53 Association of Research Libraries Research Library Issues 299 — 2019 that should be permitted and encouraged vs. those that should be managed.20 A Role for Research Librarians in AI in University Research In a recent survey that examined publications from 21 leading scientific conferences in the field of AI in 2018, only 18% of the authors were women. These researchers mostly have PhDs and represent the research underway in AI throughout the world. The US continues to graduate PhDs whose publication rates dominate other countries, with 44% of the 2018 publications produced by scholars who earned their PhD in the US, followed by China at 11%, the United Kingdom at 6%, Germany at 5%, and Canada, France, and Spain each at 4%. Furthermore, the survey also found that the AI talent pool is very mobile, with approximately one third of the researchers employed outside of the country where they received their PhD.21 When we look at faculty in AI, both tenured and non-tenured, African American representation is only 1.7%.22 This lack of diversity in the university research population is troubling when algorithms and training data sets are being developed and selected by a mostly homogeneous group of primarily white men. Given the widespread lack of diversity among this population, the normal research review process of peer review will draw from this same population, further exacerbating issues that may be present, such as unconscious bias and data training sets that may be skewed to unfairly represent certain populations. Research emerging from universities can make its way into industry products and visibility into the algorithms and data training sets can be hidden from the end users. Universities have a responsibility to ensure that the research emerging from the institution is of the highest integrity. Having policies that require accountability of algorithms, When we look at faculty in AI, both tenured and non- tenured, African American representation is only 1.7%.