Can more accurate models of famine, supported by historical data and the most up-to-date tracking tools, lead to better predictions and save more lives?
Creating these tools is the focus of a series of data hackathons hosted by the Feinstein International Center, which is housed at the at the Friedman School of Nutrition Science and Policy at Tufts University, and the Friedman School's Division of Nutrition Epidemiology and Data Sciences. Part of the new effort involves deepening the dialogue around data quality, availability, accessibility, and completeness and eliminate bias, distortion, and mistrust.
The first two hackathons took place from February 10-12 and March 10-12 on Tufts' Medford / Somerville campus, while the third is scheduled for May 25. The series has been led by Elena Naumova, a data scientist and professor at the Friedman School, and Paul Howe, Irwin H. Rosenberg Professor of Nutrition and Human Security, professor of the practice, and director of the Feinstein International Center.
"What we're trying to do is look at famine as a complex system, how it forms, how it evolves, and how it collapses - a little bit like a hurricane but with the added challenge of factoring in human decision-making," Howe said.
The sooner a potential famine is identified, the sooner nations and NGOs can take action to avert its worst consequences. "There's a hope that if we understand how a famine forms, we might be able to gain new insight into the signals that one is developing and identify points of leverage within that complex system that allow us to 'rebalance' dynamics, and prevent mortality before it occurs," Howe said.
As a result of the "three C's" - climate, conflict, and the ongoing effects of COVID-19 - organizations like the World Food Programme (WFP) have predicted an unprecedented rise in global hunger in 2023. The degree to which individual countries experience hunger varies, with humanitarian catastrophe/famine ranked as the most dire. Currently, more than 900,000 people struggle under "famine-like" conditions, the WFP says.
The Integrated Food Security Phase Classification, or IPC, defines famine as extreme food deprivation likely to lead to mass starvation and death. To prevent this worst-case scenario, famines must be predicted, or forecast, the first step toward alerting the international community to respond with the proper amount of aid. While sophisticated early warning systems such as FEWSNET are well established and provide updated monitoring of contexts across the globe, the hope is that better models of the dynamics of famine formation will further enhance and support these efforts.
Supported by an intramural grant program sponsored by the Office of the Provost, the Office of the Vice Provost for Research, and Tufts School of Medicine, the May hackathon will gather data enthusiasts from across campus; students, staff, and alumni working remotely; and participants with and without data analysis experience.
"A typical hackathon assumes people will come once and they will get the task done, and the groups don't truly meet again," said Naumova. "What's new about our idea is that we make it a systematic and iterative process, so teams can learn from and build on their previous experience, with the understanding that not everything is possible in a single session."
Teams will also share the questions they see emerging from the data that they are most interested in exploring. "It's a challenge, but it's also an incredible opportunity to generate new ideas and insights," Howe said.
One of the broader goals of the hackathon series is to create a reproducible model, said Kyle Monahan, a senior data specialist at Tufts and a key organizer of the hackathons. "The aim would be that anyone across the world could go to our site, pull down our content, and reproduce a hackathon on a different topic. And we would provide a framework to create a hackathon as a research tool, a tool for enabling students from many walks of life and many different levels and backgrounds," Monahan said.
Hackathons are just one tool in the researchers' broader effort to develop accurate datasets that support upcoming research, with the hope that better data, new models, and eventually improved forecasting could contribute to saving people from catastrophic hunger in the future.
It's common to underestimate the true value of famine forecasting, Naumova said, because when a full-on food crisis is averted, initial alarms appear unfounded. But we should not measure the success of a forecast based on whether or not an event occurs, she cautioned.
"If people and decision makers are listening and acting, and putting all their efforts together, ideally a famine is mitigated," Naumova said. "I don't know that we have that mentality yet, but my belief is that we should change the general attitude toward projections, and recognize that we need to be more thoughtful in producing forecasts."
Above all, Naumova stressed, those forecasts need to be both clear and actionable. "There are a lot of powerful models out there," she said. "The question is how we act on them."
