Our cover image is “Machine Learning” by Ben Davis from NounProject.com
As libraries, archives, and museums shift from conceptual awareness to operational use of data science, machine learning, and artificial intelligence (AI), they require domain-specific resources to make the transition. Readily-available, off-the-shelf workflows, applications, and models built by big tech will often perform poorly with cultural heritage and historical data. Beyond the issue of accuracy, they pose specific dangers when applied to culturally-sensitive collections, with potential for amplifying harms at scale. Domain-specific resources are crucial for operational use that supports library, archive, and museum aspirations to increase equity for the communities they serve. From 2019-2020 a series of machine learning and AI cultural heritage research agendas, state of field analyses, and reports were produced by Thomas Padilla, Elizabeth Lorang, Ryan Cordell, Oonagh Murphy, and Andrew Cox, among others.
These resources share in common foundational focus on equity and justice in computational work. Recommendations from this body of work span nearly all aspects of library and cultural heritage organization activity, including description and discovery, public services, collection development, infrastructure development, and sustaining interdisciplinary and interprofessional collaboration. This publication aims to build on that body of work, advancing from high level recommendations to the creation of concrete resources that support responsible operational use of data science, machine learning, and AI in libraries, archives, and museums. This resource will be a primary asset to libraries, archives, and museum practitioners as they seek to responsibly operationalize data science, machine learning, and AI.