37 Association of Research Libraries Research Library Issues 299 — 2019 the communities we serve.”20 This is echoed by Catherine Coleman in her assertion that librarians can be co-creators of “an intelligent information system that respects the sources, engages critical inquiry, fosters imagination, and supports human learning and knowledge creation.”21 There are numerous examples, such as Hamlet from MIT, the AI Lab at the University of Rhode Island, and the Stanford Library AI initiative, where machine learning in research libraries is occurring with an emphasis on explainability and accountability.22 Developments such as these highlight Chris Bourg’s 2017 suggestion that “we would be wise to start thinking now about machines and algorithms as a new kind of patron.”23 In doing so, research libraries need to consider not merely how the data can be exposed to algorithmic systems, but the new obligations with respect to data privacy and reuse that may come from this. These implications may extend beyond what is currently considered in research data management protocols. An illustration of why research libraries need to accelerate their involvement in AI and XAI arises from a recent breakthrough in the unsupervised text mining of the scientific literature, which demonstrated “that latent knowledge regarding future discoveries is to a large extent embedded in past publications.”24 This insight was observed previously during the formative years of Medline25 and has motivated the current “knowledge validation engine” of Project Aiur from Iris.ai.26 Each of these projects acknowledges that the structure of scientific communications (for example, the nature of abstracts) enables machine-learning analysis and highlights the need to verify the outcomes by examining the processes. They also emphasize the challenges of explainability when the research literature is being utilized and interpreted using complex and often opaque methods. Algorithmic decision-making is already pervasive in information tools and services acquired, provided, or developed by research libraries.