Public Policy Ph.D. Students Use Machine Learning to Solve Sustainability Problems in Hackathon

Primary tabs

A group of Ph.D. students in the School of Public Policy participated in the Organization for Economic Co-operation and Development’s (OECD) June 2022 hackathon. The international event featured teams from seven universities, who worked to produce policy decisions based in data science.

The Georgia Tech team consisted of five public policy Ph.D. students: Vincent Xinyi Gu, Yifan Liu, Daniel Marchetto, and Sergio Pelaez, and Matteo Zullo. Professor Philip Shapira connected the School of Public Policy to the event, and Assistant Professor Omar Asensio helped field the team of students. OECD presented them with the question, “To what extent are countries' green transition goals, as set out in their strategies, reflected in their science, technology, and innovation (STI) policies?”

“The hackathon was an opportunity for us to understand real-world problems and to have a group of scholars thinking about solving these problems with advanced and updated methodologies,” Pelaez said.

Other hackathon teams explored such topics as goals and budgetary commitments outlined in policy strategies, policies that encouraged publicly-funded and open-access research data, and policies that focus on responsible use of emerging technologies.

The OECD provided the team of Yellow Jackets with 313 STI strategies from OECD countries, as well as another dataset with over 9,000 policy initiatives from the same nations. They used a neural network (a subset of machine learning at the heart of deep learning algorithms) to classify the sentences in each document as either environment-related or not.

The team then analyzed the sentences related to the environment so they could measure the extent to which environmental topics were reflected between the two datasets. They found little evidence that green transition strategies were reflected in and actually applied to green transition policy initiatives.

Paleaz explains that by highlighting the disconnect between environmentally-friendly intentions and actual policy implementation, the Georgia Tech team’s results can help governments explore why this is the case.

After presenting their findings, the OECD expressed interest in continuing to work with the team from Georgia Tech and dive deeper into how they approached solving the problem with machine learning.

“It was valuable to get to know the questions that policymakers seek to answer,” Pelaez said. “It was also a huge help to have the opportunity to work with big data from actual policy strategies and initiatives around the world.”


  • Workflow Status:Published
  • Created By:gwyner3
  • Created:06/23/2022
  • Modified By:gwyner3
  • Modified:06/23/2022


  • No categories were selected.