This disclosure describes one or more embodiments of systems, methods, and non-transitory computer-readable media that determine a degree of diversification for item recommendations to a user based on the user's input and generate diverse item recommendations for the user according to the degree of diversification. For instance, the disclosed systems can receive a diversification metric from a client device based on a user interaction with a selectable tool (or another interactive element) in a graphical user interface. From among data segments representing users clustered according to item affinities, the disclosed systems can subsequently use the diversification metric to identify a data segment that is diverse from a reference data segment for the user. The disclosed systems further rank items associated with the diverse data segment to select an anomalous item as an item recommendation for display on the client device.