Eunhyeah is a PhD student at USC Dworak-Peck School of Social Work. Although she majored in business, a non-computer science major, her passion for eradicating child maltreatment led her to learn how to harness data. Her dissertation topic is to use machine learning to understand patterns of child maltreatment and to devise preventative measures to create better environments for children.
- What are your undergraduate and graduate majors?
I majored in business in my undergrad and worked in a tech company as a marketer for two years. After leaving that job, I decided to do a Masters in Social Work (MSW) in Australia hoping to work for children in need, and now I am pursuing a PhD in Social Work at USC.
- What was your turning point (event, person, or work) that motivated you to study data science?
In my first year in MSW, I started thinking about using data analytics to inform policy, knowing that data had been already widely used in the private sector for profit. I learned that data are kept in silos and not being effectively used in the social sector, so I decided to study data science. In order to manage and use data in the social sector more efficiently, I believe it is important for social workers to harness data science.
- Have you worked in the field of data science (either a work or a research experience)? Please pick one work experience that you enjoyed the most and explain it in detail.
Over the summer, I was fortunate to participate in the Data Science for Social Good Fellowship, held at Imperial College in London. Our team partnered with Memphis Fire Department (MFD) to predict individuals who are at high risk of becoming high utilizers of emergency medical services (EMS) for proactive interventions. We used incident level data provided by the City of Memphis Fire Department from 2016 to 2019, which includes patient demographics, medical history, service engagement history, and incident records. Using the data, we trained a few binary classification models to predict individual level risk of becoming high utilizers. The best performing model was Random Forest producing a list of 20 individuals predicted to be at high risk with feature importance. After validation trials, our model will be integrated in current workflow to predict high utilizers so that they can be prioritized for preventative care. We hope this can free up EMS resources for high level emergencies.
- Looking back to the beginning of your journey, do you have any advice for students or beginners who want to learn more about data science?
I would like to say that yes, it is possible to learn data science without having a background, and no, it is never too late. When I decided to study data science as a social work major, I was not sure whether I could learn computer programming as a non-computer science major. I reached out to computer science professors who were interested in social science to ask about my plan and also participated in a Datathon event just to see how I could contribute to the field of data science as a social worker. After some research, I learned that domain knowledge is very valuable in data science, and if you have clear goals and understanding of your field, learning data science is a huge advantage.
- How will you apply your skills to solve real-world problems? Why do you care about solving this problem?
I hope to use my social work domain knowledge and data science skills to answer issues around maltreated children and their families. I am interested in addressing questions of equity among under-surveilled and under-served sub-population and their life-long cross-sector service trajectories. To respond to these questions, rigorous ethical questions must be asked about algorithmic accountability and data bias. These ethical questions should not remain as abstract ideas but have to be delivered in actionable ways to ensure that people behind the data can have a better understanding and control over their own data. To do so, I think a community participatory framework can be one of the effective approaches. A community participatory framework guides us to enable people to collectively build an algorithmic policy for their own communities based on social-choice. WeBuildAI by Lee and colleagues (2019) is an example.
- Why do more people need to study data science?
I think it helps us to better understand society. Understanding how data are collected, stored, managed, and used around us is critical in all domains.