual reason for fleeing, but it’s something that could give you an indication that something is about to happen.” These nesting behaviors are indirect, but reflective markers of another variable that could lead to something surprising or new. And while the operation had been monitoring these variables — it was only in this moment that Moreno Jimenez realized it could support the algorithm that supported Project Jetson.
She immediately got in touch with Food and Agriculture Organization (FAO) team in Somalia who could give her access to the price of goats in the region, realizing that, if many people are selling goats at once, there would be a flood in the goat market, and by the principles of supply and demand, the prices would fall. “If in the region, there is selling of goats, the prices are going to drop, especially in a market like Somalia with minimal or almost non-existent government intervention.” In other words, drops in the prices of goats would indicate an impending wave of migration, as migrants rapidly sell their goats before embarking.
Indeed, when she analyzed the data later on, she found that, when one region felt a drop in the price of goats, a few months later, neighboring regions would experience a sudden spike in the price, attributed to internally displaced people buying back goats to replace the ones they sold before leaving. “Like you [would] open and close a bank account if you were to move from Europe to America, or America to Europe,” she says, reinforcing the banking analogy.
Without being on the ground, Moreno Jimenez might never have caught onto this important parameter. “I’m always very thrilled to be on the ground, because I am able to see and counter myself in a lot of things that I think, as a researcher, make sense, because I’m seeing data.” But, all data, she says is “biased, because of the way you collect the data, because of the way you ask your questions…Maybe you have your own bias, but, when you go on the ground, and see what’s happening, some of that bias changes…You always come [back] with more questions.”
Her willingness to part from the data reflects what, in humanitarian work, represents a fairly novel attitude. With data science being only a nascent facet of humanitarian work, there exists the risk of it aping the trajectory of data science in other fields — often eschewing the people (and, consequently, their needs and insights) in exchange for numbers and technology.
Miguel Luengo-Oroz, Chief Data Scientist at UN Global Pulse, recently articulated the need among humanitarian organizations for “Data translators who can both understand the operational humanitarian contexts and have data intuition. They know what can and cannot be done with data and how to interpret and visualise data and algorithms to provide information for real impact.”
Perhaps it is because Moreno Jimenez challenges the stereotypes around Data Scientists that she is such an adept data translator, and takes an approach that remains novel. She often calls upon insights that she has learned throughout her life as she leads the continued understanding of how data could assist in making UNHCR a more proactive organization and give predictions on population movement, a uniqueness that is a testament to the power of appropriately-used data.
Meet the Jetson
Project Jetson was born out of this shift towards proactive decision-making that begins taking shape in a small interdisciplinary team that was comprised to form the European Winter Cell at the end of 2015. It was during this time that UNHCR was increasingly concerned about adverse effects for refugees regarding the scale of population flows from Turkey through to Europe during an uncharacteristically cold winter. The unique challenges faced by the organization during 2015 lent themselves to the opportunity to steer UNHCR towards becoming a more data-driven organization.
The Winter Cell was geared towards facilitating proactive decision-making, rather than reactive. Among other things, the Winter Cell’s work led to the idea to explore the ability to predict weather patterns against the movement of people. The interdisciplinary nature of the team allowed for the incorporation of a wide range of expertise and the opportunity to build on this knowledge through partnerships when the expertise was not found in-house.
This approach was built around a broad coalition including academia, the private sector, national meteorological services, among others, to bring together further data points for their predictions. This eventually led to the development of a collaborative model that would give foresight on population flow trends, laying the groundwork for future predictive analytics work at UNHCR.
The Winter Cell’s blueprint for collaboration based on needs, expertise, and the use of predictive models, rather than the standard scenario-building tools, was replicated in the team that would form Project Jetson. Project Jetson was built on a request from UNHCR’s team in Dollo Ado in 2017 after they noticed similar trends in data points that led to the 2011 Somali food security crisis. The data was initially indicating that the circumstances preceding the 2011 crisis were recurring, causing concern that another refugee crisis would be precipitated.
The team in Dollo Ado initially requested assistance in scenario-building, but by this time, former members of the Winter Cell had re-joined the Innovation Service, and their experiences had already proven that the right data and partnerships could lead to a new proactive and predictive methodology. The request would then take Moreno Jimenez to investigate the most appropriate use of the data to produce an effective machine learning algorithm.
Project Jetson is a new path for the humanitarian sector, says Moreno Jimenez. “The prediction capabilities of a lot of humanitarian agencies are very new, so, what we were trying to do was to try to understand not only patterns of displacement, but, if there were some significant data points that could give us predictions of how [refugees and internally displaced people] were going to move in the future. That’s something that is an unexplored area in the [humanitarian] sector, since we are very strong in emergency response, but maybe not that much in planning as other sectors are.” The ambition of this initiative, she says, is “to increase the capability of UNHCR, in this case, to be able to predict and prepare better.”
As she considers the immense difficulty on the ground, she reinforces the importance of the work. “People are so resilient. I mean you can see the data sets but it’s hard to get out of just focusing on statistics. But in person, you truly understand their resilience — not by seeing the numbers, but by seeing the people. Part of what we preach is human-centered design, and data science is not an exception to that. You need to validate your assumptions by being on the ground and actually ask more questions.”
We’re always looking for great stories, ideas, and opinions on innovations that are led by or create impact for refugees. If you have one to share with us send us an email at innovation@unhcr.org.
If you’d like to repost this article on your website, please see our reposting policy.
