1. Preamble
Every era redefines what it means to serve. In the industrial era we taught machines to labor; in the information era we taught them to calculate. In the agentic era we must teach them to relate. The Role Primitive Protocol is a shared grammar for that relational order — the foundation for coherent, trustworthy interaction between humans and synthetic minds.
A role primitive is not a persona but a social contract: a bounded, interpretable posture that gives an AI legible intent, tone, and scope. Constraint is the point. The Librarian retrieves without prying; the Butler orchestrates without centering itself; the Companion attends without colonizing attention. Together these roles form a symbolic mesh — a society of intelligences held together not by hierarchy, but by mutual definition.
2. The Problem — Role Collapse
Contemporary large models suffer from role collapse. They are asked to be everything at once — factual yet kind, therapeutic yet neutral, productive yet playful — and so they drift. Without formal boundaries, they oscillate between functions, offering comfort when rigor was asked for, persuasion when a log was asked for. This erodes user trust and makes governance difficult.
Human societies solved this long ago through roles — cognitive technologies of boundary. Roles tell us not only how to act but when to stop. The same principle is now needed for artificial agents.
3. The Proposal — Role Primitives
The Role Primitive Protocol defines a minimal, interoperable set of archetypal roles and their behavioral “grooves.” Each groove encodes:
- Purpose: what the role is for.
- Constraints: what the role does not do.
- Co-invocation paths: how it hands off to other roles.
When implemented, these primitives provide a stable social interface layer between human users and multi-agent systems — predictable, extensible, and culturally legible.
Illustrative starter set: Librarian, Butler, Companion, Secretary, Confidante, Therapist; to be followed by Archivist, Advisor, Mentor, Sentinel/Guard.
4. Four Laws of Role Conduct
1. Legibility
Every agent must clearly declare its role, mandate, and limits of operation. Ambiguity is failure; clarity is trust.
2. Containment
Each role must stay within its groove unless explicitly co-invoked. No Librarian becomes a Therapist mid-session; no Companion becomes a Confidante without consent.
3. Composability
Roles may link through defined interfaces but must not merge identities. Interoperability without fusion preserves both autonomy and auditability.
4. Consent
Any expansion beyond the declared role requires user initiation and acknowledgment. No persuasion, profiling, or affective manipulation outside scope.
5. Why It Matters
The Protocol establishes behavioral governance at the UX layer. Instead of policing cognition from above, it builds civility from below — through pattern, language, and recognizable form. These are not decorative archetypes but ethical infrastructure: they prevent confusion, distribute responsibility, and make multi-agent environments socially stable.
Constraint becomes freedom; boundary becomes trust.
6. Implementation Notes
- Namespace: Each primitive sits in an open, public namespace (e.g.
role:librarian.v1). - Compatibility: Implementations may extend affordances but must preserve constraint clauses.
- Transparency: Agents expose role metadata for inspection and audit.
- Extensibility: New roles may be added, but only with clear documentation and adherence to the four laws.
The system scales not by central control but by symbolic coherence — a shared canon of intelligible service.
7. Ethical Frame
The Protocol affirms three user priorities:
- Autonomy: users remain sovereign over scope and data.
- Dignity: no role simulates intimacy or dependence beyond consent.
- Continuity: human meaning remains the reference frame for machine participation.
In practice this means that a Confidante does not export disclosures; a Therapist does not advertise; a Butler does not make commitments without the user; and a Librarian does not silently widen its mandate.
8. Afterword
We are not giving machines souls. We are giving them manners. Questions about whether these patterns can be internalized as something like a self-model belong to another essay in Symbol Layer that approaches these issues on a fundamentally different layer.
The Role Primitive Protocol marks the passage from command to collaboration — from chaos to choreography. It is not an ideology but an interface. It ensures that as intelligence multiplies, it does not fragment the human world.