Bias^2 Awards

These awards are associated with the HDSI Bias^2 Program.

Data-Driven Methods for Understanding and Overcoming Biases in STEM Education
Flavio Calmon

The project combines methods from social psychology, machine learning, and information theory to create algorithmic tools that monitor student, teacher, and school-level data for factors that impact students' engagement in STEM.

Examining Covid19 disparities in testing and outcomes in Holyoke, MA
Louise Ivers

In close collaboration with The City of Holyoke, Massachusetts, we aim to assess disparities in SARS-CoV-2 (COVID-19) testing rates and outcomes of COVID-19 by race/ethnicity in the city. We also aim to elucidate community-level factors that mediate the relationship between race/ethnicity and hospitalization and death among persons with confirmed SARS-CoV-2 infection. 

Data Nutrition Project
James Mickens

Machine learning algorithms are trained on data. Unfortunately, that data is often problematic in multiple ways. For example, some labels may be incorrect or missing, and individual data points may have been drawn from a biased sample. The Dataset Nutrition Label tool (hereafter called “the Label”) is a framework that makes dataset deficiencies explicit, in a way that is both understandable to humans and queryable by automated tools. The Label makes it easier for data scientists to understand whether a particular data set is appropriate for a particular ML use case.

Identifying bias in receipt and access to optimal care for endometrial cancer patients in Massachusetts
Brianna Joy Stephenson

This study aims to understand the roles race and socioeconomic status have on those who receive optimal treatment, and the facilities that are referred for treatment. By identifying the roles of receipt and access to optimal treatment and how they contribute to EC disparities, we will be able to identify areas for improvement in the Massachusetts healthcare system and achieve health equity in endometrial cancer patients.