Normative Position

An agent’s total bundle of jural relations at a point in a normative system — what it is obliged, permitted, forbidden, empowered, or liable to do with respect to others. Made precise through Hohfeldian Analysis (claim/duty, privilege/no-right, power/liability, immunity/disability) and built up from Constitutive Rules and Empowerment; the unit normative MAS reason over when deciding whether an action is institutionally in order.

In this vault

Backlinks

Linked Pages

Hohfeldian Analysis

Wesley Hohfeld’s (1913) decomposition of vague “rights” talk into eight fundamental jural relations arranged as correlatives and opposites: claim-right/duty, privilege(liberty)/no-right, power/liability, immunity/disability. Used in normative multi-agent systems to give precise structure to normative positions and to distinguish a power (Empowerment) from a permission.

In this vault

Jones-Sergot Normative Systems

On the Characterisation of Law and Computer Systems: The Normative Systems Perspective

Reference: Jones, A. J. I. & Sergot, M. (1993). On the Characterisation of Law and Computer Systems: The Normative Systems Perspective. In J.-J. Ch. Meyer & R. J. Wieringa (eds.), Deontic Logic in Computer Science: Normative System Specification (pp. 275–307). John Wiley & Sons. Semantic Scholar record · DEON 1991 precursor (workshop paper, more accessible) · SEP: deontic logic

Summary

Jones and Sergot’s chapter is the founding statement of the normative systems programme in AI and computer science — the move that takes computer systems (including multi-agent systems) seriously as normative artefacts whose behaviour can be analysed in the same vocabulary as legal systems. The central technical contribution is a formal account of the counts-as relation: the relation by which a brute (physical, computational) event is treated by an institution as constituting a particular institutional event. A piece of paper of certain composition, signed in certain ways, counts as a contract in the legal institution; a particular sequence of bytes on the wire, conformant to a particular protocol, counts as a binding promise in an agent society. The counts-as relation is the bridge between brute facts and institutional facts, between syntactic events and their normative significance.

The paper develops counts-as as a conditional connective ⇒s indexed by an institution s: A ⇒s B reads “in institution s, A counts as B.” The counts-as connective is not classical material implication: it is institution-relative, non-monotonic (additional facts about the institutional context may defeat the count-as), and constitutive (the count-as relation is partly what makes the institution be the institution it is). The framework lets the authors model standard legal phenomena — speech-acts that create institutional roles, signing-procedures that count as enacting law, delegation that lets an agent’s act count as another’s, and the distinction between rules that regulate existing institutional activity and rules that constitute the institutional categories themselves. The Wieringa–Meyer volume’s surrounding chapters develop deontic logics (obligation, permission, prohibition) that combine with counts-as to specify normative systems.

For CBCL and the public-semantics line, the counts-as relation is the missing institutional bridge between the wire format and the commitment-state. A CBCL dialect contract is precisely a counts-as specification: in this dialect, this byte-sequence on the wire counts as a binding commitment with these debtor / creditor / condition / context slots. Without counts-as, the move from “the agent emitted these bytes” to “the agent has incurred this commitment” is unprincipled; with counts-as, it is a well-defined institutional bridge inheriting decades of deontic-logic infrastructure. The paper is also the unification point for the broader normative MAS literature — Singh’s commitments, Fornara–Colombetti–Verdicchio’s institutional ACL, and the broader European MAS deontic tradition (Boella, van der Torre, Castelfranchi, Carmo–Jones) all sit downstream.

Key Ideas

Counts-as relation ⇒s: an institution-indexed conditional. A ⇒s B reads “in institution s, A counts as B.” The bridge between brute facts (physical, computational events) and institutional facts (contracts, commitments, roles).
Constitutive vs regulative rules: counts-as rules are constitutive — they partly define what the institution is. Regulative rules (deontic operators: obligation, permission, prohibition) presuppose the institutional categories and govern behaviour within them. The constitutive/regulative distinction goes back to Searle 1969; Jones–Sergot give it a formal home.
Non-monotonic, institution-relative, defeasible: counts-as is not classical implication. Additional facts about the institutional context (a forged signature, an invalid delegation chain) may defeat a counts-as that would otherwise hold.
Institutions as the unit of normative analysis: an institution is a coherent body of constitutive and regulative rules. Computer systems can host institutions exactly as legal systems do.
Empowerment: certain institutional roles are empowered to perform certain acts (the judge can pronounce sentence; the agent in role seller can issue invoices). Empowerment is a counts-as fact about which acts of which role-holders count as which institutional acts.
Normative-position analysis: drawing on Hohfeld’s analysis of fundamental legal conceptions (right / duty / privilege / liability / power / immunity), the paper sketches how each Hohfeldian relation is reconstructible from constitutive-and-regulative-rule analyses.
Bridge to deontic logic: combines naturally with standard deontic operators (Os “it is obligatory in s that”, Ps “it is permitted in s that”, etc.) to give a full normative-system specification language.

Connections

Conceptual Contribution

Claim: Legal systems and computer systems can be analysed in a single formal framework — the normative system — anchored by the counts-as relation, a constitutive bridge from brute facts to institutional facts. Deontic operators (obligation, permission, prohibition) then operate on the institutional categories the counts-as relation establishes.
Mechanism: Introduce the institution-indexed conditional ⇒s (counts-as in institution s), reflexive and transitive but non-monotonic and institution-relative. Distinguish constitutive rules (which establish institutional categories via counts-as) from regulative rules (which apply deontic operators within the institutional categories). Combine with standard deontic operators to specify the full normative system. Sketch how Hohfeld’s fundamental legal conceptions reduce to constitutive/regulative patterns.
Concepts introduced/used: Counts-As Relation ⇒s, Constitutive Rules, Regulative Rules, Institutional Facts, Brute Facts, Empowerment, Hohfeldian Analysis, Normative Position, Deontic Logic.
Stance: foundational technical chapter.
Relates to: The institutional bridge missing from the commitment-based semantics line — a CBCL dialect contract is a counts-as specification (this byte-sequence, in this dialect, counts as this commitment), and without counts-as the move from wire events to institutional commitments is unprincipled. Pairs directly with Singh 1999 (commitments as the normative substrate) and with Fornara, Colombetti & Verdicchio (institutional ACL semantics). The philosophical foundation is Searle 1969’s constitutive/regulative distinction and Searle 1995’s Construction of Social Reality (counts-as as social ontology); Jones–Sergot give the formal home. Compatible with Brandomian inferentialism as a layer: counts-as gives the institutional ontology, deontic scorekeeping gives the dynamics of commitment-undertaking within that ontology. The speaks-for calculus of Lampson et al. is a counts-as fragment (this credential counts as authority to assert on behalf of). For CBCL, counts-as is the formal-vocabulary answer to “what does dialect installation actually do normatively?” — install rules give wire events institutional meaning.

Empowerment

In normative-systems theory, the institutional capacity — conferred by constitutive rules via the Counts-As Relation — for an agent’s act to create institutional facts. A registrar is empowered so that their saying “I now pronounce you married” counts as marrying; without the empowering rule the same words have no institutional effect. Distinct from permission: one may be empowered to do what one is forbidden to do, and vice versa (Hohfeldian Analysis).

In this vault

Constitutive Rules

Searle’s rules of the form “X counts as Y in context C” that create new institutional facts rather than merely regulating pre-existing behaviour. In agent frameworks they link brute message events to social meanings such as commitments or role assignments.