NOTE-395: Hindriks & Guala — Institutions, rules, and equilibria

Contribution

The paper sets three accounts of institutions side by side: institutions as rules (North, Weber, Hodgson), as equilibria (Lewis, Schotter, Sugden, Binmore), and as systems of constitutive rules (Searle). It argues that the first two are each inadequate and jointly adequate. That combination is credited to Aoki and to Greif & Kingston. The paper’s own step is the second half: Searle’s constitutive rules are shown to be regulative rules re-expressed with a new term. If so, all three accounts are one theory. The idea is credited to Hindriks’s 2005 dissertation (n. 11).

Key insight

A constitutive rule is what a regulative rule looks like after you name the status it confers. “If you occupy land first, you may use it exclusively” becomes “first occupancy makes land your property*” plus “property* means exclusive use”. The name lets one word carry a bundle of rules, which is useful. It creates nothing that the rule, held in equilibrium, did not already settle.

Assumptions

• Players and games. One-shot games with rematching each round (n. 7); the mixed-strategy equilibrium is set aside. Institutions are confined to coordination games with several equilibria, in which strategies must be correlated on a signal.
• Representation. A third condition: players represent the equilibrium in symbolic form, as rules. It is introduced to exclude animal conventions (§4) and is a stipulation about what “institution” should cover; the reply calls the human/animal line a working hypothesis, not a requirement.
• Collective acceptance is read as “whatever set of intentional states is needed for institutions”. Ordinary preferences, expectations and common knowledge may qualify (§5 and n. 14). The paper is neutral on whether collective intentions reduce to individual ones.
• “Counts as” is interpreted by two biconditionals: X counts as Y iff X is collectively accepted as Y, and X is collectively accepted as Y iff X is Y (§5).
• Definitions must meet Belnap’s (1993) criteria, eliminability and conservativeness, for the reduction to go through (§6).

Key results

• Against pure rules (§2). A rule that is known and not followed (a 1799 French ban on women wearing trousers; a 65 mph limit in a 75 mph equilibrium) shows that stating a rule does not make an institution. What does is an incentive structure that keeps behaviour where it is.
• Against pure equilibria (§§3–4).
  • Not every equilibrium is an institution. Mutual defection in a prisoner’s dilemma needs no coordination. Hare–hare in a stag hunt needs no correlation. Both are excluded.
  • Property* is modelled as the correlated equilibrium of hawk-dove (Fig. 1: (U,U) = (0,0), mixed cells (2,1) and (1,2), (NU,NU) = (1,1)) with first occupancy as the correlation device. Expected payoff is 1.5 each when occupancy is equally likely.
  • Pararge aegeria butterflies also play a “bourgeois” correlated equilibrium. They have no institution, because they cannot condition on new signals or invent new equilibria. Representation is the difference.
• The transformation (§5).
  • [R] If one is the first to occupy a piece of land, one has the right to its exclusive use.
  • [B] If a person first occupies a piece of land, then it is her property*.
  • [S] If a piece of land is someone’s property*, she has the right to use it.
  • [C] = [B] and [S], with the form “if C then X is Y, and if Y then Z”. This is the XYZ schema, the “transformation view”.
• Consistency (§6). The Y-term is conservative (nothing new follows that lacks it) and eliminable (reverse [C] → [B],[S] → [R]). Behaviour under [R] is extensionally equivalent to behaviour under [B] and [S]. So the constitutive-rules theory is “reducible to” the rules-in-equilibrium theory.
• Four virtues (§6). Two linguistic: institutional terms are what ordinary talk uses, and they economise cognition. Two ontological: multiple realisability corresponds to multiple equilibria, and the theory is parsimonious.
• Against Searle’s necessity claim (§6). If constitutive rules were necessary for institutions, practices of etiquette, and in the reply driving on the right, would not be institutions. The authors see no ground for denying that they are.
• Reference (§6). On Lewis’s account of theoretical terms, “if the theory that introduced it is true it does name something”. So if the constitutive rules are correctly transformed from true regulative rules, “property” names something and the institution exists.
• Reply (JIE 11(3):515–522).
  • Common knowledge is not needed for coordination. Correlated equilibrium was chosen to highlight the variety of correlation devices, not to require Aumann-style epistemics.
  • Normativity does not require constitutive rules. “No one is allowed to use my garden” is a regulative rule with normative force. Rights are other people’s obligations, so “property” names a bundle of regulative rules.
  • Searle’s “desire-independent reasons” is a dilemma for him. On the weak reading it is a platitude compatible with equilibrium accounts. On the strong reading it is an unargued anti-Humean thesis.
  • A theory of institutions should stay neutral among theories of normativity. Lewis, Bicchieri and Sugden (expectations and resentment), Gilbert (joint commitment), and emotion and reasons accounts are listed.

Claims

id	claim	strength	support
C1	A rules-only account cannot tell effective from nominal rules	moderate	two examples, §2 (trousers law, speed limit after Greif & Kingston 2011)
C2	An equilibrium-only account is too permissive: animal conventions are correlated equilibria but not institutions	moderate	one biological case (Pararge aegeria, after Davies 1978 and Maynard Smith 1982); the human/animal line is conceded in the reply to be a working hypothesis
C3	Institutions are correlated equilibria of coordination games, represented as rules	a definitional proposal	motivated by C1–C2 and by intuitions about the prisoner’s dilemma and stag hunt, §§3–4
C4	A constitutive rule is a regulative rule split by a new term into base and status rules	strong for the toy case	§5, [R] → [B],[S] → [C], worked once for property*
C5	Every constitutive rule reduces in this way, so the constitutive-rules theory reduces to rules-in-equilibrium	weak–moderate	§6, Belnap’s criteria argued for the one case; constitutive rules with no prior regulative rule (Searle’s chess) are not treated
C6	Multiple realisability of institutional kinds is multiple equilibria	weak	illustration, §6: “at least in many cases, they come down to the same thing”
C7	Searle’s view must deny that etiquette is an institution, which is implausible	moderate	internal argument, §6; extended to traffic rules in the reply
C8	Institutional terms refer if their introducing rules are true	conditional	Lewis’s “How to Define Theoretical Terms” (cited from Philosophical Papers I, 1983), §6 and n. 16
C9	Normativity does not need constitutive rules: a right is a bundle of others’ regulative obligations	moderate	reply, one example (Bob’s garden); the translation is asserted for “property” generally

Concepts

• Rules-in-equilibrium. An institution: a correlated equilibrium of a coordination game, whose strategies the agents represent as rules.
• a-rule / o-rule. An agent-rule summarises and guides one’s own behaviour; an observer-rule summarises others’. Equilibrium theories are observer theories (§4).
• Base rule / status rule. A base rule says what it takes to have a status (X is Y); a status rule says what having it means (if Y then Z).
• Transformation view. Constitutive rules are regulative rules transformed by introducing a Y-term (after Hindriks 2005, 2009).
• Property*. The toy proto-institution: exclusive use by the first occupant, as a correlated equilibrium of hawk-dove.

Connections

• Lewis (LIT-473) supplies the equilibrium account of convention and, separately, the account of theoretical terms that §6 uses for reference. The paper generalises Lewis’s symmetric coordination equilibria to correlated equilibria. Lewis’s own book is unread here, so what it says is taken from this paper and the SEP entry.
• North (LIT-489) is the rules account’s exemplar, quoted from his 1990 book. The 1991 essay in the record states the same definition. Read against it, C1 is a fair objection: North explains compliance by enforcement cost and leaves informal constraints as an open question.
• Searle. The target is The Construction of Social Reality (LIT-485, filed alongside this, Deferred) and Making the Social World, which is not in the record. Searle’s Chinese Room (LIT-400) is a different subject, and this paper does not touch it.
• Bicchieri (LIT-493) is cited only in the equilibria list (n. 4) and, in the reply, for normativity from mutual expectations.
• Guala’s book (LIT-472). Its chapter 1, the publisher’s sample, repeats this paper’s §2 cases. Its index lists base rules, status rules, “institutional term” and Hindriks (pp. 58, 62–69), which suggests that chapter 5, Constitution, carries §§5–6. That is inferred from the index, not read.
• Epstein (LIT-478, LIT-487). His précis treats frame principles as a generalisation of constitutive rules and collective acceptance as one theory of anchoring. This paper also finds Searle’s form too narrow, but replaces it with equilibrium rather than with a separate anchoring relation. The précis lists Hindriks among the “status model” theorists of group agency, whom it criticises; that is a different Hindriks work, not this paper.
• Tomasello (LIT-125, NOTE-185) takes obligation to come from a joint agent “we”. The reply’s position is that a theory of institutions can stay neutral on where normativity comes from, and it lists Gilbert’s joint commitment as one option.
• Dennett, “Real Patterns” (LIT-220). The reply’s claim that ontology should follow the best explanatory theory, and §6’s “if the theory is true, the term names something”, is a realism about institutions licensed by explanatory success. That is close in spirit to Dennett’s mild realism about patterns, though the paper does not cite him; the link is mine.

Bearing on the record

• THEORY-043: indirect, and neutral. The account is about institutions, not groups, and it is deliberately neutral about collective intentionality (n. 14). It neither needs nor excludes a group-level subject. Its one point of contact is methodological: it treats an institution as real when a true theory with that term explains behaviour. A group-mind claim would need a further argument that the group is the bearer of the states, which is exactly what THEORY-043 says no one has supplied for experience.
• Group agency (LIT-391, LIT-401). List & Pettit’s group agents are organised collectives; this paper’s institutions are equilibrium patterns that need no collective agent. The two are compatible and do not address each other.
• No instruction for machine-learning practice appears, so nothing here points to the Anthology of the SOTA.

Limitations

• The reduction is shown once. C4 is a worked example. C5 claims every constitutive rule arises this way, but no case is given where a status is created with no regulative rule behind it. Searle’s chess appears only as his example of a constitutive rule, in the argument that etiquette is an institution too.
• Collective acceptance is made easy to satisfy. Defining it as “whatever set of intentional states is needed for institutions” lets the “counts as” biconditionals hold by construction. The reduction then inherits that looseness.
• Equivalence is extensional. [R] and [B]+[S] are shown to prescribe the same behaviour. Whether they confer the same rights is the point Searle presses; the paper answers it only in the reply, and only by example.
• No normativity in the paper. The authors concede this in the reply and point to a companion paper (Guala & Hindriks 2014, Philosophical Quarterly), not read.
• Representation is not analysed. What it takes for agents to “represent the equilibrium in symbolic form” is left at Sterelny’s decoupling of stimulus and response.

Open questions

• Is there a constitutive rule with no regulative pre-image, such as a rule creating a status whose only content is further statuses? If so, C5 fails for it.
• Does the transformation preserve deontic content, or only behaviour? The reply’s bundle-of-obligations claim (C9) is the answer offered, and it needs a case where the bundle is large or open-ended.
• Where exactly does the representation condition draw the line between human and animal conventions, and can it be tested? The reply proposes intervention speed as one test.

Corrections

• none to a seeded skim (there was no seed)
• Hodgson (2006) is mis-cited in both the paper and the reply as Journal of Economic Issues 15: 1–23. Crossref gives 40(1): 1–25, DOI 10.1080/00213624.2006.11506879.
• Searle’s money example names the Bureau of Engraving and Printing as the issuer. The paper corrects this itself (n. 12): the bills are issued by the Federal Reserve.
• The reply cites the main paper as 2014, its online-first year. The print citation is JIE 11(3), 2015, and both are the same article.
• The reply’s text spells “Skorupsky” where its reference list has Skorupski (2010).