9th Biennial ACSPRI Social Science Methodology Conference

Aaron Willcox

Aaron Willcox is a dynamic and innovative Data Scientist at the Social Research Centre with extensive experience in research software development, data management, and data analysis. Specializing in R programming, data pipelines, and version control systems, Aaron brings a multidisciplinary approach to solving complex data challenges. His career is marked by a passion for leveraging data to drive research outcomes and inform strategic decision-making. Previously, Aaron was a Research Fellow in Data Science at the DARPA-funded repliCATS Project within the Interdisciplinary Meta-Research Group at Melbourne University. In this role, he coordinated and managed a complex research data pipeline, contributed to the development of research software, and conducted statistical analyses of qualitative and quantitative data from over 4000 psychological and social & behavioral science claims.


Session

Thursday 28th November 2024
12:00
15min
Mapping Civic Engagement: Methodologies and Insights from the Australian Civic Opportunities Index Project
Benjamin Phillips, Aaron Willcox

On behalf of the Scanlon Foundation Research Institute, the Civic Opportunities Index aimed to develop local government authority (LGA)-level estimates of civic opportunities available to Australians, adapting methodologies from de Vries et al.’s (2024) U.S. civic opportunities index. This index classifies organisation-level opportunities for events, membership, volunteering, and taking action. The project utilised a combination of data sources: the Australian Charities and Not-for-profits Commission Charity Register, Annual Information Statement Register, and programs register, as well web-scraped data from non-profit organisations’ websites. Organisational opportunities were predicted from the web-scraped corpus using various machine learning approaches (lasso, random forests, and XGboost). This presentation will explore the various methodologies employed in the project to detect and categorise civic activities. The methodologies include traditional web scraping techniques, the use of large language models (LLMs), and human coding. We also address potential shortcomings and limitations in the work, as well as areas of potential improvement.

Generative AI / LLMs in social research
Sutherland Room