This article demonstrates how a technique called Agent-Based Modelling can address a significant challenge for effective interdisciplinarity. Different disciplines and research methods make divergent assertions about what a satisfactory explanation requires. However, without a unified framework analysing the implications of these differences systematically, debate cannot transcend mere competing assertion. Using a sequence of examples, I demonstrate that Agent-Based Modelling provides such a unified framework by showing how pairs of models may display quantitatively distinct behaviours when differing only in (for example) including a social network. The ability to quantify differences arising from divergent assumptions about what models should include makes them subject to empirical evaluation (rather than mere contention). Although the article uses social network examples for accessible presentation, the approach of building paired models is quite general and can, therefore, illuminate other significant social science controversies (like the role of rationality and the importance of ethnographic detail).