In Realism, Anti-Foundationalism and the Enthusiasm for Natural Kinds, Richard Boyd defends the use of cluster properties as natural kinds and natural kind terms, while refuting Ian Hackings claim that they are intrinsically social because they fail to satisfy the condition of being defined by a set of necessary and sufficient properties. While acknowledging that there are differences between kinds defined by cluster properties and kinds defined by sets of properties, and inherently, between natural and social properties, he believes that there are still excellent reasons for incorporating many property-cluster kinds and social kinds to the model of natural kinds. As such, he aims to defend this assimilation by forwarding a historical account of the tradition while highlighting why cluster properties are viable and make sense in their current form.
First of all, Boyd believes that the core of the current conception of natural kinds rests upon three central claims that will hold of any property-cluster kind used in either explanation or induction. They are:
a. The definition of the kind in question is determined a posteriori.
b. The referential use we make of the kind in induction requires that it be defined by a set of properties whose membership is determined by the causal structure of the world.
c. The relationship between the term and the kind in question depends on the existence of an epistemically relevant pattern of causal relations between instantiations of the kind, and the use of its related term.

To further this, he presents a historical account on the conjecturing of natural kinds beginning with Locke. Boyd claims that the tradition is largely nominalistic simply because it is verificationist. He states, “the nominalism in question is an outgrowth of the traditional empiricist skepticism.” (Pg. 129) As such, under the Lockean position, the classification of entities into natural kinds is solely “the workmanship of the understanding.” (pg. 130) This is problematic and compromised, however, since the only constraint is the logical consistency of the ideas bonded to form it. And, according to Boyd, this is arbitrary because it ultimately causes a tension between empiricist nominalism and how we account for induction. He states, “it is a natural fact about a particular natural kind that reference to it is appropriate for certain sorts of inductive generalizations and this natural fact is explained by property correlations which obtain independently of the understanding.” (Pg. 130) So, if observable properties of entities are determined by “the primary qualities of their insensible parts”, our inductive generalizations are unjustified. (Pg. 131) In other words, we fail to justify our belief that since K’s observed properties have been reliably displayed in the past, they will continue to be displayed reliably in the future.
Boyd now shifts his discussion to the twentieth century and its theory-dependent culture. The newfound enthusiasm for natural kind terms is no longer rooted in empiricism, but rather in scientific realism. This is because the actual method of science has more or less led to the confirmation of Locke’s worst nightmare- the means by which we acquire our scientific knowledge is unavoidably theory-laden. So, scientific realism rests on the abductive case that we must make an inference to the best explanation when assessing facts in the world. Boyd states, “It seems possible to argue that inductive inferences in science about observables are reliable only because they are guided by methodological principles which reflect previously acquired (approximate) knowledge of unobservable real essences.” (Pg. 133) It is from this doctrine that the current enthusiasm for natural kind terms has sprung, since we simply cannot eliminate the a posteriori element from any scientific kind or entity.
As such, Boyd wants to show that whether one subscribes to a sophisticated empiricist account of science, or a realist account of science, there exists justifiable reasons to extend the notion of a natural kind term to include both property-cluster kinds, and social kinds that play a significant role in our inductive methods. He further claims that the failure of foundationalism- a failure based on the fact that all inductive reasoning rests upon a posteriori projectability judgments- motivates this. Moreover, this failure is additionally connected to a radical contingency in epistemology that we must rely on inferential principles that lack a priori justification. And, according to Boyd, it is this crucial fact that should determine the scope of our theory on natural kind terms. Scientific classifications must be defined a posteriori if they are to reflect the actual causal structure of the world, and because we cannot identify projectable generalizations without doing so. So, “ natural kinds reflect a strategy of deferring to nature in the making of projectability judgments.” (Boyd, Pg. 139) Yet, how broad is the spectrum of cases that require veneration to nature?
Boyd believes the answer to this question is always. Since our projectability judgments are inherently a posteriori, “ we should always require the sort of semantic machinery indicated by the theory of natural kinds when our aim is induction or explanation.” (Boyd, Pg. 139) So, prima facie, the scope appears to be quite wide. In defining a natural kind, “we should be required to defer to the world just in case and to the extent that reference to the kind in question is to be part of an inductive or explanatory project.” (Boyd, Pg. 140) Contrarily, in the cases where we are concerned with merely workable conventions, deference to the world is unnecessary. As such, according to Boyd, the kind “gene” falls into the first category, and the kind “fish fork” falls into the second. Importantly, Boyd also thinks that there should be more complex kinds whose definitions combine both the naturalistic and the conventional.
It is for this reason that Boyd criticizes Hacking. Since Hacking maintains that a natural kind term should be defined by a set of necessary and sufficient conditions, he believes that property-cluster kinds can never be natural kinds. Yet, according to Boyd, if a natural kind is associated with a large group of properties, what defines the kind is causal and not conceptual. So, “the natural definition of one of these homeostatic property cluster kinds is determined by the members of a cluster of often co-occurring properties and by the (“homeostatic”) mechanisms that bring about their co-occurrence.” (Boyd, 141) And, importantly, it is an a posteriori question that answers which properties, and their homeostatic mechanisms, count.
Moreover, there exist too many paradigmatic cases of natural kinds that are staunch counterexamples to the claim that natural kind terms satisfy a set of necessary and sufficient conditions. Specifically, Boyd backs up his claim by illustrating the case of biological species, since biological species are defined by homeostatically-related and imperfect morphological, physiological, and behavioral properties. Specifically, “for sexually reproducing species, the exchange of genetic material between populations is thought to be essential to the homeostatic unity of the other properties characteristic of the species and it is thus reflected in species definition.” (Boyd, Pg. 142) This indeterminacy is a necessary entailment of evolutionary theory, since speciation relies on the existence of intermediary populations between the parent species and the emerging species. Any classification that eliminated this indeterminancy would clearly be unsuitable and disingenuous.
Following, Boyd now makes the case for the inclusion of social kinds into the theory of natural kinds. He addresses two counterarguments that might exist: the metaphysical, and the epistemological (a view he doesn’t discuss). The metaphysical counterargument centers upon an assumption that scientific realists claim, and social constructionists deny- that the reality conceptions. Clearly, certain phenomena that scientists study are causally dependent on social practice. So, it is a mistake to think that realists and constructionists differ in the extent to which they view the theory dependence of reality. Where they differ is over the nature of that mind dependence. Realists deny that the “adoption of such conceptions makes any non-causal, logical or conceptual contribution to the causal powers of the objects of study.” (Boyd, Pg. 144) The constructionists, contrarily, believe that there is an actual social construction of the causal properties of the object of study. For both the realists and the empiricists, however, social practices are metaphysically innocent. In other words, the causal mechanisms of science supervene causally on the differences in the resulting social practices.
So, are the concepts of race, gender, and sex, homeostatic property cluster kinds? If we use Boyd’s criteria to answer these questions, we get a resounding yes. Since, in addition to the biological parts of the definition, we can view the social aspects as confirmable a posteriori. For example, the property cluster kind term “sex” could be perhaps defined by the following properties: male/female, xy/xx chromosomes, masculine/feminine, lack of breast tissue/breast tissue, no ability to become pregnant/pregnancy, dominance of testosterone/dominance of estrogen, ect. Accordingly, the property cluster kind for “race” could be defined by some the following properties: ethnic origin, phenotype, cultural identity, morphology, ancestry, social identity, ect.
In conclusion, I think that Boyd makes a strong case for extending our natural kind theory to social kinds, and for extending our use of homeostatic cluster kind terms. For, without this broader category we risk going to far in either direction. There are simply to many definitions that require a necessary indeterminancy in science for us to continue with the Aristotilean version. Moreover, since many of our concepts contain properties that exist in both the natural world and the social world, it makes sense for our theory to recognize this supervenience. There are deep epistemological and metaphysical reasons, that relate to induction and explanation, for doing so.