How do you count the dead during a bloody and destructive conflict? A Duke statistician may have some answers.

Current efforts to determine the death toll in Syria’s ongoing civil war are hampered by conflicting body counts and ambiguous levels of uncertainty. Beka Steorts, assistant professor of statistical science, has been working to develop statistical methods that can more accurately count the dead and paint a better picture of the devastation in Syria as a result. Her research lab on campus, which involves Duke graduate and undergraduate students, is investigating these death counts in collaboration with the Human Rights Data Analysis Group and other researchers across the country by analyzing hundreds of thousands of death records in multiple databases.

“The goal is to be able to come up with the death count in Syria and to understand the notion of uncertainty around the death count,” she said.

Analyzing death count data is often difficult because one person may be documented in multiple databases in slightly different ways, leading to duplicated records, Steorts explained. For instance, typos and inconsistent spellings could result in several entries for the same individual.

This problem is further complicated by the fact that the Syrian death record database—which consists of about 300,000 death records—only has reliable information regarding the full name, date of death and location of death for each person. Even this information is often missing or inaccurate, Steorts added.

“There should be some sort of standard error, and we would like to be able to quantify that uncertainty,” she said.

To achieve this goal, Steorts and the other researchers created an algorithm that makes it easer to identify the likelihood that two records in the Syrian databases actually represent the same person by reducing the number of record comparisons needed. She noted that her algorithm reaches 99 percent accuracy and can comb through the data in just 10 minutes—whereas other methods commonly feature 50 to 60 percent accuracy and take days or weeks to run.

Steorts noted that these statistical estimates are beneficial because they can assist human rights groups in their work.

“We’re trying to provide justice for a group of people who can no longer fight for justice themselves,” she said.

Steve Fienberg, a professor of statistics and social science at Carnegie Mellon who introduced Steorts to the project, said the data may be used in future war crime trials after the conflict is over.

“It’s important for the world to understand the devastation of this conflict,” he said.

The project was not without its challenges, Steorts noted. Due to the sparse nature of the data sets, finding a statistical method that provides high accuracy was a long and difficult process.

Steorts was recently selected by the MIT Technology Review as one of their “Innovators Under 35”—an award for individuals who show potential in making scientific breakthroughs in their fields.

“[The project] gives a voice to those who have been killed,” she said. “It’s rewarding to work on an applied topic that matters.”