Andrea Navarrete Rivera Has Worked on Three Data-for-good Projects
Andrea Navarrete Rivera is devoted to using data science for good. A student in the master’s program at the Data Science Institute, Navarrete has already worked on three significant data-for-good projects.
First, she did a UChicago Data Science for Social Good Fellowship, during which she helped build a machine-learning system that allows the Mexican government to respond more efficiently to online requests from citizens. Mexican citizens have the right to submit requests to their government, which by mandate must respond to them quickly. The Mexican government created an online submission system to handle the requests, but in the year following its launch system the number of requests doubled. With scare resources to manually process each request, the office handling them struggled to contend with the increased volume. Navarrete was part of a team that built the machine learning system to help the government respond to the requests faster and more efficiently.
The machine-learning system is intended to help at each stage of the requests. It can automatically accept or reject 85 percent of citizen requests; send 39 percent of them to a technical helpdesk; and automatically route 49 percent to the appropriate federal agency. Together, the system can allow the Mexican government to process 24.4 percent more requests, enabling officials working with limited resources to better address citizen needs. Navarrete’s team documented the success of its machine-learning model in a paper published in ACM Digital Library.
After her fellowship, she stayed on at the University of Chicago to work as a research assistant at its Center for Data Science and Public Policy. There, she belonged to a team that built a machine learning-based Early Intervention System (EIS) which identifies officers in the Charlotte Mecklenburg Police Department mostly likely to have negative interactions with the public. Such adverse police interactions with citizens include things such as unjustified use of force, civilian injury, or discourteous behavior. Many police departments use early intervention systems but they are inaccurate at predicting which officers need the most help. Navarrete’s team was the first to build a data-driven, machine learning-based Early Intervention System. By combining comprehensive data from various sources inside a police department (human resources, internal affairs, dispatches, arrests, stops, etc.), the Early Intervention System can identify which officers would benefit from interventions such as training, supervisor counseling, and psychological counseling. The system allows police departments to target limited resources on the officers most in need of help. Their data-driven Early Invertvne System system has proved so effective that it’s being used by several police departments across the United States.
“It makes me feel proud to have worked on these data-for-good projects,’ says Navarrete, who came to DSI from Instituto Tecnológico Autónomo de México in Mexico City. “The projects were interesting and made me realize the huge opportunities we have to apply data science for good. It definitely makes me feel good that different government agencies and organizations are using data science to have a better understanding of their problems as well as to generate more informed decisions.”
During the spring of 2019, Navarrete worked on yet a third data-for-good project, this time at Columbia University. She helped Professor Colleen Chen at the Columbia Law School on a study of diversity in technology. The specific aim of the study was to identify the gender and ethnicity of authors who publish articles in the area of artificial intelligence. For the study, the two generated a database of AI authors by scraping names from the Association for the Advancement of Artificial Intelligence Conference. They also did keyword searches in publications that report on AI areas such as machine learning, robotics and symbolic learning. Afterwards, using known gender attribution and ethnicity attribution methods, they were able to identify the author's name and ethnicity and gain a better understanding of which gender and ethnicity dominates AI publications.
This summer, Navarrete has a summer internship at Audible, in Newark, N.J, the audiobook company owned by Amazon. There, she’ll help design customer-acquisition models, using machine learning to identify potential customers of the company’s products on international markets and translating those models into useful insights for the marketing teams. It’s her first time working for a company, and she’s eager to see how data science works in the private sector. The internship will also help her decide what kind of job she’d like to have after she graduates in 2020 from the master’s program. But whatever area she decides to work in – nonprofit or private industry – she says she’ll always strive to use data science for the betterment of society.
“I’m devoted to democracy,” says Navarrete, “and I see a big opportunity to use of technology and data science to help overcome the gaps of knowledge that exist in our societies and to help governments become more responsive to their citizens.”
–By Robert Florida