“…Using inferred networks as a null model may show how the mechanistic models need to be refined, for example, including the effect of traits such as seed shape, integument thickness or lipid content (Gaba et al., 2019 ; Honek et al., 2007 ; Sebastián‐González et al., 2017 ), or other local drivers, for example, total seed abundance, vegetative cover, or the presence of alternate invertebrate prey. Competition for alternate prey or density dependence in the risk index would lead to prey switching (Gendron, 1987 ; Manly et al., 1972 ), which has recently been inferred for carabids from empirical data (Carbonne et al., 2020 ; Gray et al., 2021 ). Therefore, when suitable data are available, each of these assumptions could be tested in future models.…”