Abstract
Both research and practice rely on data for advancement, which elevates concerns around ethics and equity in technological advances and research implications. Previous efforts have focused more on improving methodologies than confronting social systems and processes that lead to inequity in technological outcomes. Furthermore, few have studied how to support people with overlapping social identities through the social science framework of intersectionality in data science.To this end, we propose a toolkit and prototype tool for supporting explicit consideration of intersectionality in data-driven processes. With our prototype workflow support application, Quintessence, data scientists can integrate intersectional reflection into the data science process, from pipeline design to modeling and interpretation, with reflection questions at each stage. We also introduce a Python library, QuinterPy, that calculates and visualizes the representation of overlapping, user-selected categories in a dataset. The library also can calculate the change in the representation of these categories before and after a processing step, e.g., before and after data cleaning. The Python library can be used on its own or as part of the application.In the future, we plan to evaluate the workflow support application and library with practicing data scientists to determine which aspects data scientists consider most useful and how to best incorporate intersectionality into data science practice.
How to Cite:
, J., , A. & , B.,
(2022) “Quintessence: A Quantitative Intersectional Data Analysis Support Framework”,
Journal of Student-Scientists' Research 3.
doi: https://doi.org/10.13021/jssr2021.3220
Files
Downloads are not available for this article.