The US Census Bureau xD team, in collaboration with the 10x program, is building a library of machine learning and artificial intelligence resources to help combat bias introduced by bias and algorithms. We aim to create toolkits to diverse audiences encompassing both technical and non-technical users.
WHY: AI and machine learning-based tools and techniques are quickly being adopted and deployed across governments at all levels. The rapid pace of adoptions raise ethical concerns of using these tools in civic settings, particularly when they could be invoked to make high-stakes decisions. Invisible quality issues in datasets, biased data collection methods, bad data governance problems, and issues in model development have the potential to introduce enormous bias into decision-making processes that adversely and systematically affect vulnerable populations served by government programs
WHAT: This collection currently includes a syllabus for beginners, executable Jupyter Notebooks, an annotated list of resources, a list of books used to build these resources and other prototpye components including a Question and Answer flow used to guide users toward the appropriate resource list.
A full release is planned for 2021.