54 Association of Research Libraries Research Library Issues 299 — 2019 showing not just the algorithm itself and the process followed when using data, but an explanation of the extent to which the data used had an influence on the decision outcome.23 Data sets that are used to train the systems must also be open to inspection to uncover potential biases and lack of true representation. Research librarians who are part of AI research teams can be sensitive to the need for well-documented and open systems. Librarians are likely to be aware of how other existing policies will influence outcomes. Librarians working in this space will need a sufficient understanding of algorithms so that they can validate the documentation that explains the algorithm and its intended impacts with the data that is fed into it. To demonstrate replicability and consistency, the algorithms, their explanations, and associated data training sets should be archived by the institution, a role that is well suited to research libraries and a function that data librarians often perform as members of research teams. Making these materials openly available will allow other researchers to replicate the findings and make improvements to further advance research. Concerns researchers may have about others claiming credit by using these algorithms and data can be mitigated by archival restrictions such as embargoes or limited-access restriction if necessary. Establishing institutional policies around documenting algorithms and archiving both the algorithms and training data will benefit the larger research community and can provide a safeguard for the university against the risk of claims associated with harm caused by the use of the technology. A Role for Research Librarians in AI in University Education The underrepresentation of women and people of color in AI at the PhD level is a reflection of the underrepresentation that exists at the undergraduate level in students’ choice of majors and the courses they take. Undergraduate enrollments in computer science have increased significantly, with a growth of 136% among full-time computer science majors between 2006 and 2015.24 With this growth, there has not