"Research evaluation practices linked to social impact have important systemic effects on the prioritization and organization of research while at the same time leading to the delivery of higher social value. Amidst growing criticisms, global research evaluation has evolved in a different direction, characterized by quantitative metrics and mimetic behavior. The article deals with the forces that sustain the prevailing research evaluation system, asks why it has proven to be so resilient, and discusses alternative proposals. A new argument for building an alternative is put forward: the need for a developmental role for universities, introducing the notion of ‘connected autonomy’ allowing universities to productively and in a nonsubordinated way collaborate with a broad set of actors to achieve desirable social changes. An outline is presented for how to make research evaluation practices and the pursuit of developmental goals more compatible, an important issue for knowledge public policy."