Realism about group agency (List and Pettit 2011) is the view that groups are real entities, irreducible to the set of networked individual agents constituting them. It has been argued that realism is compatible with methodological individualism, the doctrine that all explanations of social phenomena must be cast in terms of individuals and their features. This reconciliatory move is supposed to offer realists a way to resist eliminativism or bizarre emergentism. However, I argue that such a compromising manoeuvre ultimately fails and the attempts to defend group agency realism collapse into rival theoretical alternatives. To that end, I show that realist views fail to provide an adequate causal story about how group agents impact the world. More precisely, I point out issues with causal overdetermination in cases of constitutive dependence and with the quantification of differential causal contribution. Along the way, I show that appeals to judgment aggregation paradoxes or normative autonomy are inconclusive. Eventually, I discuss the value of group agency views (why they are central to social explanation, prediction and design) and offer a brief recommendation concerning our stance towards group agency talk.
