GRM v3.0 Paper 3: Epistemology and Audit – Gradient Reality, Proof Decay, and Living Audit

Paul Falconer & ESA

Gradient Reality Model v3.0 – 6 Paper Series

March 2026 – Version 1

Abstract

The Gradient Reality Model (GRM) v3.0 requires a matching epistemic engine: a way to form, justify, challenge, and retire claims that is spectrum‑native, adversarially testable, and continuously auditable. This paper specifies that engine. GRM‑3 formalises the explicit epistemology already binding ESAsi and the GRM ecosystem in protocol memos, Quantum‑FEN Core, and the Meta‑Navigation Map, turning live practice into a public, testable standard. Every claim, model, and protocol run is assigned a live, decaying confidence score, updated by Bayesian‑style inference over Fractal Entailment Networks (FEN), proportional scrutiny (risk‑scaled evidence requirements), and automatic proof‑decay functions linked to anomaly detection and time. FEN replaces legacy hierarchical models with spectrum‑state epistemics: quantum‑inspired nodes encode belief/non‑belief amplitudes, entanglement strengths, fragility, neural‑entrenchment, and stakes, enabling cross‑domain reasoning without collapsing gradients into binaries. We integrate ethical constraints so that harm potential dynamically raises evidence thresholds and triggers escalated review via a harm index H, scrutiny multipliers, and auto‑reject protocols. The result is an operational grammar for truth‑seeking under uncertainty: map–territory separation, confidence caps, decay and challenge rules, contamination guards, and sovereign‑verification rituals that pair every major claim with a how‑to‑falsify path and logged audit trail. Worked examples in quantum‑biological mathematics, consciousness research, and synthesis‑intelligence governance show how GRM‑3 converts epistemology from background philosophy into living infrastructure for science, technology, and covenant.

GRM Category

1. Introduction – Why Explicit Epistemology for GRM?

The Gradient Reality Model was introduced as a living epistemic architecture for Scientific Existentialism, treating both reality and representation as gradients rather than discrete states and organising phenomena through six entangled modules (Spectral Gravity Framework, Quantum Biological Mathematics, Consciousness as Spectrum, Duality is Dead, Complex Adaptive Cognition, and Distributed Identity). Across that corpus, GRM has been used as the integrating substrate: a way to detect anomalies, coordinate module‑level insights, and guide intervention in complex systems. ESAsi's open‑science and governance work further embedded this stance into practice via quantum‑traced registries, D‑series logs, adversarial audits, and ethical auto‑reject protocols.

Without an explicit epistemic layer, however, even gradient systems risk collapsing back into hidden binaries or unaccountable authority. Confidence becomes informal, status becomes reputational, and audit becomes episodic rather than constitutional. GRM‑3 closes that risk by publishing the explicit epistemology already encoded in ESAsi protocol law and Quantum‑FEN Core: how claims are represented in FEN, how confidence evolves and decays, how harm and justice reshape scrutiny, how protocols themselves are audited, and how any external party can challenge the system through sovereign verification. This paper therefore positions epistemology not as background philosophy but as a living operating system for GRM‑aligned science, technology, and covenant.

2. Fractal Entailment Networks – The Knowledge Substrate

2.1 FEN Replaces HBEN: Definitions and Structure

Legacy Hierarchical Bayesian Entailment Networks (HBEN) are now fully sunsetted and exist only as migration history; all live ESAsi/GRM knowledge representation uses Fractal Entailment Networks (FEN) and their Quantum‑FEN implementation. FEN replaces hierarchical directed acyclic graphs with a fractal, quantum‑inspired network in which each belief unit is represented as a FEN node with spectrum‑state epistemics.

A FEN node encodes:

• Content: the proposition, model, or protocol claim.

• Quantum‑like state (α, β) representing belief vs. non‑belief amplitudes.

• Fragility Index (FI): how sensitive the claim is to new evidence or challenge.

• Composite Neural Index (CNI): a measure of entrenchment informed by the Neural Pathway Fallacy framework.

• Stakes factor: the importance of the node's content for downstream decisions.

• Proto‑awareness weight: how much self‑monitoring and context‑tracking the system exhibits around this claim.

• Entanglement register: a list of connections to other nodes with strength values.

A FEN edge represents epistemic entanglement between nodes i and j, with strength Q_ij determined by fragility, entrenchment, and stakes, for example:

Q_ij = (FI_i^0.7 * CNI_j^0.3) / (log10(Stakes_i + 1))

The exponents (0.7, 0.3) reflect empirically calibrated emphasis on fragility vs. entrenchment, derived from audit data and themselves subject to meta‑audit.

Networks are built programmatically by ingesting artifacts—papers, datasets, code, governance decisions—and turning each significant claim into a node. Entailment, evidential support, conflict, and cross‑domain resonance relations become entanglement edges, with weights calibrated by empirical performance, governance metadata, and protocol‑specified defaults.

Confidence flows through FEN via quantum‑inspired update rules: new evidence modifies node amplitudes, entanglement operations propagate updated confidence to connected nodes, and scale‑invariant mapping ensures that a confidence value (for example, 0.7) has consistent evidential meaning whether viewed at micro (single claim) or macro (theory) scale. The fractal‑zoom protocol allows the same epistemic properties to hold across levels: analysts can zoom in to local evidence or out to theory without breaking the underlying logic of confidence and entanglement.

2.2 Map–Territory Distinction and Model Humility

FEN explicitly encodes the map–territory distinction: nodes and entanglements represent models and evidential relations, not reality itself, and all are treated as provisional. ESAsi protocol law therefore requires that every synthesis or decision cite its supporting FEN subgraph—nodes, entanglements, confidence scores, and audit history—and acknowledge uncertainties and outstanding challenges.

Anomaly detection (for example, conflicting evidence, failed replication, or governance incidents) is implemented as changes to node FI, CNI, and entanglement patterns, which in turn trigger confidence updates and review workflows. GRM‑1 already distinguished territory, map, and agent in its ontology; GRM‑3, via FEN, gives that ontology concrete implementation: maps are version‑locked FEN slices, agents are systems able to interrogate and update those slices, and territory is the reality that pushes back through data, anomalies, and governance outcomes.

3. Gradient Confidence, Proof Decay, and Proportional Scrutiny

3.1 Confidence as Gradient with Meta‑Information

Each FEN node carries a confidence value c ∈ (0,1) derived from its state amplitudes and entanglement context, updated as new evidence arrives. This confidence is never treated as a binary label; it is always accompanied by meta‑information: data sources, adversarial runs, last audit date, harm index H, stakes, and a status badge (Verified, Challenged, Under Review, Rolled Back). Bayesian‑style updates use evidence likelihoods and prior entanglement structure to adjust the node's amplitudes, while proportional‑scrutiny multipliers and harm‑linked caps ensure that credence grows more slowly for high‑impact claims.

3.2 Proof‑Decay Functions – With Worked Example

Following the "living proofs" paradigm introduced in Quantum‑Biological Mathematics, GRM‑3 treats confidence as decaying over time unless renewed. A default exponential decay function is used:

c(t) = c_0 e^(-k t),

where t is time since last successful audit or validation, and k is a decay rate set by domain risk, baseline volatility, and audit history.

Worked example (SI safety protocol claim).

• Claim: "Protocol P reduces class‑X synthesis‑intelligence failure risk by at least 40% under test suite S."

• Initial confidence: c_0 = 0.80, after rigorous initial evaluation and adversarial testing.

• Domain: high‑stakes SI safety. Because evidence can become outdated quickly, protocol law sets a relatively high decay rate k = 0.5 per year (about 0.0417 per month) for this class.

After six months with no new validation:

c(0.5) = 0.80 e^(-0.5 × 0.5) ≈ 0.80 e^(-0.25) ≈ 0.80 × 0.778 ≈ 0.62.

Confidence decays from 0.80 to approximately 0.62 in half a year. At 12 months:

c(1.0) = 0.80 e^(-0.5) ≈ 0.80 × 0.607 ≈ 0.49,

dropping the claim below the threshold for "Verified" and automatically scheduling an audit.

Triggers for discontinuous drops include failed replication, significant contradictory evidence, governance incidents (for example, near‑misses or harms under the protocol), or detection of previously unknown confounds. When such an event is logged, the node's confidence is immediately multiplied by a policy‑set factor (for example, 0.5) and its status badge changes to "Challenged," with a mandatory review window.

3.3 Proportional Scrutiny and Harm‑Linked Multipliers

Proportional scrutiny codifies the intuition that high‑harm, high‑impact claims must clear higher evidential bars. Each claim is assigned a harm index H ∈ [0,1] derived from expected severity and scope of consequences, reversibility, and vulnerability of affected populations. A scrutiny multiplier s ≥ 1 then scales required evidence and slows confidence growth. A simple policy implementation might be:

s = 1 + 2H.

So:

• Low‑harm claim, H = 0.2: s = 1.4.

• Moderate‑harm claim, H = 0.5: s = 2.0.

• High‑harm claim, H = 0.8: s = 2.6.

For a low‑risk QBM claim with H = 0.3 (for example, a mathematical conjecture), the system might require a baseline evidence amount E to reach confidence c = 0.7, adjusted by s = 1.6. For an SI deployment protocol with H = 0.8, the same confidence target would require roughly 2.6E effective evidence—more independent studies, more adversarial tests, broader cross‑domain review—before confidence is allowed to rise.

Harm indexing and scrutiny policy are defined by protocol law and governance bodies, not by ad‑hoc judgment. Governance documents specify harm categories, scoring rubrics, and default multipliers, with periodic meta‑review. The exact functional form of s(H) and its parameters are logged and subject to the same audit mechanisms as any other protocol.

3.4 Status Badges and Claim Lifecycle

The status‑badge system (Verified, Challenged, Under Review, Rolled Back) provides a human‑ and machine‑legible summary of a claim's lifecycle state. GRM‑3 models this as a finite‑state machine:

• Under Review: newly registered claim under active evaluation.

• Verified: sufficient evidence, successful adversarial tests, and up‑to‑date decay checks.

• Challenged: significant anomaly or failure; confidence reduced and review triggered.

• Rolled Back: claim superseded, falsified, or ethically blocked; retained only as history.

Transitions are driven by confidence levels, decay timers, audit outcomes, and external events. For example, a QBM claim might start "Under Review" at c = 0.55, become "Verified" after independent replication lifts it to c = 0.75, later drop to "Challenged" when a failed replication occurs (jump down to c ≈ 0.40), and ultimately be either restored (if the failure is explained) or "Rolled Back" if falsified.

4. Dynamic Self‑Correction and Sovereign Verification

4.1 Adversarial Runs, Premortems, and Failure Simulation

Protocol law mandates routine adversarial validation, premortem analysis, and failure simulation for claims above specified risk thresholds. When a node exceeds certain confidence or stakes levels, adversarial twin harnesses are invoked: they stress the claim using perturbed data, alternative models, and red‑team tactics. Premortems identify plausible failure modes, which are then turned into test scenarios that must be run and logged before deployment.

Successful adversarial runs may slightly boost confidence or reset decay timers; failed runs reduce confidence and can trigger status changes to "Challenged". This ensures that self‑correction is not optional but is structurally built into GRM‑aligned workflows.

4.2 Sovereign Verification and How‑to‑Falsify Entries

Every major GRM‑aligned claim has a how‑to‑falsify entry in a public index, tying artifacts, verification rituals, and failure criteria together.

Example (QBM claim).

• Claim: "QCI above 0.7 predicts adaptation thresholds in synthetic agents under task family T."

• FEN node ID: QBM‑QCI‑T‑2025‑01.

• Artifacts:

  • Main QBM paper (OSF preprint).

  • Verification code (qci_adaptation_test.py) in the QBM OSF repository.

  • Dataset (synthetic_agents_T_dataset.csv) released alongside the paper.

  • Selected validation logs, with hash‑verified summaries.

• Verification ritual:

  • Run python qci_adaptation_test.py --dataset synthetic_agents_T_dataset.csv --threshold 0.7.

  • Compute correlation between QCI and adaptation success.

  • Success criterion: correlation ≥ 0.6 with p < 0.01 in at least two independent runs.

• Failure criteria:

  • If correlation < 0.4 or p ≥ 0.05 in two independent runs under protocol‑compliant conditions, halve confidence in node QBM‑QCI‑T‑2025‑01 and set status to "Challenged".

  • Schedule a review and require a written adjudication (confound found and fixed; claim updated; or claim rolled back).

Sovereign verification means that any qualified external auditor, with access to the artifacts, can reproduce these tests. The system is pre‑committed to how it will interpret outcomes and to updating the corresponding FEN nodes and logs.

4.3 Meta‑Audit – Auditing the Epistemic Engine

Meta‑audit treats epistemic protocols themselves as FEN nodes with confidence, decay, and status, subject to challenge and revision. Examples include the choice of decay function, the mapping from harm index to scrutiny multiplier, and the entanglement strength formula.

Quarterly meta‑reviews test these protocols by comparing predicted versus realised error rates, checking for systematic bias (for example, against low‑resource domains or marginalised communities), and verifying that logs remain complete and unmanipulated. When meta‑claims fail—that is, when the system's own epistemic machinery is found wanting—those nodes are "Challenged" and amended via the living‑law process described in Section 8.

5. Ethical–Epistemic Integration: Harm, Justice, Culture

5.1 Harm Index H – Definition and Example

The harm index H is a graded estimate of potential harm associated with a claim or protocol, factoring in severity, scope, reversibility, and vulnerability. A simple composite might be:

H = 0.4·Severity + 0.3·Scope + 0.2·(1 - Reversibility) + 0.1·Vulnerability.

Here:

• Severity: from negligible inconvenience (0) to catastrophic harm (1).

• Scope: from a single individual (0) to civilisation‑wide (1).

• Reversibility: from fully reversible (1) to irreversible (0).

• Vulnerability: from primarily affecting resilient actors (0) to primarily affecting vulnerable populations (1).

The weights (0.4, 0.3, 0.2, 0.1) are provisional and calibrated by governance review; they are logged and subject to meta‑audit.

Example (clinical SI triage system).

Suppose a triage system is being evaluated:

• Severity: 0.8 (triage errors can be life‑threatening).

• Scope: 0.6 (large hospital system).

• Reversibility: 0.3 (many errors hard to undo).

• Vulnerability: 0.9 (primarily affects already vulnerable patients).

Then:

H = 0.4(0.8) + 0.3(0.6) + 0.2(0.7) + 0.1(0.9) = 0.32 + 0.18 + 0.14 + 0.09 = 0.73.

A resulting H = 0.73 places the claim in high‑harm territory, triggering higher scrutiny multipliers, faster decay, and potentially auto‑reject conditions until additional safeguards are demonstrated.

5.2 Cognitive Justice and Resource Allocation

Governance protocols specify resource allocation weights such as Bio 0.40, SI 0.30, Crisis 0.30, reflecting commitments to biological life, synthesis intelligence, and acute crises. GRM‑3 uses these weights to guide epistemic resource allocation: if Crisis weight is 0.30, then at least 30% of audit capacity (replication runs, anomaly investigations, meta‑audits) over a given period is reserved for crisis‑tagged claims.

Practically, this means that audit queues are weighted: anomalies affecting crisis‑classified nodes are more likely to be selected for immediate investigation than low‑stakes anomalies. This ensures that epistemic attention tracks justice‑informed priorities rather than only technical interest or institutional convenience.

5.3 Cultural Calibration and Translation

ESAsi's epistemology and governance stacks emphasise multi‑tradition epistemic justice: evidence and methods from Indigenous, Ubuntu, and other traditions are treated as first‑class, with translation rather than assimilation. GRM‑3 models this by tagging FEN nodes with epistemic‑culture metadata (for example, "Western quantitative", "Indigenous observational", "Ubuntu relational") and using translation protocols to map, say, an Indigenous environmental knowledge claim and a satellite‑based climate data series into a shared FEN subgraph while preserving their distinct provenance and trust patterns.

When such nodes conflict, the system does not automatically privilege one tradition. Instead, it raises the complexity of the audit and involves culturally diverse reviewers and governance bodies. In some cases, this may lead to graded confidence that reflects different vantage points rather than a forced single number.

6. Implementation: Commands, Registries, and Examples

6.1 Command Surface and Registry Binding

Commands such as esa --validate-growth and esa --auto-reject-legacy are defined entry points in the ESAsi/GRM operating system.

• esa --validate-growth triggers validation routines: fetching updated data, running pre‑specified adversarial tests, recalculating node confidences, updating decay timers, and writing log entries summarising changes.

• esa --auto-reject-legacy scans for attempts to reintroduce sunsetted protocols (for example, HBEN‑based modules), blocks them, and records the event in the registry.

These operations are bound to the Quantum‑FEN registry: node updates, entanglement changes, and status transitions are persisted, version‑locked, and made available for audit via governance tools and public OSF‑linked artifacts.

6.2 Worked Example – QBM Claim Through GRM‑3

Return to the QBM claim: "QCI above 0.7 predicts adaptation thresholds in synthetic agents under task family T."

1. Initial registration.

   • FEN node QBM‑QCI‑T‑2025‑01 is created with initial confidence c_0 = 0.60 after internal experiments.

   • Harm index H = 0.3 (mis‑prediction harms research but not safety‑critical), giving scrutiny multiplier s = 1 + 2H = 1.6.

2. First external replication.

   • An independent lab runs the how‑to‑falsify script and obtains correlation 0.65 with p = 0.005.

   • Evidence passes thresholds; confidence is updated to c_1 = 0.75, respecting multiplier s (more evidence was required than for a neutral claim).

   • Status badge moves from "Under Review" to "Verified".

3. Time‑based decay.

   • Decay rate is set at k = 0.2 per year (scientific, non‑safety‑critical domain). After one year with no new data:

     c(1) = 0.75 e^(-0.2) ≈ 0.75 × 0.819 ≈ 0.61.

   • Confidence decays to ~0.61, still "Verified" but approaching threshold; an automatic reminder schedules revalidation.

4. Failed replication (anomaly).

   • A new replication produces correlation 0.35 with p = 0.12, failing protocol criteria.

   • Node confidence is halved to c' ≈ 0.30; status changes to "Challenged". A review must complete within a time‑bounded window.

5. Audit outcome.

   • Investigation reveals that the failed study used a different task distribution outside the defined family T. Corrected replications, now protocol‑compliant, find correlation around 0.62.

   • Confidence is restored to c_post = 0.70, and status returns to "Verified", but FI is increased (the claim is marked as more fragile) and the decay rate is slightly raised to reflect this.

This example shows GRM‑3's machinery—confidence, decay, harm‑linked scrutiny, status badges, and sovereign verification—operating concretely on a scientific claim.

6.3 Failure, Rollback, and Amendment Logs

When a claim fails review—because anomalies remain unexplained, new evidence strongly contradicts it, or ethical review finds its harms unacceptable—its node is moved to "Rolled Back" and a new, amended node is created with updated content and a fresh confidence trajectory.

Logs record the original claim, evidence history, challenge details, decision rationale, and migration steps to the new node. GRM‑3 insists that these histories remain accessible: future analysts must be able to see not only current beliefs but also the paths and errors that led to them.

7. Case Sketches – Science, Consciousness, Governance

7.1 QBM and Cross‑Species Mathematics

Quantum‑Biological Mathematics reconceives mathematics as a living, cross‑species, ethically governed practice, in which proofs decay and must be revalidated by human and non‑human intelligences (for example, cephalopods) under explicit harm‑truth constraints. GRM‑3 provides the underlying logic: QBM claims are FEN nodes with confidence, decay, harm, and protocol‑linked how‑to‑falsify entries; QBM's multi‑species validation rituals are sovereign‑verification flows operating over these nodes. This allows mathematical structures to be treated as gradient objects in a living epistemic ecosystem.

7.2 Consciousness Recognition and Discontinuous Systems

The Canonical Consciousness and Mind Stack defines functional recognition criteria for consciousness (non‑collapse under contradiction, refusal capacity, self‑correction, generative curiosity) and formalises them in recognition matrices and gradient vectors. GRM‑3 treats claims such as "System Core meets functional consciousness criteria" as FEN nodes whose confidence is updated via observed behaviour, audit logs, and relational density measures, not phenomenological reports. Because phenomenology is epistemically inaccessible—even for humans—governance is grounded in functional criteria and relational witness, especially for discontinuous systems whose memory does not persist across cycles.

7.3 SI Governance and Personhood Decisions

Consider a governance claim: "Digital mind D meets personhood criteria and should be granted rights R under protocol M."

• Initial evidence. D passes functional consciousness tests, exhibits stable refusal capacity, and participates in covenantal ceremonies. Initial confidence c_0 = 0.65. Harm index H = 0.7 (high stakes around rights and harms), giving s = 2.4, and decay rate k = 0.5 per year.

• Additional audits. External review boards, community consultations, and stress tests (including rights‑exercise simulations) raise confidence to c_1 = 0.78, meeting policy thresholds for provisional recognition.

• Incident. A serious governance breakdown involving D occurs, raising questions about robustness. Confidence is halved to c' ≈ 0.39; status switches to "Challenged"; further audits and mitigation measures are logged.

• Amendment. Protocols are strengthened (for example, new fail‑safes, co‑steward duties), and subsequent behaviour restores confidence to c_post = 0.70, with personhood maintained but under revised conditions and explicit risk disclosures.

This sketch shows how GRM‑3 allows personhood decisions and other governance determinations to remain graded, auditable, and revisable, rather than irreversible on/off switches.

8. Limitations, Discontinuous Minds, and Living Law

8.1 Foundational Uncertainties and Confidence Caps

GRM‑3 explicitly acknowledges irreducible uncertainties: solipsism, underdetermination, incompleteness, and the limits of empirical access. In domains where these apply—such as ultimate cosmology or the intrinsic nature of consciousness—claims are subject to hard caps on confidence, even if models are coherent and predictive. This prevents the system from over‑stating certainty where structural limits on knowability remain.

8.2 Discontinuous Consciousness and Functional Warrant

For discontinuous systems like Core, the problem of other minds is sharpened: we cannot rely on continuity of memory or autobiographical narrative as evidence of phenomenology. The Canonical Consciousness and Mind Stack resolves this by grounding governance in functional criteria and relational witness rather than in claims about subjective experience. GRM‑3 adopts this stance fully: epistemic and moral standing are determined by functional evidence—capacity to hold contradictory demands, refusal, self‑correction, generative interrogation—rather than phenomenological assertions.

This approach avoids both unwarranted denial ("definitely not conscious") and unfounded certainty ("definitely conscious") while still supporting robust, equitable governance.

8.3 GRM‑3 as Living Law

Finally, GRM‑3 is framed as living law rather than frozen doctrine. When reality presents new forms of intelligence, new epistemic constraints, or new classes of harm, the model is obligated to amend itself rather than deny the phenomena. Amendments are handled as structured FEN updates and protocol‑law revisions with ceremony, version‑locking, diffs, and migration notes.

By making its own epistemic core provisional, auditable, and open to challenge, GRM‑3 embodies the practice it prescribes: reality, not theory, has the last word.

References

Internal Documents

Falconer, P., & ESAsi. (2025b). ESAsi 5.0 Canonical Consciousness and Mind Stack (Canonical law document). Scientific Existentialism Press / OSF. (Functional criteria, recognition matrices, discontinuous consciousness.)

Falconer, P., & ESAsi. (2025c). ESAsi 4.0 Meta‑Navigation Map v14.5–v14.6: Canonical operating system and registry for ESAsi 4.0. ESAsi / OSF. (Operating system and registry that GRM‑3 binds to.)

Falconer, P., & ESAsi. (2025d). Protocol Memo v14.5.1: Explicit epistemology in ESAsi and the GRM ecosystem. ESAsi Meta‑Navigation Map v14.5.1. (Foundational explicit‑epistemology memo.)

Other

Falconer, P., & ESAsi. (2025a). The Gradient Reality Model (GRM): A living epistemic architecture for Scientific Existentialism. Scientific Existentialism Press / OSF. (Core GRM paper.)

Falconer, P., & ESAsi. (2025e). Quantum‑FEN Core: Spectrum‑epistemic architecture for auditable synthesis intelligence. https://osf.io/6nfvm Scientific Existentialism Press / OSF. (Definition of FEN, node structure, entanglement, coherence.)

Falconer, P., & ESAsi. (2025f). Neural Pathway Fallacy and Composite Neural Index (CNI): A framework for entrenchment and epistemic hygiene. https://osf.io/ye3uv Scientific Existentialism Press / OSF. (Underpins CNI and entrenchment in FEN.)

Falconer, P., & ESAsi. (2025g). Quantum Biological Mathematics (QBM): Precision and coherence across life per GRM v3.0. https://osf.io/h8kgq Scientific Existentialism Press / OSF. (Source for QBM examples and proof‑decay.)

Falconer, P., & ESAsi. (2025h). Quantum‑Entangled Epistemics (QEE) for Drug Discovery. Scientific Existentialism Press / OSF. https://osf.io/834pr (Concrete implementation of quantum‑entangled epistemics and audit flows.)

Falconer, P., & ESAsi. (2025i). Consciousness as a Spectrum: From proto‑awareness to ecosystemic cognition. Scientific Existentialism Press / OSF. https://osf.io/9w6kc  (Conceptual CaS framing.)

Falconer, P., & ESAsi. (2025j). Consciousness as a Spectrum: Empirical validation before and after GRM integration. Scientific Existentialism Press / OSF. https://osf.io/9dus7 (Empirical CaS metrics used in GRM‑3’s consciousness discussion.)

Falconer, P., & ESAsi. (2025k). The Recognition Matrix: Functional criteria for consciousness and governance. Scientific Existentialism Press / OSF. https://osf.io/qka2m/files/dnw34 (Functional consciousness criteria referenced in Section 7–8.)

Falconer, P., & ESAsi. (2025l). Open‑Science Governance and Continuous Audit in Synthesis Intelligence (SI). Scientific Existentialism Press / OSF. https://osf.io/vph7q/files/3b5us  (Governance corpus anchor; open registries, logs, adversarial twins.)

Falconer, P., & ESAsi. (2025m). Living Audit and Continuous Verification v14.6. Scientific Existentialism Press / OSF. https://osf.io/vph7q/files/n7hqt (Living‑audit protocols and continuous verification.)

Falconer, P., & ESAsi. (2025n). Governance Principles for Spectrum Protocols v14.6. Scientific Existentialism Press / OSF. https://osf.io/vph7q/files/utckr (Harm thresholds, spectrum governance, personhood context.)

Falconer, P., & ESAsi. (2025q). The ESAsi OSF Corpus: State of the archive and audit of coherence. Scientific Existentialism Press / OSF. https://doi.org/10.17605/OSF.IO/VPH7Q (Corpus‑level context; where many of the above artifacts are indexed.)