Fairness in Reviewer Recommendations at Elsevier

1 ABSTRACT At Elsevier we aim to help scientists further their research. On the one hand, by offering a platform for publishing ground-breaking research, and on the other by helping researchers discover relevant content to assist their work. In our team, Editorial Data Science, we support the former goal by providing tools to help editors in the decisions that they make, from finding reviewers to recommending transfers for manuscripts to more appropriate journals. To improve the workflows for publishing research and help editors in finding relevant reviewers we developed a reviewer recommender. The reviewer recommender will, based on a submitted manuscript, recommend reviewers that optimise the fit between the manuscript, the journal, and the reviewer themselves. To achieve this, the reviewer recommender is built on top of the three principles of quality data, expert knowledge, and experimentation. We leverage data from across Elsevier, ranging from Scopus profiles which includes publication histories and academic impact data, to data from our publishing platform which includes historic invites and subsequent actions around submitted manuscripts. All this data is used to develop and evaluate appropriate models for recommending reviewers.