29 Association of Research Libraries Research Library Issues 299 — 2019 The use and development of machine learning applications in research libraries will only continue to grow in volume and influence. As AI reconfigures much of scholarly communications, it will be essential that libraries, and their users, have trust in the cognitive delegation of many tasks and processes. Mariarosaria Taddeo notes that “delegation without supervision characterises the presence of trust.”5 Approaching that state will require artificial intelligence to exhibit, and be open to, new levels of transparency and accountability. One critical element of that is explainability. Defining Explainable AI (XAI) The US Defense Advanced Research Projects Agency (DARPA) definition of XAI is widely referenced. The purpose of XAI is to enable human users “to understand, appropriately trust, and effectively manage the emerging generation of artificially intelligent partners.”6 To this user-centric XAI definition, DARPA adds the expectation that AI systems “will have the ability to explain their rationale, characterize their strengths and weaknesses, and convey an understanding of how they will behave in the future.”7 Examining these definitions yields both insights and complications. The user-centric definition has three key concepts: understanding, trust, and management. Understanding can mean a range of ideas from simple awareness or acceptance to acknowledgement and finally to detailed knowledge. While the idea of trust seems straightforward, the modifier “appropriately” suggests a conditional situation where the granting of trust is contextual. Managing indicates a relationship between the user and the AI and implies that the human user is, or should be, in a position of reasonable control. However, referencing AI systems as “partners” suggests a more cooperative and quasi-independent relationship. As AI reconfigures much of scholarly communications, it will be essential that libraries, and their users, have trust in the cognitive delegation of many tasks and processes.