O0-RP-005

Operational Definitions: Substrate, Boundary, Agent, Observer

**Version:** 1.0

**Research status:** APPROVED

CLAIM STATUS: METHODOLOGICAL
EVIDENCE TYPE: DEFINITIONAL STANDARD
PHYSICAL VALIDATION: NONE
INDEPENDENT REPLICATION: N/A
PHILOSOPHICAL PROVENANCE: O/0 ARCHIVE
ARCHIVE ENDORSEMENT: LIMITED TO REPORTED RESULT

Abstract

This document provides precise, measurable operational definitions for the four central terms used throughout the O/0 Research Program: substrate, boundary, agent, and observer. Each definition specifies measurement criteria, boundary conditions for applicability, explicit exclusions (what the term does NOT mean), and the formal relationship between the operational definition and any informal usage in the philosophical archive. These definitions are binding across all research records and ensure that claims made in one document are commensurable with claims in another. Without such definitions, the program risks equivocation between the philosophical and scientific uses of its core vocabulary.

Source proposition

The O/0 archive uses terms like "substrate," "boundary," "agent," and "observer" in a phenomenological and sometimes poetic register. The research program requires that each term be stripped of its contemplative connotations and defined solely in terms of measurable quantities.

Scientific audit

The terms "substrate," "boundary," "agent," and "observer" each carry heavy philosophical baggage and are used inconsistently across disciplines. In physics, "observer" has a specific meaning in quantum mechanics that differs from its meaning in relativity, which differs again from its meaning in information theory. The operational definitions below resolve this ambiguity for the specific context of the O/0 Research Program.

Research question

What minimal, measurable, discipline-neutral definitions of substrate, boundary, agent, and observer are sufficient to support the program's computational research while avoiding philosophical overcommitment?

Operational definitions

1. SUBSTRATE

**Definition:** A substrate is a computational or mathematical structure S = (X, T, U) consisting of a state space X, a topology T defined over X, and an update rule U: X → X that determines the temporal evolution of the system. The substrate is characterized by the following properties:

• **Continuity**: The topology T is connected (there exists a path between any two points in X).
• **Homogeneity**: The update rule U is translationally invariant (no point in X is privileged by U itself, though initial conditions may break this symmetry).
• **Determinism or stochasticity**: U may be deterministic or stochastic; if stochastic, the noise distribution must be specified.

**Measurement criteria:**

• Continuity is verified by confirming that the adjacency graph of the discretized state space is connected.
• Homogeneity is verified by confirming that U produces identical outputs for identical local configurations regardless of absolute position.
• The substrate exists as a formal object and is fully specified by its source code and parameters.

**Boundary conditions for applicability:**

• The definition applies only to computational models within the program. It does not claim that physical reality has these properties.
• Substrates may be finite (grid-based) or infinite (analytically defined), but all simulations operate on finite approximations.

**What SUBSTRATE does NOT mean:**

• It does NOT mean "the fundamental stuff of reality."
• It does NOT mean "consciousness" or "awareness."
• It does NOT mean "the Void" or "O" as used in the philosophical archive.
• It does NOT carry ontological commitment. It is a formal object in a model.
• It is NOT claimed to be isomorphic to any physical system unless explicitly demonstrated.

2. BOUNDARY

**Definition:** A boundary B within a substrate S is a region of state space where a measurable gradient exceeds a specified threshold θ. Formally: B = {x ∈ X : ||∇φ(x)|| > θ}, where φ is a scalar field derived from the system's state (e.g., density, activity level, information content) and θ is a researcher-specified threshold that must be reported in all publications.

**Measurement criteria:**

• Boundaries are detected computationally by calculating the spatial gradient of a specified observable φ.
• The threshold θ must be justified relative to the system's noise floor (typically θ > 3σ of the background gradient distribution).
• Boundary persistence is measured in units of system time-steps. A boundary is considered "stable" if it persists for more than τ_min time-steps (researcher-specified, reported).

**Boundary conditions for applicability:**

• This definition applies to spatially extended substrates where gradient operations are well-defined.
• For graph-based substrates, the gradient is replaced by the difference between adjacent node states.
• The definition is observer-relative in the sense that different choices of φ and θ yield different boundary sets.

**What BOUNDARY does NOT mean:**

• It does NOT mean "an ontological division between self and world."
• It does NOT mean "separation" in any metaphysical sense.
• It does NOT imply that the boundary has causal power independent of the substrate dynamics.
• It does NOT mean the system is "divided" — it means a measurable gradient exists.
• It is NOT a binary property; boundaries have degrees (gradient magnitude) and durations (persistence).

3. AGENT

**Definition:** An agent A within a substrate S is a spatially localized region R ⊂ X that satisfies all of the following criteria simultaneously for a duration exceeding τ_agent:

• **Persistence**: The region R maintains structural coherence (measured by autocorrelation > ρ_min across successive time-steps).
• **Dissipation**: The region R exhibits net entropy export to its surroundings (measured by comparing internal entropy change to boundary entropy flux).
• **Information asymmetry**: The mutual information I(R_interior; R_exterior) is significantly greater than expected by chance (p < 0.01 against a shuffled null model).
• **Active boundary maintenance**: Perturbations to the boundary of R are followed by restorative dynamics (measured by return-to-baseline statistics after controlled perturbation injection).

**Measurement criteria:**

• Each criterion above is quantified by a specific statistical test with a pre-registered significance threshold.
• Agent detection is performed algorithmically with no human judgment in the classification loop.
• All thresholds (ρ_min, τ_agent, significance levels) must be reported and justified.

**Boundary conditions for applicability:**

• This definition is deliberately minimalist. It identifies agent-like behavior without requiring goal-directedness, internal models, or consciousness.
• Systems satisfying these criteria include biological cells, convection cells, and certain software processes. The definition is intentionally broad.

**What AGENT does NOT mean:**

• It does NOT mean "conscious being" or "sentient entity."
• It does NOT mean "intentional system" (no goals or beliefs are attributed).
• It does NOT mean "observer" (agent and observer are distinct; see below).
• It does NOT imply moral status or ethical consideration.
• It does NOT mean "self" in the phenomenological or Buddhist sense.
• It is a functional classification, not an ontological category.

4. OBSERVER

**Definition:** An observer O is an agent (satisfying all agent criteria above) that additionally satisfies:

• **Internal model**: O contains a subsystem M ⊂ R whose states correlate with external states beyond O's boundary with a mutual information I(M; E) significantly above chance, where E denotes the environment beyond O's boundary.
• **Model-environment decoupling**: The internal model M can be in states that do not correspond to current environmental states (i.e., M has representational capacity beyond direct mirroring — it can represent counterfactuals, memories, or predictions).
• **Asymmetric information flow**: Information flows preferentially from environment to model (observation) rather than from model to environment (action), as measured by transfer entropy T_{E→M} > T_{M→E} during observation episodes.

**Measurement criteria:**

• Internal model identification uses information-geometric methods to detect low-dimensional manifolds within the agent's state space.
• Model-environment decoupling is tested by holding the environment constant and measuring whether M still exhibits autonomous dynamics.
• Transfer entropy is computed using standard time-series methods with appropriate embedding dimensions and lag selection.

**Boundary conditions for applicability:**

• This definition identifies systems that functionally observe, not systems that subjectively experience observation.
• A thermostat trivially satisfies weak versions of these criteria. The thresholds must be set to exclude trivial cases while including interesting ones.

**What OBSERVER does NOT mean:**

• It does NOT mean "conscious observer" or "experiencing subject."
• It does NOT mean "observer" in the quantum mechanical sense (wavefunction collapse).
• It does NOT imply phenomenal awareness, qualia, or subjectivity.
• It does NOT mean "the observer" as used in contemplative traditions.
• It is a functional-information-theoretic classification, not a claim about the nature of experience.

Hypothesis

Not applicable. This is a definitional governance document.

Null hypothesis

Not applicable.

Competing explanations

Alternative definitions exist in the literature (e.g., autopoiesis for agents, Bayesian brain for observers). The definitions above are chosen for their measurability and computational tractability, not for their philosophical completeness. They are not claimed to be the only valid definitions, only the standard ones for this program.

Formal model

The four definitions form a nested hierarchy: Substrate ⊃ Boundary ⊂ Agent ⊂ Observer, where each successive level adds measurement criteria. Formally:

• Every observer is an agent: O ⊂ A.
• Every agent has boundaries: ∀a ∈ A, ∃B such that B = ∂R_a.
• All structures exist within a substrate: A ⊂ S, B ⊂ S.

Methods

Definitions were constructed by surveying relevant literature (dynamical systems theory, information theory, theoretical biology, artificial life), identifying the minimal measurable criteria for each concept, and stripping away all properties that require subjective judgment or philosophical commitment.

Controls

Each definition was tested against edge cases:

• Does a rock satisfy "agent"? No (fails dissipation and active boundary maintenance).
• Does a candle flame satisfy "agent"? Borderline (passes persistence and dissipation, marginal on information asymmetry).
• Does a thermostat satisfy "observer"? Marginally (passes weak model criterion, fails decoupling).

These edge cases verify that the definitions have discriminating power.

Predictions

Consistent application of these definitions will eliminate ambiguity in cross-document claims and enable automated consistency checking across the research program's outputs.

Falsification criteria

A definition is inadequate if: (a) it classifies systems that researchers unanimously agree should not be classified (e.g., classifying empty space as an agent), or (b) it fails to classify systems that researchers unanimously agree should be classified (e.g., failing to classify a living cell as an agent given appropriate substrate representation).

Results / Expected Outcomes

Four operational definitions have been specified with full measurement criteria, explicit exclusions, and formal relationships. These definitions are now binding across all O0-SIM, O0-MATH, and O0-OBS records.

Uncertainty

Threshold values (θ, τ_min, ρ_min, τ_agent) are necessarily somewhat arbitrary. Different threshold choices will yield different classification results. All analyses must report sensitivity to threshold selection.

Limitations

These definitions are designed for computational systems. Extending them to physical or biological systems requires additional bridging assumptions that are not specified here. The definitions are also silent on phenomenal consciousness by design — this is a feature, not a limitation.

Replication status

Governance document. Definitions have been applied in O0-SIM-001 through O0-SIM-005 and found to be consistently applicable.

Data and code

A reference implementation of the agent-detection algorithm (Python, NumPy/SciPy) is maintained in the program's code repository.

Relationship to philosophical archive

The O/0 archive uses "substrate," "boundary," "agent," and "observer" in rich phenomenological senses that these operational definitions deliberately do not capture. The research program acknowledges this reduction and does not claim that the operational definitions exhaust the meaning of these terms. Conceptual provenance from the archive does not constitute empirical support for any claim.

References

• Maturana, H. & Varela, F. (1980). Autopoiesis and Cognition.
• Friston, K. (2013). Life as we know it. Journal of the Royal Society Interface.
• Schreiber, T. (2000). Measuring information transfer. Physical Review Letters.
• Tononi, G. & Sporns, O. (2003). Measuring information integration.
• Beer, R.D. (2014). The cognitive domain of a glider in the Game of Life.
• Krakauer, D. et al. (2020). The information theory of individuality.

Revision history

• v1.0: Initial document generated.