ML4STS lab#
We study how to do Machine Learning for Socio-technical Systems. This means we apply data science using ML models to understand the world with other scientists, study and evaluate ML systems directly, and build tools to help data scientists to employ best practices. Most of our work right now focuses on fairness of automated decision making systems and some includes collaboration with other scientists.
News#
See the news page for a more complete list.
2024-10-30 : Funding from the Survival and Flourishing Fund
We received funding from the SFF Fairness Track to fund work developing a benchmark for fair data-driven decision making with LLMs. We will develop a benchmark for both data analytics style decision-making and training fair models on data. Then we will assess state of the art AI systems engaging in these tasks across various levels of AI agency.
2024-10-08 : Dr. Brown is speaking in the Digital Defense Fund fall seminar series
She is co-leading a session on AI basics titled, “Unmasking AI: the Mysteries and Magic of Artificial Intelligence”.
2024-08-23 : Dr. Brown is presenting at NSBE PDC
NSBE’s Professional Development conference is a venue for working professionals to skill up. She is presenting the building a portfolio website workshop by invitation
2024-03-29 : Dr Brown will participate in the AI lab Fireside chat series
title How to anticipate risks in AI-enhanced research
abstract ML-powered AI is increasingly appearing in all areas of research. However, ML is still in many ways a research output, more of a prototype than a polished, reliable, product. Even commercially released systems, because they are software, can be released without the same rigorous testing that commercially available research equipment has been traditionally be subject to. In this talk, I will draw connections between the ML research process and potential risks in applied ML and use of commercial AI in research.
2024-03-22 : Dr. Brown will present a workshop at NSBE 50
The workshop will be a hands on workshop to build a profile website using github pages.