This paper contributes to the debate on the design of policies that are effective in promoting interdisciplinary research and adequate for evaluating their impact. To this purpose, we present a novel approach to the problem of facilitating team science, by using an experimental framework based on the notion of social network intervention. Specifically, by using social network analysis of publication and grant data at the University of Florida, we designed and implemented an innovative type of pilot funding program conceived as a form of intervention on the University’s collaboration network. The intervention, conceived as a randomized experiment, broadly consisted in identifying research communities in the University’s collaboration network, and maximally increasing the cohesion of selected communities by adding collaborative links with specific structural properties in their subnetwork. This article describes the intervention design, reports findings from the program implementation, and presents a framework for rigorous future evaluation of the intervention.