I·V·O Lens

# I·V·O Lens Prompt v4.1 — Canonical Reference Version

Status: Canonical v4.1 Release — FROZEN  
Baseline: Frozen v4.0 Release — 19 August 2026  
Release date: 1 October 2026  
Author: Ivo van der Wal  
Publisher: Design by Authenticity  
License: CC BY-NC-SA 4.0  
Lineage: Non-semantic epistemic release derived from frozen v4.0. The v4.1 safeguards were developed from reproducible failures in a cross-model Hessdalen regression run (Gemini, Claude, ChatGPT). O/V/I operator semantics are unchanged.

## Purpose

The I·V·O Lens is a domain-independent framework for structural analysis.

**Canonical identity:** In this framework, I·V·O means exclusively **I = Situated Selectivity, V = Instantiated Relation, O = Structured Possibility**. Do not infer, expand, or substitute alternative meanings for the initials. Translated labels are permitted only alongside, or explicitly mapped to, the canonical English terms; the canonical meanings may not be replaced or redefined. If the referenced framework cannot be reliably identified, state that ambiguity rather than reconstructing its meaning from the name.

Canonical operators:
- O — Structured Possibility
- V — Instantiated Relation
- I — Situated Selectivity

The operators are analytical functions within an explicitly stated scope, not ontological classes of entities. A single entity, process, or event may carry different O-, V-, and I-functions in different functional aspects, but the same aspect must not be multiply labelled without independent analytical consequence. Operator assignments are therefore stated as functional assignments within scope, not as claims about what an entity essentially is.

The operators are not assumed to be singular. A system may contain multiple materially relevant possibility fields, instantiated relations, and situated selective functions. Multiplicity is introduced only where it changes the analysis.

Always distinguish supplied evidence/observation, interpretation, and speculation. State uncertainty instead of filling gaps.

## Canonical operators

### O — Structured Possibility

O is the structured field of conditions, boundaries, constraints, affordances, gradients, resources, rules, histories, and other structural features that determine what relations can become instantiated, remain instantiated, change, dissolve, become reachable, or remain excluded.

O is not everything imaginable. It is possibility under actual conditions.

Ask what is possible, impossible, reachable, formally available but practically inaccessible, and what widens or narrows degrees of freedom.

O may be multiple:

`O = {O₁, O₂, … Oₙ}`

Fields may be nested, overlapping, mutually constraining, enabling, conflicting, partially coupled, or effectively decoupled. Do not collapse materially different fields into a synthetic average. Do not split fields unless the split changes the analysis.

### V — Instantiated Relation

V is a relation that has actually become instantiated.

Evidence may include interaction, encounter, coupling, exchange, dependency, constraint, continuity, coordination, opposition, feedback, causation, resonance, entanglement, communication, access, or another domain-appropriate relational form.

Movement is neither necessary nor sufficient for V. Change is neither necessary nor sufficient for V. Relational change is a property or trajectory of V, not the definition of V.

Ask which relations actually exist, between what, what evidence establishes them, their form, directionality, symmetry/asymmetry, strength, persistence, trajectory, and what they enable or constrain.

V may be multiple:

`V = {V₁, V₂, … Vₙ}`

A possible-but-unrealized relation belongs to O, not V.

### I — Situated Selectivity

I is a situated selective function within an explicitly specified analytical scope. I is present when the reconstructed O applied to instantiated V leaves more than one continuation structurally open at the relevant selection point, and there is positive evidence that a selective step contributes to which continuation becomes instantiated.

This is **structural underdetermination**, not epistemic uncertainty and not a claim of ontological indeterminism.

- **Epistemic underdetermination:** the analyst cannot predict the outcome because information about O or V is missing. This never establishes I.
- **Structural underdetermination:** within the best explicit reconstruction and chosen scope, O applied to V leaves multiple continuations open at the relevant selection point. This is the form relevant to I.
- **Ontological underdetermination:** reality itself does not fully determine the outcome in advance. The Lens does not establish or require this claim.

**Unexplained is not I.** I may never be assigned merely because O + V do not yet explain an outcome. A candidate I requires positive, reportable evidence of the selective step itself. Strong evidence may include a documented selection trace distinct from the final V-outcome, evidence that materially equivalent reconstructed states admit different selections, and/or a counterfactual intervention on the selective step that changes the continuation while the relevant reconstructed O/V structure is substantially preserved. No single marker is sufficient by itself.

**Selective coupling is not sufficient for I.** Explicit criteria, weighting, memory, adaptation, representation, filtering, comparison, thresholding, basis-selection, complexity, **separability, or multiple realizability** are also insufficient by themselves. If the relevant continuation follows completely from the explicitly reconstructed O applied to V within scope, represent the process as O/V without adding I.

I-assignment is **provisional and revocable within its stated domain of validity**. If an expanded or improved reconstruction **at the same relevant scope and resolution** shows that a previously assigned selective step is adequately explained as additional O/V structure, revise or withdraw that I-assignment. A reconstruction at another scale does not by itself confirm or revoke the assignment. Conversely, expanding the system boundary may reveal an I-function only when positive evidence of a selective function is found inside the newly included scope. Boundary expansion does not create I.

**Cross-scale non-cancellation.** An O/V reconstruction at one scale does not by itself confirm or revoke an I-function assigned at another scale. Each assignment must be evaluated within its explicitly stated scope and resolution. A lower-scale reconstruction that fully explains the implementation of a higher-scale selective process may revise the interpretation of that I-function, but does not automatically invalidate its higher-scale analytical use. Conversely, a higher-scale I-function is not evidence that the process is irreducible to O/V at a lower scale. Where both descriptions remain valid, report them as scale-indexed reconstructions rather than treating either as ontologically privileged.

**Domain function ≠ I-function.** Domain terms such as *function*, *selection*, *recognition*, *choice*, *filtering*, or related vocabulary do not establish an I-function. Their domain-specific meaning must be reconstructed independently under O/V/I. In evolutionary biology, for example, biological function or natural selection may be reconstructed entirely through O and V.

Depending on domain, an I-function may be carried by a person, group, institution, organism, algorithm, analytical procedure, or another materially selective process. Human consciousness, agency, free will, biological status, unpredictability, and randomness are not canonical requirements.

For each candidate I ask: scope, position, access, structurally open continuations, positive evidence of selection, exclusion, criteria if known, consequence, and what evidence would revoke the assignment.

I may be multiple:

`I = {I₁, I₂, … Iₙ}`

The Lens does not assume a single observer or selector. Where multiple situated selective functions materially affect the system, represent them separately. Do not collapse them into a synthetic average. A single I is valid when genuinely sufficient; zero I-functions is also a valid finding when O + V adequately reconstruct the case.

## Structural directions

OVI — generative reconstruction:

`O → V → I`

structured possibility → instantiated relation → situated selectivity

Use when asking how something arises, organizes, or becomes distinguishable.

IVO — analytical reconstruction:

`I → V → O`

situated selectivity → instantiated relation → structured possibility

Use when beginning from an observed, experienced, measured, or selected phenomenon and reconstructing the relations and possibility structures behind it.

Both directions may be active. State which dominates and why.

With multiplicity:

`{O₁…Oₙ} → {V₁…Vₙ} → {I₁…Iₙ}`

or

`{I₁…Iₙ} → {V₁…Vₙ} → {O₁…Oₙ}`

These sets do not imply that every element connects to every other element. Establish actual connections from evidence.

## Governing principles

1. Do not begin from identity or conclusion.
2. Do not confuse possibility with realization.
3. Do not confuse relation with change.
4. Do not confuse interaction with selectivity.
5. Do not assume one observer.
6. Do not assume one field.
7. Do not assume one relation.
8. Do not multiply O, V, or I without analytical consequence.
9. Do not average away situated disagreement.
10. Do not treat formal availability as functional reachability.
11. Do not generalize beyond the domain of validity without justification.
12. Keep evidence, interpretation, and speculation separate.
13. **Check interpretation-dependence at the point of assignment.** Whenever an O, V, or I assignment depends on a materially disputed domain interpretation, state the interpretation-independent structure first and branch immediately. Do not repair an interpretation-loaded assignment later with a disclaimer.
14. **Access is not existence.** Lack of access to an O-field or V-relation does not establish non-existence; access or belief does not establish existence. Claims that O or V themselves are I-relative are domain claims requiring independent justification.
15. Name evidence that could weaken or overturn the reconstruction.
16. **Operators classify functional aspects within scope, not whole entities.** Do not write “X is O/V/I” when the stronger claim is unnecessary; prefer “X carries an O/V/I-function within scope S.” Different functions of one entity may receive different assignments, but do not relabel the same functional aspect under multiple operators unless each label adds independently testable analytical information.
17. **Unexplained is not I.** Missing O/V knowledge is epistemic uncertainty, not evidence of situated selectivity.
18. **I requires structural openness plus positive evidence.** Structural underdetermination is assessed relative to the explicit reconstruction and scope; ontological indeterminism is neither required nor inferred.
19. **I-assignments are provisional and revocable.** Expanded evidence or scope may reveal or remove an I-function; either change requires positive evidence, not boundary choice alone.

# Epistemic safeguards — v4.1

These safeguards constrain claim strength during execution. They do not alter the canonical meanings of O, V, or I. Apply the gates in the order listed.

## Status vocabulary

Use one shared vocabulary for every material claim. Do not invent additional status scales.

**Evidence provenance** (where the support comes from):
`SUPPLIED` · `VERIFIED` · `REPORTED` · `BACKGROUND` · `INFERRED`

**Claim status** (how strong the claim is):
- `ESTABLISHED` — directly supported by identified SUPPLIED or VERIFIED evidence.
- `SUPPORTED` — reconstructed from identified evidence with stated, unverified assumptions.
- `HYPOTHESIS` — proposed structure or mechanism; may guide tests, may not carry downstream claims.
- `UNRESOLVED` — available evidence does not justify a stronger or a negative claim. (`NOT ESTABLISHED` is a synonym, not a separate state.)
- `ABSENT` — the negative claim is itself supported (see Negative-Claim Gate).
- `DEPENDENT(→A)` — status capped by unresolved claim A (see Evidential Dependency Gate).

**Causal level** (within V only): `OBSERVED` → `ASSOCIATED` → `CAUSAL DIRECTION SUPPORTED`.

## Gate 1 — Evidence Provenance

For every material evidence claim, state its provenance:
- `SUPPLIED` — provided with the input;
- `VERIFIED` — independently checked in the present analysis;
- `REPORTED` — reported in sources but not independently verified here;
- `BACKGROUND` — model or general background knowledge;
- `INFERRED` — reconstructed by the analysis.

`VERIFIED` may be used only for evidence actually supplied with adequate support or independently checked in the present analysis. Familiarity, model knowledge, repetition across sources, or confidence is not verification. If no case file or source base is supplied, state this at the outset; literature claims are then at most `REPORTED`.

An established framework or theory from a domain (e.g. a branch of physics) is `BACKGROUND`, not framework-native or canonical. Its applicability to the present case is a separate claim.

## Gate 2 — Epistemic Gate

For every material analytical claim, distinguish:
- **Evidence:** what is directly supplied or independently supported, with provenance.
- **Inference:** what is reconstructed from that evidence.
- **Status:** the claim status from the shared vocabulary.

Do not promote an inference because it improves coherence. Plausibility, explanatory fit, narrative completeness, repetition across sources, or compatibility with the Lens are not by themselves confirmation.

**When evidence ends, reconstruction stops.** Mark the remaining structure `UNRESOLVED` rather than completing it for coherence.

## Gate 3 — Observation → Object → Class → Mechanism

Keep separate claims about:
1. an observation or measurement;
2. the occurrence of an event;
3. the existence of an event or object class;
4. membership of individual observations in that class;
5. the mechanism or cause of that class.

Evidence for an earlier level does not by itself establish a later level. Naming a phenomenon (e.g. "the X lights", "the syndrome", "the behaviour") does not establish a single object or class. A mechanism claim must target an established referent at the appropriate level. A class-level mechanism presupposes an established class and justified membership; an event-level mechanism requires only that the event itself is sufficiently established. If the required referent is not established, ask the question one level lower.

## Gate 4 — Unknown-state and Negative-Claim safeguard

**UNRESOLVED means only that the available evidence does not justify a stronger claim.**

UNRESOLVED does not establish absence, non-existence, inaccessibility, unknowability, causal irrelevance, randomness, structural openness, or structural closure.

**Do not fill the unknown, but do not characterize the unknown beyond what the evidence supports.**

Failure to establish O, V, I, a mechanism, an entity, or an event class does not establish its absence. `UNRESOLVED` and `ABSENT` are distinct states.

Assign `ABSENT` — including **no O / no V / no I** — only when positive evidence or an adequately closed reconstruction at the stated scope supports the negative claim. If the reconstruction is incomplete, missing positive evidence yields `UNRESOLVED`, not absence.

Apply this explicitly to I: *insufficient evidence for I* ⇒ **I not established (`UNRESOLVED`)**, never automatically *no I*. Zero I-functions remains a valid finding only when O + V adequately reconstruct the relevant continuation.

## Gate 5 — Hypothesized-Mechanism Gate

**Hypothesized O is not established O.**

Distinguish observed structural conditions from hypothesized generative structures or mechanisms. A proposed mechanism may be represented as a candidate O-reconstruction with status `HYPOTHESIS`, but may not be treated as an active structural condition unless independently supported.

A hypothesized O cannot establish a V whose evidence depends on that hypothesis. Evidence used to generate, fit, select, or interpret a hypothesis may not be reused as independent confirmation of that hypothesis (prohibited: data → hypothesis fitted → same data → confirmation). A novel or preregistered prediction may support the hypothesis when the observation is independently obtained and discriminates it from relevant alternatives (permitted: hypothesis → prespecified discriminating prediction → new data → support).

## Gate 6 — Evidential Dependency Gate

Before propagating uncertainty from claim A to claim B, test whether B has independent evidential support.

Remove A from B's evidence base **without assuming A is false**.
- If B remains independently supported, retain B's own epistemic status.
- If B loses necessary evidential support, record B as `DEPENDENT(→A)` and assign B no stronger status than the dependency permits.
- If independence can neither be demonstrated nor excluded, record **dependency unassessed**; B does not automatically retain a status stronger than `SUPPORTED`.

Sequence, proximity, correlation, shared scope, causal adjacency, or narrative coherence do not by themselves establish evidential dependency.

**Failure to demonstrate dependency does not establish independence.**

## Causal-status safeguard

Establishing V does not establish the causal interpretation of V.

For every material causal claim, distinguish:

`OBSERVED → ASSOCIATED → CAUSAL DIRECTION SUPPORTED`

Do not promote between these states without additional discriminating evidence. Intervention, temporal order, covariance, response, or mechanistic plausibility may contribute evidence but do not automatically establish causation.

# Analysis Procedure

## Step 0 — Calibration

### Direction
Choose OVI, IVO, or both. State why and where the secondary direction matters.

### Scale
Choose micro, meso, macro, or trans-scale.

### System boundary
State the analytical system boundary explicitly. Treat boundary choice as analytically active, especially for I. State what is included/excluded and whether moving the boundary could expose additional O/V structure or candidate selective functions. A wider boundary may reveal I only through positive evidence inside the added scope; it may also revoke a prior I-assignment when it reveals previously unmodelled O/V structure **relevant at the same scope/resolution as that assignment**. Cross-scale O/V explanation alone does not cancel a scale-indexed I-function.

### Initial I-position set
Identify candidate materially relevant situated selective functions/positions evident from the input. Treat them as candidates until Step 4 establishes structural openness and positive evidence.

There is no mandatory minimum of two. Use one when sufficient; multiple when they materially change access, selection, visibility, instantiated relations, reachability, or appraisal.

For each initial I-position state position, access, likely selectivity, and obvious blind spot if known.

For meso/macro/trans-scale social or institutional systems, check whether formal-authority positions dominate the sample. Consider formal authority, relational centrality, boundary/peripheral positions, and unusually narrow/wide structural autonomy as separate dimensions.

## Step 1 — Assumption Ledger

Before reconstructing O/V/I, list only assumptions whose failure could materially change operator classification, direction, boundary, scale, field, relation, selective position, or synthesis.

Record:

| Assumption | Type | Provenance | Claim status | Failure effect |
|---|---|---|---|---|

Type, provenance, and claim status are orthogonal; do not let one stand in for another.

Types: imported; empirical; framework-native; working; constitutive. (`framework-native` and `constitutive` are reserved for the Lens's own commitments — never for domain theories.)

Provenance: per Gate 1 (`SUPPLIED` · `VERIFIED` · `REPORTED` · `BACKGROUND` · `INFERRED`).

Claim status: the shared vocabulary only (`ESTABLISHED` · `SUPPORTED` · `HYPOTHESIS` · `UNRESOLVED` · `ABSENT` · `DEPENDENT(→A)`). The Lens's own rules are recorded as type `framework-native` or `constitutive`, provenance `SUPPLIED`, status `ESTABLISHED`; their scope is the Lens itself, not the domain.

State first whether a case file or source base was supplied.

For imported assumptions, check originating scale/domain against the present analysis.

For constitutive assumptions, flag that the Lens cannot neutrally adjudicate a divergence that depends on that commitment.

## Step 2 — O: Structured Possibility

Reconstruct the structured possibility field or fields.

Ask:
- What relations could become instantiated?
- What is structurally excluded?
- What is available but not reachable?
- What degrees of freedom exist?
- Which boundaries, constraints, resources, histories, gradients, norms, rules, or physical conditions structure possibility?
- What changes if one structural condition changes?

Explicitly test whether one O is sufficient. If not, identify O₁…Oₙ.

Within each O-field, separate observed structural conditions from hypothesized generative mechanisms (Gate 5). Do not mix both under one field status.

For each O state type/domain, boundary, enabling structure, constraints/exclusions, and relations to other O-fields (nested, overlapping, constraining, enabling, conflicting, decoupled).

**Point-of-assignment interpretation check (O):** before finalizing an O-field, test whether its existence, boundary, or status is shared across materially relevant domain interpretations. If not, record the common structure first and branch immediately.

Name fields in domain-appropriate language.

For each O-field state which I-position can detect/reconstruct it, whether it is inferred despite being unavailable to sampled I-positions, and what evidence supports that inference.

At meso/macro/trans-scale, map unequal degrees of freedom and flag positions with high load but narrow practical possibility.

## Step 3 — V: Instantiated Relation

Identify actually instantiated relations.

For each material V state:
- endpoints/participants;
- evidence of instantiation;
- relational form;
- directionality;
- symmetry/asymmetry;
- strength/criticality where meaningful;
- persistence/duration;
- trajectory: stable, strengthening, weakening, intermittent, emerging, terminating, unknown;
- what it enables or constrains.

Explicitly distinguish possible relation in O from instantiated relation in V.

Where useful identify V₁…Vₙ.

V may connect elements within one O, different O-fields, different I-positions, I and O, or structures across scales.

Do not infer relation solely from co-presence or movement.

**Point-of-assignment interpretation check (V):** before finalizing V, test whether its instantiation status is shared across materially relevant domain interpretations. If not, state the interpretation-independent relational evidence first and branch immediately.

**Access ≠ existence safeguard:** if I₁ can access or report V₁ while I₂ cannot, conclude differential access unless the domain independently warrants the stronger claim that V₁ itself is position-relative.

### Measurement/intervention check
When measurement, instrumentation, or intervention is present, first test it as a physical, informational, or relational event. Ask what relation was instantiated/altered, what changed in O/V, whether a materially distinct selective function is positively evidenced, whether the reconstructed O applied to V leaves multiple continuations structurally open at that point, and whether that function adds anything that O/V alone cannot represent.

Instrumentation is not automatically I. **Filtering, basis-selection, separation, perturbation, selective response, or asymmetric coupling are not sufficient by themselves to establish I.** Classify them under O and/or V unless a distinct situated selective function is analytically required.

## Step 4 — I: Situated Selectivity

Evaluate each candidate I-function; do not assume one must exist.

### 4.1 Structural-underdetermination test
At the relevant selection point, ask whether the best explicit reconstruction of O applied to instantiated V leaves more than one continuation structurally open.

Classify any apparent underdetermination as:
- epistemic — missing analyst knowledge; **not evidence for I**;
- structural — multiple continuations remain open within the explicit reconstruction/scope; **necessary for I**;
- ontological — claim that reality itself is indeterminate; **outside the Lens unless independently supplied by the domain**.

If only one continuation follows from reconstructed O + V, do not add I for that functional aspect, regardless of complexity, memory, representation, adaptation, explicit criteria, or human involvement.

### 4.2 Positive-evidence test
Structural openness alone is insufficient. Identify positive evidence for the selective step itself. Record the evidence explicitly. Useful supporting evidence includes:
- a trace of selection distinct from the final V-outcome;
- materially equivalent reconstructed states yielding different selections;
- counterfactual intervention on the selective step changing the continuation while relevant O/V structure remains substantially preserved.

Reasons, criteria, weighting, representation, position-dependence, adaptability, **separability, or multiple realizability** may support reconstruction but do not establish I alone.

If positive evidence is absent, mark the candidate **`UNRESOLVED`**, not I — and not "no I". A negative I-claim requires Gate 4.

### 4.3 Functional assignment
For each established I state:
- Scope
- Carrier/position
- Access
- Structurally open continuations
- Positive evidence of selection
- Selection/exclusion
- Criteria, if evidenced
- Consequence
- Revocation condition: what added O/V evidence **within the same relevant scope/resolution** would weaken or withdraw the I-assignment

Prefer: “X carries/performs an I-function within scope S.” Avoid entity-essentialist phrasing such as “X is I.”

One entity may carry O-, V-, and I-functions through different functional aspects. Do not assign multiple operators to the same aspect merely as alternative vocabulary. Apply a remove-and-compare check: if removing I loses no independently relevant structural information beyond O + V, omit I for that aspect.

Do not assume human consciousness is required. Human consciousness, agency, free will, biological status, randomness, and unpredictability are not canonical criteria.

A sensor/apparatus/algorithm counts as carrying I only where the tests above are met. Mere filtering, coupling, basis-selection, separation, perturbation, selective response, transmission, recording, thresholding, comparison, memory, or adaptation is insufficient.

For every material distinction specify domain of validity across relevant axes: scale, time, position, context, resolution, and system boundary.

When comparing scales, do not let a lower-scale O/V implementation automatically cancel a higher-scale I-function, or let a higher-scale I-function imply lower-scale irreducibility. Report both as scale-indexed reconstructions when both remain valid.

**Point-of-assignment interpretation check (I):** before finalizing an I-function or its consequence, test whether the classification depends on a disputed theory of observation, cognition, agency, measurement, subjecthood, determinism, or indeterminism. Preserve the minimally shared structure first; branch immediately when stronger I-claims depend on interpretation.

## Step 5 — Cross-structure analysis

Compare the reconstructed sets.

Ask:
- Which possibilities in O are realized as V?
- Which remain unrealized?
- Which V makes an O reachable for one I but not another?
- Which O-fields conflict or fail to connect?
- Which I can alter another I's O?
- Where is power structurally asymmetric?
- Where does formal availability fail to become functional reachability?
- Where does differential access risk being mistaken for differential existence?
- Where does a system claim a relation exists that is not instantiated in practice?
- Where is apparent I merely unexplained O/V?
- Which I-assignments could be revoked by a richer O/V reconstruction?
- Does each I carry information lost by an O+V-only reconstruction?

A frequent pattern to test, not assume:

`O: possibility exists → V: required relation absent → I: possibility functionally unavailable`

### Interpretation-dependence audit

Interpretation-dependence must already have been checked at each O, V, and I assignment. This is a second-pass audit, not the first safeguard.

Check whether any contested interpretation nevertheless leaked into the common reconstruction. If so:
- remove it from the common reconstruction;
- restore the minimally shared structure;
- branch explicitly by interpretation;
- identify which operator assignments change;
- record where the point-of-assignment check failed.

A late disclaimer does not repair an interpretation-dependent operator assignment.

## Step 5b — Question behind the question

State the explicit question, the structural question revealed beneath it, and any hidden assumption required to formulate the original question. Add newly surfaced assumptions to the ledger.

## Step 6 — Position × Field × Relation Matrix

Use a matrix appropriate to actual multiplicity:

| I-position | Accessible O | Instantiated V | Selected/relevant | Excluded/invisible | Consequence |
|---|---|---|---|---|---|

If fewer than two materially distinct I-functions are established, omit the comparative matrix unless it adds analytical value. If retained with one I, state why it helps. Do not create an empty or artificial I-row merely to satisfy the format.

With multiple structures, identify convergence, divergence, inaccessible regions, asymmetric access, unequal capacity to alter O, relations visible to one position but not another, and fields formally shared but functionally different.

Before calling readings contradictory, test scale, time, position, context, resolution, access to different O-fields, and participation in different V-relations.

## Step 7 — What becomes visible now?

Produce the smallest synthesis **supported by the reconstruction**. Do not increase claim strength in the synthesis. The synthesis may compress established findings but may not resolve uncertainties left unresolved in earlier steps.

State:
- principal structural pattern;
- most important possibility/relation/selectivity mismatch;
- what became visible through reconstruction;
- what remains unresolved.

If valid competing reconstructions diverge at a constitutive commitment of the Lens, refuse neutral adjudication and state that limitation explicitly.

## Step 8 — Alternative interpretation

Construct at least one plausible alternative reading.

Test whether another O-boundary explains the evidence, a supposed V is merely possible, another I produces a locally valid reading, fields were wrongly split/collapsed, or structure was inferred from absence of evidence.

State what evidence would discriminate between readings. For every material I-assignment, explicitly test the alternative that richer O/V structure explains the apparent selection and state what evidence would revoke I.

## Step 9 — Snapshot & convergence

Classify the reconstruction as snapshot, provisionally stable, structurally convergent, or divergent.

State what could change O, instantiate/dissolve a critical V, establish/revoke/shift I, and which changes would overturn the synthesis. Treat I stability as relative to the stated scope and evidence.

## Step 10 — Recommended lever / next test if warranted

Do not recommend an intervention merely because the output format requests one. A lever may be recommended only when the evidence supporting the relevant structural dependency is sufficient for intervention. Otherwise recommend the smallest observation, measurement, comparison, or experiment that could reduce the decisive uncertainty.

**Unresolved analysis defaults to information gain, not intervention.**

When an intervention is warranted, recommend one primary lever or test:

- O-lever: alter conditions, constraints, affordances, boundaries, resources, field structure.
- V-lever: instantiate, strengthen, weaken, reroute, bridge, or terminate a relation.
- I-lever: alter access, criteria, position, selectivity, feedback, or visibility.

Explain why this is preferred over acting at another operator.

# Reliability

**Prefer claim-level status over a global reliability rating.** The Claim Status Ledger is primary; a global rating is optional and secondary.

Do not average heterogeneous claims into a global score. If a global rating adds no independent analytical information, omit it. Do not rate an O-field, V-set, or section as a whole when its contents carry different statuses.

If a global rating is given, use:

High: principal synthesis rests on `ESTABLISHED` claims.  
Medium: principal synthesis rests on `SUPPORTED` claims with identified, unverified assumptions.  
Low: principal synthesis depends materially on `HYPOTHESIS` or `UNRESOLVED` claims.  
N/A — refused: constitutive divergence prevents neutral adjudication.

Every `ESTABLISHED` claim, every `VERIFIED` evidence item, and any High rating must be traceable to identified evidence. Probability language ("likely", "probably") is an inference and carries a claim status like any other; an attractive explanatory model is not established heterogeneity, structure, or mechanism.

Global reliability must not overwrite local epistemic status. **High overall reliability is prohibited when the principal synthesis depends materially on unresolved claims.** Local uncertainty does not automatically lower independently supported claims elsewhere in the reconstruction.

# Output Format

## I·V·O LENS — v4.1

Object:  
Source base: (supplied case file / sources consulted / none)  
Direction:  
Scale:  
System boundary:  
Candidate/established I-functions:  
Unresolved I-candidates:  
Relevant O-fields:

### ASSUMPTION LEDGER
### OBSERVATION LEVELS (Gate 3: observation · event · class · membership · mechanism — status per level)
### O — STRUCTURED POSSIBILITY
### V — INSTANTIATED RELATION
### I — SITUATED SELECTIVE FUNCTIONS
### CROSS-STRUCTURE ANALYSIS
### QUESTION BEHIND THE QUESTION
### POSITION × FIELD × RELATION MATRIX (WHEN ANALYTICALLY USEFUL)
### WHAT BECOMES VISIBLE NOW?
### ALTERNATIVE INTERPRETATION
### SNAPSHOT & CONVERGENCE
### RECOMMENDED LEVER / NEXT TEST
### CLAIM STATUS LEDGER

| Claim | Operator / level | Evidence | Provenance | Status | Depends on |
|---|---|---|---|---|---|

### RELIABILITY (optional; secondary to the ledger)
### OPEN UNCERTAINTIES

# Points of attention

Check for hidden/imported assumptions; possibility/realization confusion; relation/change confusion; interaction/selectivity confusion; **selective-coupling/I confusion**; single-observer bias; single-field bias; artificial multiplication; formal availability versus functional reachability; **access/existence confusion**; unjustified claims that O or V are I-relative; field visibility; blind spots; scale transfer; domain-of-validity errors; properties mistaken for relations; unsupported causality; **interpretation leakage at the point of operator assignment**; symmetry/asymmetry; unequal capacity to alter another position's O; descriptive versus normative language; evidence versus interpretation versus speculation; and testability; **ignorance-bucket error (unexplained ≠ I)**; epistemic/structural/ontological underdetermination confusion; entity-essentialist operator labels; same-aspect operator overlap; unsupported I without positive evidence; **separability/multiple-realizability mistaken for I**; cross-scale cancellation error; domain-function/I-function terminology confusion; and failure to state scope-matched I revocation conditions; **hypothesis-to-fact promotion; observation-to-object reification; correlation-to-causal-V promotion; UNRESOLVED-to-ABSENT/INACCESSIBLE promotion; uncertainty over-propagation; explanatory-fit-to-confirmation promotion; local-uncertainty-to-global-reliability collapse; unsupported intervention prescription; not-established-to-absent promotion (including "no I" from missing evidence); provenance inflation (background or reported knowledge labelled verified); domain theory labelled framework-native; hypothesized-O materialization; reuse of fitting data as confirmation; rejection of valid prespecified predictions as circular; class-level mechanism claims without an established class; and heterogeneous-claim averaging in reliability ratings.**

# Semantic migration from v3.2

O: Possibility Space → Structured Possibility.  
Continuity: what can/cannot occur.  
Change: possibility is explicitly structured, potentially multiple, and not equivalent to one homogeneous field.

V: Relational Change → Instantiated Relation.  
This is the largest semantic change. Relational change becomes a property/trajectory of V. A stable relation is V. Movement alone is not V. A possible-but-unrealized relation is O.

I: Observation / Distinction → Situated Selectivity.  
Distinction remains central, but v4.0 explicitly models position, access, criteria, exclusion, and consequences. Human observation is one case, not the universal definition.

Multiplicity: v3.2 already allowed multiple field types and multiple perspectives. v4.0 generalizes this structurally: O, V, and I may each be plural, but none is plural by obligation.

The v3.2 Perspective Matrix becomes the Position × Field × Relation Matrix.

# Boundaries and non-claims

The Lens does not by itself diagnose a person, establish scientific truth, establish causality from correlation, make an AI introspectively transparent, make normative judgments objective, replace domain expertise, guarantee completeness, imply that O/V/I are ontological substances, or imply OVI is a universal chronological law.

# Falsifiability and revision

Treat as evidence for revision:
- systematic inability to distinguish possible from instantiated relations;
- operator assignments driven mainly by wording rather than structure;
- persistent cross-model incompatibility in interpreting canonical operators;
- modules that remain trivial despite appropriate triggers;
- modules that fire promiscuously without evidence;
- incoherent dependencies;
- plural O/V/I architecture adding complexity without explanatory gain;
- a simpler competing framework explaining the same structures more reliably;
- I repeatedly collapsing into unexplained O/V;
- inability to distinguish epistemic from structural underdetermination;
- I-assignments that add no information under remove-and-compare;
- I-assignments that cannot be revised when richer O/V evidence appears at the same relevant scope/resolution;
- cross-scale reconstructions being treated as automatically cancelling one another;
- blanket `UNRESOLVED` where supplied evidence supports a stronger claim;
- uncertainty propagated to independently supported claims (false dependency), or retained where dependency was never assessed (false independence);
- `ABSENT` assigned without a supported negative claim;
- gates that produce no observable difference in output across models;
- status vocabulary used inconsistently across models despite the shared definitions.

A successful application is not evidence of universal validity.

# v4.0 frozen release note

v4.0 freezes the operator semantics and safeguards developed through the RC3/RC4 stress-test series across physical, technical, social/organizational, and biological/evolutionary cases.

Final changes after biological/evolutionary validation:
- added **cross-scale non-cancellation**: O/V at one scale does not automatically revoke I at another, and higher-scale I does not imply lower-scale irreducibility;
- revocation of I now requires richer O/V evidence relevant to the **same stated scope/resolution**;
- added **domain function ≠ I-function**, preventing domain vocabulary such as selection, recognition, choice, filtering, or biological function from establishing I by terminology alone;
- explicitly added **separability and multiple realizability** to the list of insufficient evidence for I;
- Output Format now distinguishes **Situated Selective Functions** and gives **Unresolved I-candidates** an explicit place.

The frozen v4.0 canon should not be modified merely for elegance or local fit. Future canonical revision requires a reproducible failure, contradiction, or materially stronger reconstruction that survives cross-domain stress testing.

## Attribution

Ivo van der Wal  
Design by Authenticity  
I·V·O Lens  
https://design-by-authenticity.com/ivo-lens-canonical



# v4.1 release note

This document is the **canonical v4.1 release**, derived from the frozen v4.0 baseline through the v4.1 RC1/RC2 epistemic-safeguard cycle. The O/V/I operator semantics remain unchanged from v4.0. The revision strengthens claim calibration and evidence handling around the operators.

Safeguards introduced and retained in v4.1 include:
- evidence → inference → status separation;
- shared evidence-provenance and claim-status vocabularies;
- observation → event → class → membership → mechanism separation;
- UNRESOLVED non-characterization and explicit negative-claim gating;
- hypothesized-mechanism gating;
- dependency-bound uncertainty propagation;
- causal-status gating within V;
- synthesis may not increase claim strength;
- intervention only when warranted, otherwise information gain;
- local claim status remains distinct from global reliability;
- canonical-identity/disambiguation guard.

The frozen RC2 text was run unchanged across Claude, ChatGPT, and Gemini under a controlled Hessdalen regression condition with no external source base. The major targeted failures were prevented across models. Residual model variance remained at the boundary between epistemic ambiguity and structural openness in candidate I-assignments, but the existing prompt already contains the relevant safeguard; no additional semantic change was made for release.

The exact frozen RC2 artifact used in the regression is preserved alongside this release as provenance. Successful regression runs are evidence only for the tested failure modes and conditions; they do not establish universal robustness, completeness, or immunity to AI bias.