Classification and Natural Kinds

Conclusion
Difficulties with boundary drawing between mechanisms and between natural kinds have surfaced in the practice of scientific classification, forcing scientists to employ a significant amount of conventional elements not only in very early stages of classification but even in order to be able to begin classifying at all. Tobin argues that the scientific practice undermines the HPC account of natural kinds, and if the prevalence of conventional elements allows one to maintain realism about the resulting natural kinds at all, then it can only be promiscuous realism. This essay has argued that even though the complexity of the causal structure of the world allows for a variety of ways of classifying objects into natural kinds based on structural similarity, it is not self-evident that mere structural similarity would necessarily constitute an ontologically committing natural kind. Furthermore, such extremely liberal notion of natural kinds tends to redundancy, since it does not seem to be anything above certain structural similarity between random chunks of the causal structure of the world. It has been suggested that adaptation of ontic structural realism to biology could (1) to an extent avoid the issue of boundary drawing between mechanisms and between natural kinds, (2) avoid conventionalism by maintaining realism about the underlying structures of the world, and (3) avoid ontological commitment to a vast amount of natural kinds, objects and individuals by eliminating the notion of biological objects and, hence, natural kinds.
…