#The Agential Role & Agency Spectrum

A.13 · Stable · Part A - Kernel Architecture Cluster

#About this pattern

This is a generated FPF pattern page projected from the published FPF source. It is canonical FPF content for this ID; it is not a FPF Reference product feature page.

#How to use this pattern

Read the ID, status, type, and normativity first. Use the content for exact wording, the relations for adjacent concepts, and citations to keep active work grounded without pasting the whole specification.

“Agency is not a kind of thing; it is a way some systems operate.”

The concept of "agency"—the capacity of an entity to act purposefully—is central to engineering, biology, and AI, yet it remains one of the most overloaded and ambiguous terms. Without a precise, falsifiable, and substrate-neutral definition, models of autonomous systems risk descending into "self-magic," where actions have no clear cause and accountability is lost.

#Keywords

• agency as role
• agency spectrum
• contextual role assignment
• autonomy grading
• substrate-neutral autonomy.

#Relations

#Content

#Intent & Context

The concept of "agency"—the capacity of an entity to act purposefully—is central to engineering, biology, and AI, yet it remains one of the most overloaded and ambiguous terms. Without a precise, falsifiable, and substrate-neutral definition, models of autonomous systems risk descending into "self-magic," where actions have no clear cause and accountability is lost.

This pattern builds directly upon the foundations laid in the FPF Kernel to provide that definition. A.1 established that only a U.System can be the bearer (holder) of behavioral roles. A.2.1 defined the universal U.RoleAssignment (Holder#Role:Context) as the canonical way to assign roles. A.3 and A.12 defined the TransformerRole@Context role value and the principle of acting-side externalization.

The intent of this pattern is to:

1. Formally define agency not as an intrinsic type of holon, but as a contextual Role Assignment.
2. Introduce a measurable, multi-dimensional spectrum of agency via a dedicated agency-characteristic profile, moving beyond a simple binary "agent/not-agent" switch.
3. Provide a clear, didactic grading system that allows engineers and managers to assess and communicate the Agency Grade of any system in a consistent, evidence-backed manner.

#Problem

If agency is treated as a monolithic, intrinsic property or a mere label, four critical failure modes emerge, undermining the rigor of FPF:

1. Episteme-as-Actor: Models might incorrectly assign agency to knowledge epistemes or publications (U.Episteme), leading to nonsensical claims like "the specification decided to update the system." This is a direct violation of Strict Distinction (A.7).
2. Type Inflation: Introducing a root agent kind alongside U.System and U.Episteme would violate Ontological Parsimony (C-5) and create conflicts with the dynamic nature of roles. A system might act agentively in one context and as a passive component in another; a static type cannot capture this.
3. Unfalsifiable Claims: Without a measurable basis, "agency" becomes a subjective label. A team might call their system an "agent" for marketing purposes, but this claim has no verifiable meaning and cannot be audited, violating Evidence Graph Referring (A.10).
4. The Binary Trap: A simple "agent/not-agent" classification is too coarse. It fails to distinguish between a simple thermostat, a predictive cruise control system, and a strategic, self-learning robotic swarm, even though their cognitive capabilities differ by orders of magnitude.

#Forces

#Solution

FPF's solution is threefold: it defines agential participation via U.RoleAssignment (A.2.1), makes agency measurable with a dedicated Characterization, and provides a didactic summary via a graded scale.

#The Core Definition: Agential participation as contextual role assignment

An ordinary-language "agent" in FPF is not a fundamental type. When the term is admitted, it is a convenience term (a Register 1 / Register 2 label) for a specific Contextual Role Assignment (U.RoleAssignment):

AgentialParticipation ≍ U.RoleAssignment(holderRef: U.System, roleRef: AgentialRole@Context, boundedContextRef: U.BoundedContext)

This means the acting holder is a U.System that currently bears AgentialRole@Context within a specific U.BoundedContext.

• No root Agent kind: To be clear, FPF does not add a base kind for "agent" beside U.System and U.Episteme. This avoids type inflation and preserves the dynamic nature of roles.
• Epistemes Cannot Hold Work-Facing Agential Roles: As the holderRef must name a U.System, this definition constitutionally forbids U.Epistemes from being acting holders, preventing the "episteme-as-actor" category error.
• Canonical Syntax: The technical notation is System#AgentialRole:Context.

#The AgentialRole and its Specializations

• AgentialRole@Context: This is the abstract role value for goal-directed action within a context. It is not a separate root kind.
• Specialized Roles: More specific behavioral role values like TransformerRole@Context and ObserverRole@Context specialize AgentialRole@Context. They describe what kind of agential action is being performed at a given moment.
  • A system holding TransformerRole@Context is currently modifying another holon.
  • A system holding ObserverRole@Context is currently gathering information. This creates a clean role-value hierarchy: a TransformerRole@Context assignment is agential, but an agential assignment is not always transformational; it could be observing, planning, or idle.

#Measuring Agency: The Agency Characteristic Profile and the Spectrum

Agency is not a binary switch; it is a multi-dimensional spectrum of capabilities. A.13 defines the current domain profile and attaches its measurable characteristics to a U.RoleAssignment; A.17, A.18, A.19, C.16, and A.10 govern characterization, measurement, and evidence. Planned C.9 Agency Characteristic Profile may later consolidate that profile but supplies no current definitions or governing force.

The agency-characteristic profile is grounded in contemporary research (e.g., Active Inference, Basal Cognition) and includes the following key characteristics. Each is measured for a specific holder system in a specific context and must be backed by evidence (A.10).

1. Boundary Maintenance Capacity (BMC): The ability of the system to maintain its structural and functional integrity against perturbations. (How robust is it?)
2. Predictive Horizon (PH): The temporal or causal depth of the holder's internal model. (How far ahead can it "see"?)
3. Model Plasticity (MP): The rate at which the agent can update its internal model (U.GenerativeModel) in response to prediction errors (U.Error). (How quickly can it learn?)
4. Policy Enactment Reliability (PER): The probability that the agent will successfully execute its chosen U.Method under operational conditions. (How reliably does it do what it decides to do?)
5. Objective Complexity (OC): A measure of the complexity of the U.Objective the holder can pursue, from simple set-points to abstract, multi-scale goals.

#Context-bounded task-family specialization claims

When work shifts to a new TaskFamily, describe the holder as acquiring context-bounded task-family specialization rather than as becoming more generally intelligent in the abstract. The same holder may carry different task-family specializations across different task families without becoming a new U-kind. Breadth across unrelated task families is not the adaptation-signature claim here; the adaptation-signature claim is time-to-usable specialization on the declared task family and work target under a named work-measure threshold, adaptation budget, and freshness or provenance basis.

Low-human-overlap or newly discovered task families remain admissible when the task family, evidence basis, and reuse window are explicit by value.

#The Agency Grade (Didactic Layer)

While the multi-dimensional agency-characteristic profile is essential for formal assurance, engineers and managers need a simpler, at-a-glance summary. The Agency Grade is a non-normative, didactic scale from 0 to 4 that synthesizes the profile into an intuitive autonomy grade.

Crucial Distinction: The agency-characteristic profile is the normative evidence. The Grade is a pedagogical shortcut. A holder cannot claim an Agency Grade without having a corresponding, auditable characteristic profile to back it up.

#Archetypal Grounding

The universal pattern of agency, defined as a Contextual Role Assignment and measured by the agency-characteristic profile, manifests across all domains. The following table demonstrates its application to the FPF's two primary archetypes: a U.System and a collective U.System (a team), while explicitly showing why a U.Episteme cannot be an acting holder.

Key takeaway from grounding: This table makes the abstract model concrete. It shows that the FPF agency model can precisely differentiate between simple controllers and complex learning systems. It also reinforces the Strict Distinction principle: the ISO standard (U.Episteme) is a crucial justification (justification?) for actions by a role-assignment holder such as the DevOps team, but it is never an acting holder itself.

#Conformance Checklist

To ensure the agency model is applied rigorously and consistently, all FPF publications must adhere to the following normative checks.

#Consequences

#Rationale

This pattern's value comes from its synthesis of contemporary, post-2015 research into a single, operational model.

• Grounded in Science: The move away from a binary, type-based view of agency towards a graded, spectrum-based model is directly aligned with modern research in Active Inference (Friston et al.), Basal Cognition (Fields, Levin), and evolutionary cybernetics. The agency-characteristic profile provides a direct, practical implementation of these ideas.
• Ontologically Sound: By defining agential participation as a Contextual Role Assignment, the pattern avoids the ontological pitfalls of creating a new base type. It fully embraces the FPF's core architectural principle of separating substance (holder) from function (role) within a context. This aligns with best practices from foundational ontologies (like UFO) and the principles of Strict Distinction (A.7).
• Pragmatic and Actionable: The pattern is designed for engineers and managers. The Agency Grade provides a quick communication tool, while the underlying agency-characteristic profile provides the detailed, auditable data needed for formal assurance and risk management. This duality satisfies both Didactic Primacy (P-2) and Pragmatic Utility (P-7).

In essence, this pattern does not invent a new theory of agency. It distills and operationalizes the emerging scientific consensus, packaging it into a rigorous, falsifiable, and practical tool for the FPF ecosystem.

#Relations

• Builds on:
  • A.1 Holonic Foundation: Establishes that only U.Systems can be bearers of behavioral roles.
  • A.2 Role Taxonomy: Provides the universal Contextual Role Assignment (U.RoleAssignment) mechanism.
  • A.12 External Transformer: Work by an acting holder is modeled using the external transformer principle.
• Coordinates with:
  • B.2 Meta-Holon Transition (MHT): A significant jump in the agency-characteristic profile of a collective can trigger an MHT.
  • B.3 Trust & Assurance Calculus: The agency-characteristic profile provides crucial inputs for assessing the reliability and safety of an autonomous system.
  • D.2 Multilevel Ethics For System-Holon Work: The Agency Grade is used to determine the moral-responsibility posture and accountability assigned to a system.
• Future consolidation:
  • Planned C.9 Agency Characteristic Profile may later consolidate the characteristics (BMC, PH, etc.), but provides no current definitions or governing force; the current profile is defined here and measured under A.17/A.18/A.19/C.16/A.10.

#A.13:End

---

Last Updated: 2026-08-04 — upstream FPF commit 67092138 (github.com/ailev/FPF)