Dons' Data Goes Live on ABC
In the run-up to the 2018 midterm elections, ABC News approached the USF Data Institute with a challenge: Could the USF team of students and faculty develop new strategies to help the ABC News Decision Desk make quicker predictions about specific races?
The Data Institute team came up with a plan for election night, and their approach was so successful that the Data Institute has been asked back to work with ABC News for the 2020 presidential primaries and general election.
“This is a testament to the quality of the program and also a testament to the focus of the program,” said Nick Ross, assistant professor of data science. “A lot of data science is either focused too vocationally or too academically, but we’re striking a balance between those two, where students have the data science skills they need but also have the theoretical underpinning to those skills.”
Faster election projections
On election night for the 2018 midterm elections, the USF team worked closely with the ABC News Decision Desk, a group of experts who analyze live election results to determine when they can project particular races. Traditionally, the journalists depend upon teams of people stationed at precincts and counties around the country calling in with results as they’re released. Those numbers are then worked into spreadsheets and other programs.
But the USF team focused on the data engineering side of things, writing code for a program that would automatically pull election night results as they came in precinct-by-precinct in specific states. They built and tested an online dashboard where all the data could be aggregated, viewed, and run through data models to help make accurate projections. Students got a crash course in some data tools and techniques they weren’t officially supposed to learn in class until the following semester.
“This was a very real-world experience in which our students had to apply what we were teaching them in real time to a high-stakes, high-pressure environment,” Ross said.
Cutting-edge data science
On election night, some members of the team were in New York at ABC News headquarters, working directly with the ABC News Decision Desk to provide information from the dashboards — just one piece of information the journalists took into consideration before making an official race projection on live TV. Decisions on when to project a race, said Data Institute Executive Director David Uminsky, who was in New York, are always based on conversations that involve multiple sources of information.
In San Francisco, the rest of the team focused on making sure the code was working properly and, when it wasn’t, fixing errors on the fly. Sometimes, they were called on to write entirely new code in a fast-paced, deadline-driven environment. They were in constant communication with the team members in New York.
“The whole experience gave me such a great hands-on learning experience and what was really a once-in-a-lifetime experience,” said Nicole Kacirek MSDS ’19. “Some of the technical skills I learned certainly helped me in my practicum project, but beyond that I learned how to work on a project that’s fast-paced, where teamwork and collaboration are key.”
Going forward, other news organizations are now also beginning to use new data science approaches in their election night coverage, Uminsky said.
“USF,” said Ross, “is doing cutting-edge data science work on a national level.”