In recent years, there has been a growing interest for social kinds, to such an extent that their traditional exclusion from being natural kinds has been questioned. The Homeostatic Property Cluster (HPC) account, being an alternative to the essentialist accounts of natural kindhood, paved the way for treating social kinds as candidate natural kinds. However, one of the most compelling defenses of the view that social kinds can be natural kinds is to be found in the work of Muhammad Ali Khalidi, whose view differs from the HPC account. Khalidi’s account of natural kinds takes the HPC account’s insight that the projectibility of natural kinds is grounded in causal properties and relations, but it gets rid of the homeostatic mechanism as a necessary feature, developing instead a more general causal theory of natural kinds (Khalidi 2013). Relying on this view, Khalidi claims that those social kinds whose properties are causally related such as recession or racism, are natural kinds, whereas social kinds whose properties are conventionally related such as permanent resident or prime minister, are not (Khalidi 2015).

In my work, I argue that even kinds whose properties are conventionally related can be natural kinds, since they can fulfill the epistemic role that natural kinds play. Legal and linguistic kinds are examples of categories that they can fulfill this role, given the explanatory and inductive powers in their relevant disciplines. Linguistic kinds such as noun and verb allow inductive inferences about the relations that words have with each other in a grammatical structure. Legal kinds such as civil law and criminal law, are key categories in for explaining legal systems. Although there might be a complicated causal chain that led to the creation of the conventional properties of these social kinds, the properties themselves of legal and linguistic kinds are not causally related.

If some conventional social kinds are natural kinds, then any view of natural kinds based on causality such as Khalidi’s causal account or the HPC account seems inadequate, since the properties of conventional categories are not causally related. Essentialism would not help either, given its restrictions to physical, chemical, and – sometimes – biological kinds. Thus, I claim that the best account available is the one presented by Matthew Slater, the Stable Property Cluster (SPC) account, that shifts the focus from the ontological ground for induction – the essence or the homeostatic property cluster – to the epistemic ground of the kind for induction (Slater 2015). In other words, this anti-realist and pragmatic account focuses on the stability of the kind rather than on what causes this stability. Thus, the flexibility of the SPC account allows us to include as natural kinds the conventional categories investigated by the social sciences.