Contribution
Sciences sort their objects into kinds, and to call a kind natural is to say it follows the structure of the world rather than human interests. Scientific realism holds that successful sciences uncover such kinds, and that this justifies inductive inference. The entry covers the metaphysics of natural kinds: naturalism vs conventionalism, realism, essentialism and eliminativism. It then looks at how kinds work in biology, chemistry, psychology and the social sciences, at the epistemology of kinds, and at the semantics of natural-kind terms from descriptivism to direct reference and externalism.
Skim
Abstract, figures and selected sections, read when the work was seeded. Not enough to state its assumptions or results exactly; a Read note replaces this one.
- §1 (intro): four metaphysical questions frame the whole entry, and theories are placed on a scale of commitment from out-and-out conventionalism to irreducible, essence-bearing kind entities. The authors note that intermediate views are the most popular.
- §1.1: naturalism (weak realism) vs conventionalism (weak and strong, with Mill and Locke as sources), and Dupré’s promiscuous realism (§1.1.3), which allows many legitimate cross-cutting taxonomies (the lily example).
- §1.2–1.3: strong realism, treated as an ontological commitment to kinds as entities, by analogy with the universals debate, and essentialism (Ellis, Devitt, Fine’s sense of essence). Boyd’s homeostatic property clusters are offered as a liberal alternative suited to biology.
- §1.4: eliminativism. Hacking says “Delete every mention of natural kinds” and Ludwig agrees; Conix and Chi reply that the concept still does epistemic work for non-arbitrary classification.
- §3: a naturalist, practice-led and epistemology-first turn (Slater; Ereshefsky and Reydon) that values kinds for their epistemic role independently of metaphysics.
- §4.4: whether the rigidity of kind terms (Kripke, Putnam) yields necessary a posteriori essences, or smuggles essentialism in. Salmon and Burge are cited against, and Bird 2020 for a closer tie between semantics and metaphysics.
Open questions
- Under this heading the work presumably supplies the philosophical target: a minimal sufficient statistic as a formal candidate for “the natural grouping”, i.e. the part of the data that reflects structure rather than accident. The entry itself never mentions information or compression, so that link belongs to the owner and a note should say so.
- Boyd’s HPC view (clusters kept stable by causal mechanisms) and Dupré’s pluralism are the positions most directly comparable with the idea that model-class choice determines what counts as structure (compare k04, Lemma 6.2).
- Check the Spring 2024 archived edition against the 2022 revision if the note is going to cite the edition.