Ethical Data Science and AI
MIDAS Leads Multi-University Collaboration on Data Equity Systems
A team led by MIDAS Director, H.V. Jagadish, is establishing an information system to minimize the misuse and misinterpretation of data. Dr. Jagadish and MIDAS affiliated faculty, Tayo Fabusuyi, Robert Hampshire and Margaret Levenstein, along with colleagues at other universities, have completed initial work on methods to identify data and software bias that negatively impact equity in critical domains, such as mobility, housing, education, economic indicators and government transparency. The ultimate goal is to develop a system that will take potentially biased data from both public and private sources, and make them bias-adjusted and analysis-ready. This project will establish a Framework for Integrative Data Equity Systems (FIDES) for the study of systems that enable research on sensitive data while preventing misuse and misinterpretation.
Fair Representation In Arts and In Data
”Fair Representation in Arts and Data” is a year-long collaboration between data scientists, artists and museum curators that was funded by the President’s Arts Initiative. The project analyzed the collection at UMMA using two widely used face detection algorithms to reveal biases in the algorithms and in museum artwork acquisition practices. Museum visitors can now get a first glimpse at the initial research findings through “White Cube, Black Box”, a display at the “YOU ARE HERE” exhibit at UMMA.
Reproducible Research
MIDAS Promotes Reproducible Data Science
A significant challenge across scientific fields is the reproducibility of research results, in both the narrow sense of reproducing the results with the same data and analytical methods, and in the broader sense of making the production of scientific knowledge transparent, traceable, and trustworthy. Data science research with increasingly complex data and long project pipelines faces reproducibility challenges every step of the way. MIDAS organized the 2020 Reproducibility Challenge to highlight high-quality reproducible work through examples of best practices across diverse fields and built an online resource collection. We now focus more intensely on actionable solutions that can be shared with other researchers to improve reproducibility. Our goal is to facilitate the development, validation and dissemination of tools and training to help make data science research more reproducible across many application domains.
MIDAS will host a “Reproducibility Day” celebrating the 2021 challenge in early 2022. Learn more about the 2021 Reproducibility Challenge here. For tools and resources relating to reproducible data science and research please visit the MIDAS Reproducibility Hub.