Abstract

Project selection by funding bodies directly influences the division of cognitive labour in scientific communities. I present a novel adaptation of an existing agent-based model of scientific research, in which a central funding body selects from proposed projects located on an epistemic landscape. I simulate four different selection strategies: selection based on a god's-eye perspective of project significance, selection based on past success, selection based on past funding, and random selection. Results show the size of the landscape matters: on small landscapes historical information leads to slightly better results than random selection, but on large landscapes random selection greatly outperforms historically-informed selection.