Brian Epstein (2018), Stanford Encyclopedia of Philosophy; first published 21 March 2018, substantive revision 4 March 2024 — https://plato.stanford.edu/entries/social-ontology/
Key takeaways
- Two inquiries. One asks what a social entity is and what determines it: the conditions for something to be a group belief or a woman (§3). The other asks what sets up, constructs or “anchors” the category: what makes those the conditions (§4). The entry likens them to semantics and metasemantics.
- Grounding and anchoring. Recent work treats ontological determination as metaphysical grounding. Epstein argues that social construction, “anchoring”, is a distinct relation that is neither grounding nor meta-grounding. Haslanger, Hawley and Schaffer accept the practical distinction and resist the sharp one (§2.2.2).
- The field’s maps. Ontological individualism is widely assumed and recently contested (§3). Lewis’s convention, Hart’s accepted rules, Searle’s status functions, Millikan’s reproduced kinds, Boyd’s and Hacking’s causal loops and practice theories are all theories of anchoring (§4). Groups, group agency, institutions, race and gender, artifacts, money, law and language are the main domains (§5).
Standing in the record
Filed on 2026-10-02 at the owner’s request, as a readable survey for the
new social-ontology topic (ADR-017), and as the source for what The
Ant Trap’s grounding–anchoring framework is, since the book itself is
not lawfully readable (LIT-478). NOTE-393 is the reading.
What the framework is, from §§2.2 and 4.5, with the author’s précis of the book (read for LIT-478) where it is fuller:
- A social fact is grounded by other facts: a pig in a place grounds the fact that there is an unkosher animal there.
- The conditions for grounding a fact of that kind are given by a frame principle, which holds across a frame, a set of possible worlds.
- What puts a frame principle in place is its anchor: perhaps the agreement of religious authorities, collective acceptance, practice or divine command. Anchors are heterogeneous. US legal categories are anchored by jury instructions, trial results, legislative enactments, judicial interpretation and environmental regularities (§4.5).
- Searle’s constitutive rules are frame principles of one narrow form, and his collective acceptance is one theory of anchoring (précis). Lewis’s convention, Hart’s rule acceptance, Millikan, Boyd, Hacking and practice theories are other anchoring theories (§4).
- “Socially grounded” and “socially constructed” are two dimensions. Crowds are socially grounded but not constructed; phenotypic racial categories are constructed but not grounded; fiat money is both (§2.1). The entry cites this to an “Epstein 2016” that its bibliography does not list (NOTE-393, Corrections).
How it places the other works filed with it. Lewis (LIT-473) and Bicchieri (LIT-493) are in the convention list of §4.2.2. Searle’s theory of status functions (§4.2.3) is the subject Hindriks & Guala (LIT-481) reduce to rules in equilibrium; the entry reports Searle’s view and the critics of collective acceptance in money (§5.6) but does not discuss that reduction. North (LIT-489), Guala (LIT-472) and Guala & Hindriks 2020 are placed in new institutional economics (§5.3).
Group minds and THEORY-043. §5.2 is a short survey of the group-agent literature the record already holds in part: List & Pettit’s “groups with minds of their own” (LIT-391), Huebner on complex groups (LIT-426), and the functionalist model of group agency as a realised “system of practical activity”. It cites Schwitzgebel 2015 (LIT-159) and List 2018 (LIT-401), the record’s two anchors for THEORY-043. It does not take a side on group phenomenal consciousness. What the framework implies for THEORY-043 is set out in NOTE-393; in brief, it says the whole whose mindedness is in question is not fixed by its members, and that both of THEORY-043’s routes are stated as if it were.
Realism about patterns. §4.3 surveys “pattern theories” (functional roles, Millikan’s teleofunctions, Boyd’s homeostatic clusters, Hacking’s looping kinds), on which social kinds are real because they support induction. That is the social-kinds counterpart of Dennett’s real patterns (LIT-220), though the entry does not cite Dennett in that section.
No instruction for machine-learning practice appears, so nothing here points to the Anthology of the SOTA.