Data for Good

When Ali Arab graduated with an applied mathematics undergraduate degree in Iran, he made a tough decision to leave his home country in pursuit of more education and a life where he could use his understanding of math to solve practical problems.

“Because of the political situation in Iran, the prospects of a person in graduate studies for someone like me was not very good,” he said.

He ultimately landed at Southern Illinois University Edwardsville for his master’s degree, and later the University of Missouri for his doctorate. Now, as an associate professor of statistics and director of graduate studies in the Department of Mathematics and Statistics in Georgetown’s College of Arts & Sciences, Arab is blending the personal and professional for good.

The challenge is creating statistical models of processes from sparse data to better understand our world. For example, to visualize the impact of climate change on migratory birds, he modeled data that was collected by citizen scientists before and during periods of significant climate change, such as the timing of their nesting and how long they spend in different stops during their migration.

“Understanding behavioral changes in animal species helps us better understand climate change, both from the conservation management perspective for bird populations, but also as an indicator of how things may be shifting in ways that are impacting certain ecological processes,” he said.

He’s also exploring how to create statistical models that can help predict changes in forced displacement patterns of humans. Together with Professor of Computer Science and Public Policy Lisa Singh and Donald G. Herzberg Professor of International Migration Katharine Donato, Arab received a Sonneborn Interdisciplinary Collaboration Chair, a three-year award that promotes interdisciplinary collaboration among Georgetown faculty, to explore the possibility of creating a statistical framework that predicts forced human migration.

“When you are in the early stages of a crisis that will potentially trigger forced migration of people from an area — this can be climate- or conflict-driven — there are not a lot of immediate sources of data as data collection takes time,” he said. “We’re looking at ways that combine the conventional sources of data with other sources of organic data.”

Organic data (a term coined by Interim President Bob Groves), according to Arab, is often unconventional in science. It may include conversations on social media, reports in newspapers, satellite imagery, search engine data and more.

“Sometimes there’s a need to get a forecast or an understanding of the next couple stages of a process,” said Arab, using the COVID-19 pandemic as an example. “You can’t always afford to wait for the conventional sources of data. But it comes with a downside; they’re only proxies. There are signals you can extract, but there’s a lot of noise in there, too.”

As recipients of a National Science Foundation grant, Arab and his collaborators are exploring Bayesian transfer learning — a burgeoning method that leverages machine learning — to create predictive models that span time and regions using prior knowledge and existing data sets to help understand current issues that have sparse data.

Arab describes transfer learning this way: Say someone is an experienced ping-pong player, and they want to learn tennis. Some of the knowledge and skills of ping pong will translate to tennis, but some — such as technique — will not and may even have a negative impact. With Bayesian transfer learning, though, modelers can extract knowledge from related data sets and incorporate them into their new models.

“It’s the idea of, can I learn from similar situations and transfer that learning to a new situation, understanding the differences?” Arab said. “For example, what we have learned about the forced migration in past cases — such as the forced migration of Ukrainians by Russian forces — may help us forecast displacement patterns in a future crisis.”

All of this work is for what Arab calls “science as a human right.” Outside of his role as a faculty member and researcher, Arab is committed to championing human rights. He serves as a representative of the American Statistical Association to the American Association for the Advancement of Science and Human Rights Coalition; previously served as a member of the board of directors of Amnesty International USA; and is a founding board member of Hostage Aid Worldwide, an organization that uses data-driven solutions to campaign for the freedom of hostages around the world. He is also actively involved in data-driven advocacy efforts of hostages, including freelance journalist and fellow Hoya Austin Tice (SFS’02, L’13), who was kidnapped while reporting in Syria on August 13, 2012.

“I have a personal tie to many of the projects I work on. For example, as an immigrant, the topic of migration resonates with me,” he said. “I have experienced living under an oppressive regime in Iran, which has deepened my sensitivity to social justice and human rights. I’ve had an opportunity to align my academic research with the sort of personal values that are important for me. I think, too, this is very important, especially now with the change in climate and the political unrest. I always liked the idea of using my mathematical and statistical background in order to answer practical questions.”

Moving forward, Arab predicts that data science and artificial intelligence are going to play an important role in scientific discovery, more than ever before.

“I think this will only effectively happen if we combine scientific knowledge with data,” he said. “I don’t want to ignore hundreds of years of science when I’m working on, for example, models of spread that have foundations in physics and biology. I’m interested in developing statistical models that not only benefit from different sources of data in smart ways but also draw on science-based modeling concepts and ideas. My hope is that the current AI revolution moves in that direction, so we have more effective scientific discovery, which, I think, can lead to some exponential growth in understanding nature, life and medical discoveries.”