47 Association of Research Libraries Research Library Issues 299 — 2019 Research Librarians as Guides and Navigators for AI Policies at Universities Geneva Henry, Dean of Libraries and Academic Innovation, The George Washington University Introduction Artificial intelligence (AI) is a term that is increasingly a part of daily conversations and is being discussed in many different contexts. Commercial applications of the various AI technologies (for example, natural language processing, machine learning, predictive analytics, robotics)1 are becoming part of mainstream society without people realizing that AI is at work. Searching the internet using popular search engines, for example, can employ deep learning algorithms that continually learn from previous searches. If the same or a very similar search is performed many times, with users consistently selecting the third-ranked return, the search engine will know that the ranking priority should be adjusted so that the most frequently selected result receives a higher ranking.2 Users generally do not think about how search results are returned they’re just happy to find what it is they’re searching for on the first page of the results without having to sift through the 1,000,000+ possible matches that were returned. Even if someone did want to understand how the search results were prioritized, the proprietary nature of commercial products that are using AI to have a competitive advantage in the marketplace makes it impossible to inspect the software behind the decision-making process. The end-user experience of using AI-enabled products—from search engines, to self-driving cars, to vacuum cleaners that do our housework for us—can be pleasant, but it can also be deceptive. Without visibility into the algorithms that were programmed into the systems by the software developers, the training data sets that were used to enable the algorithms to build a knowledge base, and the ongoing self- improvement processes that drive the decision-making based on