Interactions between species are the building blocks of ecosystems, yet remain a large gap in ecological understanding. Sampling species interactions is difficult: species abundances vary in space and time, and interactions can be contingent on specific environmental conditions. Because it isn't feasible to comprehensively sample every location on Earth, we need to optimize where we sample to maximize our ability to best detect interactions between species. Here we argue that by taking the perspective of a virtual ecologist first and simulating the underlying ecological process as well as the process of data collection, we can improve the quality of our sampling and the capacity for our analysis to detect ecological phenomena. We present this virtual ecology framework for optimizing spatial sampling of interactions, and associated software (BiodiversityObservationNetworks.jl) for producing sampling priority maps.