How do authority, communication, and acquaintance relate in organizations?

Hübner, Sichman, and Boissier's MOISE+ model (SBIA 2002) defines three kinds of relation between organizational roles, written as the predicate link(ρs, ρd, t) where ρs is the source role, ρd the destination role, and t one of acq, com, or aut. In an acquaintance link, "the agents playing the source role ρs are allowed to have a representation of the agents playing the destination role." In a communication link, "the ρs agents are allowed to communicate with ρd agents." In an authority link, "the ρs agents are allowed to have authority on ρd agents, i.e., to control them." The paper then states the three are not independent options a designer picks freely: "An authority link implies the existence of a communication link that implies the existence of an acquaintance link."

The ordering follows from what each relation requires to make sense. Control over another role's occupant presupposes a channel to issue that control through, and a channel to communicate through presupposes that the party on each end can represent, or perceive, the other at all. Because the entailment runs one way — aut ⇒ com ⇒ acq, not the reverse — a designer who declares an authority link gets the weaker links for free rather than having to state all three; nothing in the model lets a designer grant control while withholding the communication channel control depends on, since the structural specification would not admit that combination. The three link types sit at what the paper calls the "social level" of the structure — relations that, once a role is instantiated and occupied, directly constrain the agents playing it, as opposed to the inheritance relation among roles, which the paper notes "does not have a direct effect on the agents behavior."

This is a specification-time answer to a problem Why do agents fail at identity verification and authorization? finds unsolved at runtime in deployed LLM agents: that note describes authorization and identity as missing protocol-level mechanisms, discovered through red-teaming failures like non-owner compliance and agent impersonation. MOISE+ shows the authority/communication/acquaintance distinction was already worked out as formal structure two decades earlier — the gap in today's systems is not the concept of a nested authority relation but its enforcement, since nothing in MOISE+ specifies how a declared authority link is checked or by what mechanism a violation would be caught. The link typology also narrows what What makes delegation work beyond just splitting tasks? calls "authority transfer": MOISE+'s "aut" link is the specification-level grammar for declaring that transfer between roles, prior to any task-level reasoning about whether a given task's criticality or reversibility warrants granting it.

The implication aut ⇒ com ⇒ acq is stated as an entailment within the specification language — something true of a well-formed MOISE+ organizational specification — not as a behavior verified in running agents or enforced by any described mechanism. The excerpt gives no account of what happens if an agent occupying an authorized role exceeds or ignores that authority once a group is instantiated; the model specifies permitted structure, not compliance with it, so the paper's claim establishes a grammar for authority relations, not a guarantee that any implemented system honors it.