The Integration of Machine Learning into Research Process: Methodological Reconfigurations in Social Science-Driven Disaster Research

This project explores how disaster researchers are using machine learning (ML) and what that actually changes in disaster research. It starts by collecting a broad set of published studies where ML is used to study disasters and climate risks such as floods, heatwaves, wildfires, and storms. For each study, the project records what kinds of data are used (for example, surveys, interviews, satellite images, sensor networks, administrative records, or social media), what types of ML models are applied, and what roles these models play in the research process, from data cleaning and feature extraction to prediction, classification, and scenario analysis. The project then looks closely at how authors talk about ML in their papers: the reasons they give for choosing ML, the benefits they highlight (such as speed, scale, flexibility, or better performance), the challenges and risks they note (such as bias, opacity, data gaps, or limited interpretability), and how they position ML in relation to more traditional qualitative and quantitative methods used in disaster social science. By combining these technical and narrative perspectives, the project builds a clear and accessible picture of how ML is reshaping disaster research, identifies areas where current practices may be improved.

Faculty Supervisor:

Haorui Wu

Student:

Partner:

Hanyang University

Discipline:

Sociology

Sector:

Sustainability & the Environment

University:

Dalhousie University

Program: