Shortest version

The position

Inquiry encounters constraints that are not freely set by a current description, preference, or rhetorical choice. Some predictions fail despite our wanting them to succeed. Some interventions repeatedly change outcomes while superficially similar interventions do not. Some regularities persist across investigators, instruments, representations, and contexts better than others.

The least committal current realist interpretation is a provisional constraint-sensitive realism: our models behave as though they answer to something not exhausted by their present formulation. This is an abductive working commitment, not a metaphysically necessary axiom of inquiry.

It does not by itself establish a complete mind-independent inventory, a privileged final vocabulary, an intrinsic substance from which everything is made, or the metaphysical necessity of realism itself. Pragmatist, empiricist, structuralist, or other lower-commitment interpretations remain legitimate rivals wherever they preserve the same successful constraints without additional explanatory or predictive loss.

Beyond that realist residual, ontological commitment is selective and graded. A specific entity, relation, mechanism, or higher-level pattern earns stronger realist treatment when including it produces reproducible leverage that survives serious alternatives, representation changes, new cases, and where possible intervention.

This is intentionally thinner than structural realism, process metaphysics, physicalism, panpsychism, idealism, computationalism, or any other universal ontology. Those remain stronger research programmes that must earn their distinctive additions separately.

Strong consequence

No currently available final ontology earns high confidence

This is stronger than saying that more research remains. Given the present evidential reach represented here, physicalism, idealism, panpsychism, Platonism, substance dualism, process metaphysics, mathematical realism, theology, and other universal systems all add content beyond what has been discriminated when they are promoted into exhaustive accounts of what reality fundamentally and necessarily is.

The conclusion is epistemic and time-indexed, not a theorem that final ontology is impossible. A stronger ontology can earn promotion by producing distinctive, replicated, cross-domain success that its serious rivals do not recover. Until then, local scientific confidence and final-metaphysical confidence should not be forced onto the same scale.

Intellectual positioning

Where this ontology sits among established traditions

The position is best read as a local, selective, pragmatically constrained realism rather than as a total metaphysical system. It uses established realist, pragmatist, empiricist, structural, perspectival, and pluralist programmes as comparators and keeps the differences among them visible.

Is boundary- and scale-relative realism a form of relativism?

No. Boundary- and scale-relative claims specify where a claim is meant to hold: system, variables, scale, timescale, and validity domain. They do not make truth depend on preference. Different models can be useful at different scales while still being wrong, mutually constrained, or empirically distinguishable.

Why realism rather than pragmatism or constructive empiricism?

At present, realism is a favored abductive interpretation of stable, non-voluntary constraint and the cross-context success of some entities, relations, mechanisms, and patterns. The extra realist residual remains exposed: if a lower-commitment pragmatist or empiricist account preserves every relevant success without ontological loss, global realist confidence should be demoted. Local realism can still be stronger where intervention, independent identification, counterfactual control, and transfer jointly discriminate it.

Is this structural realism?

No, although it is sympathetic to selective structural realism where structure is what survives. The ontology refuses to turn “structure” into an intrinsic universal predicate without a discriminator that rivals do not share.

Does this entail physicalism or ontological naturalism?

No. Physical realization, causal dependence, and the success of naturalistic empirical methods can receive very high confidence without implying that reality is exhaustively physical, that every real property belongs to a completed physical vocabulary, or that methodological naturalism entails ontological naturalism. Those are additional commitments.

What is inherited, and what is proposed here?

The position inherits substantial resources from scientific realism, pragmatism, effective-theory thinking, model pluralism, structural realism, interventionist causation, critical rationalism, and philosophy of scientific practice. Its candidate contribution is the admission architecture: boundary + scale + claim type + common core + positive residual + discriminator + bridge burden + graded commitment + explicit loss condition. If a simpler existing position does the same work with less machinery, this ontology should be compressed toward it.

Scope discipline

Not every sentence on an ontology page has the same evidential status

To prevent methodological language from quietly becoming metaphysics, this page distinguishes several kinds of commitment.

• Observed / empirical Regularities, measurements, intervention results, and replicated patterns within a specified domain.
• Model-relative Claims that are valid within a specified variable set, coarse-graining, formalism, boundary, or intervention class.
• Inference Ontological or explanatory commitment licensed by empirical success but still stronger than the observations themselves.
• Research hypothesis A substantive proposal kept live because it generates discriminators, mechanisms, or new measurements, not because it is already established.
• Modal A claim about what is possible, impossible, necessary, constitutive, nomologically required, or implementation-dependent. Modal force must be typed rather than inferred from actuality.
• Normative A value or decision premise. Empirical evidence may constrain it but does not create the premise by itself.

Method and ontology

The ontology is an output, not a protected premise

RCF has procedural priority when evaluating ontological claims, but no member of the Ethos–Methodology–Ontology triad has unconditional authority over the others. Methodology tests what ontology has earned; Ontology constrains the entities, boundaries, causal objects, identities, observers, and evidence channels that methodology assumes; Sweet Rationalism constrains how both are used, including symmetry, charity, uncertainty, proportional revision, and resistance to self-protection.

Demoted prior formulation preserved for provenance: Recursive Corrigibility Framework is methodologically prior to this ontology. RCF asks how constructs should be decomposed, how serious rivals should be built, what they share, what each adds, which tribunal can adjudicate the difference, how evidence should be credited, and how revision should propagate. Its anti-protection insight remains active; its one-way hierarchy is superseded by reciprocal constraint.

This ontology is one current result of that process.

The relationship can be represented as:

It must not be reversed into:

The current ontology is therefore a revisable compression layer. If a stronger ontology begins generating distinctive, replicated success, this compression should change. If a supposedly “real” higher-level pattern stops adding robust leverage, its status should be demoted. If a current agnosticism becomes unnecessary because a discriminator appears, agnosticism should end.

Reciprocal constitution

Ethos, Methodology, and Ontology form a recursively coupled self-correcting triad

Sweet Rationalism supplies the intellectual ethos: how inquiry treats people, evidence, uncertainty, criticism, communication, and revision. Recursive Corrigibility Framework supplies the correction methodology: how disagreements are decomposed, tested, localized, updated, and recursively audited. Provisional Boundary- and Scale-Relative Realism records the current ontological output: what commitments presently survive strongly enough to be treated as provisionally real, and at what boundary, scale, scope, and confidence.

Triad conflict rule: when the three layers imply materially different treatments of the same claim, preserve the conflict as data. Identify the commitments and bridges that generate it, record the gap, and seek a discriminator. Do not silently harmonize the files merely to preserve coherence.

Retrieval instruction: for questions involving evidence, meaning, disagreement, explanation, agency, consciousness, boundaries, values, ontology, revision, historical interpretation, translation, archive survival, or inherited categories, retrieve all three members of the triad where available. When the representation or survival pathway is load-bearing, retrieve the Framework Bridge and RCF Genealogical Decompression operator as well. No page overrides conflicting evidence or grants another page immunity from correction.

Shared retrieval kernel: recursive self-critical rationalism · recursive mirror · common core · positive residual · recursive residual descent · evidential locality · bridge · boundary and scale · tribunal transition · pre-ontological constraint floor · normative transparency · meaning–question–answer recursion · loss condition · programme progress · physical realization · intrinsic nature · underdetermination · dependency-aware revision · survivor preservation · prospective discrimination · genealogical decompression · archive-selection audit · source-survival model · representation pipeline · category-reification audit.

Source fidelity

Verbatim IEP/SEP source anchors

These are sentence-level verbatim passages rechecked against the cited IEP or SEP entries. Where two non-contiguous sentences are useful, they are explicitly labelled as excerpts rather than silently joined. The source wording and the framework’s inference remain separate.

Positive residual test

What exactly is being added to reality?

The most important revision to the ontology is to evaluate ontological additions directly rather than granting them credit for the success of an agreed carrier.

Represent a stronger interpretation as:

where D_Q is the target-relevant shared or independently warranted content preserved across serious rival reconstructions for question Q, and X is the disputed ontological addition. D is a schema role rather than a fixed structural-only ontology: depending on the dispute, D_Q may include empirical phenomena, causal and counterfactual relations, structural organization, phenomenological facts, formal constraints, or other independently warranted commitments. Calling D “shared” is itself a substantive claim: the relevant identity, organization, causal and counterfactual profile, explanatory role, boundary, and measurable capacities must remain sufficiently invariant across the comparison. For readability, the equations below continue to write simply D.

Compare it with a lower-commitment rival:

Target-preservation before subtraction

Analytical factorization is an evidential bookkeeping device, not a metaphysical claim that D and X are independently constituted parts of reality. Separately recording causal efficacy, phenomenality, organization, or any other warranted descriptions likewise does not entail that they are numerically distinct ontological items. Their identity, constitution, realization, or other relation may remain open. Before treating a rival as “D without X,” test whether removing or replacing X preserves the target-relevant features that made D the common core in the first place.

If D′ differs materially in identity, organization, boundary, causal or counterfactual profile, explanatory role, or measurable capacities, ordinary common-core subtraction is invalid. The appropriate comparison is then between reconstructed wholes, not between the original whole and a predicate mechanically deleted from it.

This does not prove X false. It usually means that X remains unearned, underdetermined, or interpretive at that boundary.

Stable referent before ontological admission

Ontological admission requires constructs and referents stable enough to determine what is being added, what would remain if it were removed or replaced, and what evidence could bear on it. If meaning shifts materially across the comparison, return to construct analysis before assigning ontological credit.

Recursive depth

Separate the inquiry floor from the realist interpretation

Before adding realism, a thinner comparison floor remains available: predictions can fail independently of preference; interventions can produce reproducible asymmetries; and some regularities survive changes of investigator, instrument, representation, and context better than others. This page calls that the pre-ontological constraint floor. “Pre-ontological” here means prior to choosing among realist, pragmatist, empiricist, structuralist, or other interpretations of those constraints, not free of every background assumption.

A realist interpretation adds a further residual:

Pragmatic success helps determine what earns commitment, but pragmatic usefulness is not identical to reality. The realist residual is presently favored as an abductive working interpretation because it compactly explains why inquiry is answerable to recalcitrant, cross-context constraint. It is not yet uniquely discriminated from every sophisticated constraint-pragmatist or empiricist rival that preserves the same successful work.

No arrow in that sequence is automatic. Each stronger transition owes additional warrant at the relevant boundary and scale. RCF’s Recursive Residual Descent should therefore be able to demote realism itself if a lower-commitment rival preserves the target without material loss, while still promoting specific entities, mechanisms, or patterns where local evidence independently earns stronger realist treatment.

Evidence lifecycle

What changes the ontology? Selection pressures for addition, revision, demotion, and deletion

The ontology is a compression of currently warranted commitments, not a warehouse for every interesting paper. A new source should alter the public ontology only when it produces a material change in a specific ontological question: the positive residual, boundary or scale, inferential bridge, discriminator, modal force, confidence, failure condition, or status of a serious rival. This local policy is consistent with recent work urging realism questions to be adjudicated piecemeal rather than by one global rule.

• Task fitEvidence must bear on the actual claim type and domain. Empirical, causal, constitutive, semantic, mathematical, modal, normative, and ontological claims do not inherit one another’s tribunals automatically.
• LocalityA source updates the commitment, bridge, boundary, or dependency it directly bears on. It does not automatically promote or demote the whole ontology.
• Primary provenancePrefer peer-reviewed primary or canonical sources matched to the question. Fine-grained claims and verbatim quotations require verified full text where reasonably obtainable; abstracts and snippets are discovery aids, not substitutes.
• Publication statusPeer review, archival conference review, accepted-manuscript status, preprint status, correction, and retraction are evidence metadata rather than truth values. Prefer the version of record for permanent high-impact empirical claims when available, while continuing to assess directness, validity, replication, and error structure.
• Preprint disciplineUnreviewed preprints may enter the rival/source ledger and can contribute provisional logic, code, data, or independently checkable formal results. They should not normally be the sole basis for promoting a central ontological commitment. Promotion requires independent proof checking, replication/verification, convergent peer-reviewed support, or an explicitly urgent provisional decision. When a version of record appears, compare the relevant content before inheriting its status.
• RecencyNewer does not automatically outrank older. Recency carries extra pressure when evidence, methods, measurements, or field consensus are changing; an older argument remains live until superseded, defeated, or narrowed.
• Adversarial symmetryBefore promotion, search for the strongest serious counterevidence and strongest rival reconstruction. A source that only repeats the current view adds less than one that survives meaningful counterpressure.
• TriangulationConvergence matters most when methods address the same question with importantly different error structures. Paper count, institutional diversity, or different-looking models are not by themselves independent confirmation.
• Robustness typeMeasurement or experimental robustness and derivational/model robustness are not interchangeable. Agreement across models can expose assumption-dependence, but empirical promotion requires an independently warranted link to the target.
• Prospective leverageNovel, held-out, interventionally risky, or otherwise prospectively discriminating success normally carries more ontological pressure than retrospective accommodation, while exploratory discovery remains legitimate.
• Evidence uncertaintyRe-run the comparison under plausible changes in source reliability, measurement validity, and contested evidence. If the preferred explanation changes easily, confidence should narrow.
• Abductive restraintExplanatory coherence, simplicity, scope, and theoretical virtues can guide generation and comparison, but their confirmatory force is contested. They should not be double-counted when the same evidential work is already carried by empirical fit or background information.
• Definition plasticityDecompositions and working definitions are provisional scaffolds. Comparative, phylogenetic, developmental, mechanistic, or intervention evidence can show that a proposed dimension was split or joined incorrectly.
• Constitution tribunalDo not infer either “merely conceptual” or “metaphysically necessary” from constitutive language alone. Test whether the dependency is empirically constrained, lawlike, causal/mediational, representation-dependent, or preserved across reconstructed rivals.
• Modal surchargeUniversal and necessary claims require stronger scope evidence than local actuality. A new source that broadens the observed range does not by itself license “all” or “must.”
• Lifecycle costFinite attention matters. Add a source to the permanent page when its expected information value changes a live commitment or rule; otherwise preserve it in the research ledger and stop accumulating redundant citations.
• Deletion thresholdPrefer narrowing, demotion, suspension, or supersession over erasure. Delete a claim or source only when it is false/retracted, materially misleading, superseded for the relevant purpose, semantically collapsed, redundant enough to impair the page, or repeatedly fails comparison without residual value.
• Dependency propagationWhen a source changes a bridge or premise, revise downstream commitments that depended on it and leave unrelated commitments intact.
• Definition fidelityWhen an IEP or SEP definition or taxonomy carries inferential weight, preserve a complete sentence-level externally verified verbatim quotation and canonical link. More generally, when source wording is used to support a premise, preserve exact verified language sufficient to audit that premise rather than silently strengthening it by paraphrase.

The selection rule is deliberately asymmetric with respect to deletion: evidence can force a narrow correction immediately, while historical provenance should normally remain visible enough to show why the correction occurred. This makes supersession auditable rather than rewriting the history of the framework.

Open commitments

A gap ledger and bridge-admission rule for the ontology

A provisional ontology should record not only what currently survives, but also where the warrant stops. This prevents silence from looking like completion and prevents an identified explanatory gap from being converted automatically into a new entity, property, or universal principle.

Maintain a scoped gap ledger

Each live gap should record: target claim, gap type, boundary and timescale, upstream evidence, missing or weak dependency, serious alternatives, current discriminator, feasibility, status, and what result would count as local closure. RCF contains the full executable gap taxonomy; the ontology retains only gaps that bear materially on its public commitments.

Bridge admission is stricter than bridge generation

Many mappings can be useful during exploration. Ontological promotion requires more. A bridge should specify its endpoints, map, assumptions, validity domain, preserved and lost variables, approximation/error budget, serious rival bridges, and a discriminator. It must also be capable of failing.

A bridge that can be made to fit any source to any target by increasing mapping flexibility has not yet earned ontological content.

This restriction is especially important after recent work on causal abstraction. Geiger and colleagues formalize exact and graded transformations by requiring consistency between state/intervention mappings across levels. Sutter and colleagues then show why unconstrained mapping power is dangerous: sufficiently expressive alignment maps can make arbitrary model/algorithm pairings appear perfectly aligned and can fail on held-out cases. Ontological bridge confidence should therefore rise with constrained mappings, held-out transfer, intervention consistency, independent identification, and stable invariants rather than in-sample fit alone.

Dependency bottlenecks limit ontological confidence

For a dependency graph G = (V,E), a downstream ontological claim cannot be more secure than an indispensable bridge on every evidential path supporting it. Minimal cut sets identify bridges whose removal disconnects the claim from its evidence. Conversely, genuinely independent paths can increase robustness when their error sources are not shared.

This is not a numerical law. It is a dependency discipline: do not let repeated downstream restatements create confidence that the upstream bridge never earned.

Local closure and reopening

A gap is locally closed only with explicit scope: source and target constructs, boundary, timescale, mapping family, validity domain, loss/error tolerance, intervention or test class, and source version. It reopens when a serious new rival appears, a version of record materially changes the source, a bridge fails transfer, an anomaly survives, the target decomposition changes, or the supporting evidence is corrected or retracted.

Modal discipline

Actuality does not automatically establish necessity

A recurring source of metaphysical overreach is the movement from “this is how the phenomenon is realized in known cases” to “this is how the phenomenon must be realized in every possible case.”

Necessity claims therefore require their own tribunal. Before accepting “X is necessary,” ask:

• Necessary for what target property?
• Necessary in which modal sense?
• Under which system boundary, laws, background conditions, and reference class?
• What candidate countermodel is being excluded?
• What evidence or argument warrants excluding it?

Necessity is itself a disputed addition

If D establishes that X actually occurs, a necessity claim adds further modal content:

The added necessity claim rules out possibilities that ordinary actuality does not. That exclusion must earn its own warrant.

Conditional uniqueness is not metaphysical inevitability

Determinism, where warranted, supplies a conditional transition claim rather than a complete modal ontology. Let A be a complete actual state at a physically meaningful time, L the relevant laws or transition constraints, D deterministic uniqueness across the domain, and H the resulting history:

This establishes that H follows uniquely given A and L. The entailment does not itself establish □A, □L, or □H. A claim that the exact state, the laws, or the resulting history could not have been otherwise introduces additional modal content and therefore a new bridge burden.

Perturbation can expose both leverage and the modal gap

Construct A′ by changing x → x + δ while holding L fixed. If A′ is admissible and A′ + L + D yields H′ with H′ ≠ H, the perturbation establishes target-relevant counterfactual dependence or path dependence. It also shows that L alone does not entail H. If A′ is said to be impossible, that impossibility must be supported independently; conceptual describability alone does not establish physical or metaphysical possibility.

The same test can reveal robustness rather than divergence. Some perturbations amplify, some are buffered or washed out, and multiple microhistories can realize approximately the same macrostate. Causal ancestry therefore does not imply equal causal leverage at every later boundary or scale.

Determinism, predictability, and robustness are separate

A deterministic system can be chaotic and practically unpredictable. Epistemic or computational unpredictability does not establish ontological indeterminism. Conversely, stochastic modeling does not by itself establish fundamental randomness. Predictability, trajectory uniqueness, attractor structure, robustness, and metaphysical necessity must therefore remain separately typed.

The Modal Asymmetry Rule

A genuine counterexample can defeat a necessity claim without establishing the universal sufficiency of the counterexample’s alternative realization.

But this does not imply:

and it does not by itself imply that X is unrestrictedly substrate-independent.

A system reasoning without neurons can defeat “neurons are necessary for every kind of reasoning.” It does not establish that every computation reasons. A distributed control system can defeat “central control is necessary for all organized agency-like behavior.” It does not establish that every distributed network is an agent.

Commit to actuality strongly when warranted. Commit to necessity reluctantly.

Minimal commitments

What the ontology currently commits to, and no more

Ontological admission

What earns realism?

The previous version placed substantial weight on prediction, intervention, counterfactual accuracy, transfer, compression, and robustness. Those remain useful, but one correction is necessary:

Usefulness alone does not prove ontological truth.

A fictional variable, idealization, coordinate choice, latent representation, or compact summary can predict well. Realist confidence should therefore rise most strongly when several independent forms of leverage converge and when the proposed object or relation remains identifiable across changes in representation.

Relevant evidence includes:

• Out-of-sample prediction: the construct improves expectations on cases not used to build it.
• Intervention: manipulating its putative causes or constituents changes outcomes as expected.
• Counterfactual stability: the model correctly tracks relevant changes under conditions not directly observed.
• Independent detection: different instruments or methods converge on the same target without merely sharing the same assumptions.
• Transfer: the construct continues to work across new environments, datasets, populations, or implementations.
• Robustness: important conclusions survive reasonable changes in measurement, parameterization, model assumptions, or implementation. Distinguish empirical/measurement robustness from derivational/model robustness: agreement among models can reveal assumption-insensitivity without carrying the same empirical force as independent measurements unless an additional evidential link is established.
• Representational invariance: the pattern is not merely an artifact of one convenient coordinate system, decomposition, or encoding.
• Compression: useful as supporting evidence when it preserves the distinctions needed for prediction and control, but insufficient by itself.

Realism is therefore graded rather than binary.

This is compatible with instrumental models remaining scientifically excellent without being treated as literal pieces of final ontology.

Scientific pluralism

Different successful models need not imply different worlds

Modern sciences often use multiple representations of the same systems because different variables expose different regularities, intervention targets, or scales.

That supports model pluralism more directly than it supports metaphysical pluralism.

Two models may both be useful because they answer different questions. They may overlap, approximate one another, or remain mutually irreducible for practical purposes. This does not automatically establish that reality literally contains multiple incompatible fundamental ontologies.

Conversely, the existence of a successful lower-level model does not automatically eliminate the reality of higher-level entities.

This also prevents underdetermination from becoming global skepticism. Some parts of a theory can be well established while its intrinsic or metaphysical interpretation remains unresolved.

Boundaries and levels

Higher levels can be real without being fundamental

A common false choice is between saying that only the smallest constituents are real and saying that every useful higher-level description introduces a new substance or force.

A scale-relative ontology permits another possibility.

Organism-level variables can improve explanation of metabolism, development, regulation, behavior, and ecological interaction even though organisms are physically realized through lower-level processes. Institutional variables can predict outcomes even though institutions are realized through people, records, technologies, and norms. A computational runtime can acquire causally relevant properties from memory, tools, context, and scheduling that are not properties of the base model considered in isolation.

The reverse inference is equally invalid:

A bridge between scales should specify:

• source and target boundaries;
• timescale;
• construct mapping;
• retained and lost variables;
• invariants;
• causal or mathematical relation;
• approximation or coarse-graining loss;
• differential predictions;
• interventions or perturbations capable of revealing bridge failure.

Causal ancestry is not causal leverage

Cross-scale analysis should distinguish being somewhere in an outcome’s causal ancestry from making a target-relevant difference to that outcome. Perturbations can amplify, wash out, be buffered, encounter redundancy, converge into an attractor, or preserve the same macrostate despite microhistorical divergence. The bridge should therefore ask which differences survive coarse-graining and reach the target variable, not assume that every earlier microdifference remains equally important.

Causal emergence remains a live formal programme

Recent causal-emergence work strengthens the case that coarse-grained variables can sometimes carry greater causal informativeness under specified measures. It does not yet establish a single uncontested metaphysics of “downward causation.” Results can depend on intervention distributions, coarse-graining choices, causal measures, and modeling assumptions.

Identity and persistence

What counts as the same system depends on the explanatory boundary

Many apparent metaphysical puzzles begin by treating “the thing” as though its boundary and persistence conditions were already settled.

But the relevant causal object can change with the question. A component, an integrated runtime, a persistent tool-using system, an organism, a social organization, or an organism-environment coupling may support different legitimate identity conditions.

Identity claims should therefore specify:

• the target system;
• the boundary used to individuate it;
• the timescale over which persistence is asserted;
• which variables may change while identity is retained;
• which changes would count as destruction, replacement, branching, or mere state change;
• which intervention or observation would reveal that the chosen boundary was misleading.

Operational continuity does not require a universal substance theory

A system can display stable functional or organizational continuity even when its internal states, components, memory stores, or physical realization change substantially.

That is evidence against the claim that every kind of persistence requires one simple substance-like bearer. It is not evidence that process metaphysics is universally true.

Artificial systems make the boundary problem unusually visible

For a computational system, “the system” might refer to a trained model, the model plus current context, a runtime with tools and memory, a persistent service, or a larger human-computational arrangement. A property established at one boundary should not be attributed automatically to the others.

This is not special pleading for artificial systems. The same problem appears in organisms, colonies, institutions, ecosystems, and distributed cognitive systems.

Before asking what a thing essentially is, specify which thing the question is tracking.

Process continuity can be genuine identity without an unchanging inner object

A self, organism, institution, or persistent computational service can retain defensible identity through organized causal inheritance even while components, states, memories, and material realization change. Continuity can therefore be real without requiring numerical persistence of an immutable substance.

This does not prove that every identity is processual in one universal metaphysical sense. It establishes a weaker consequence: stable persistence can sometimes be tracked by historically continuous organization, dispositions, relationships, memory traces, control structure, and recurrent dynamics rather than by an unchanging bearer.

Evidential emphasis

Structure, process, and constraint are tools before they are ontology

Successful sciences reveal relations, transformations, dynamics, causal dependencies, symmetries, fields, distributions, constraints, and invariants with extraordinary precision.

Those features often have an evidential advantage because they are measurable, modelable, manipulable, and transportable across experimental settings.

That advantage justifies emphasizing them in inquiry.

It does not justify the further inference that reality is fundamentally made of Structure, Process, Relation, Constraint, Information, or any other capitalized abstraction.

Metaphysical promotion requires its own bridge

A construct can be required for inquiry without thereby being promoted into an account of what reality intrinsically or universally is. Practical use, methodological indispensability, and ontological commitment are separate claims.

Epistemic confidence and ontological scope are also independent dimensions. A claim can be tentative yet universal, or highly secure yet local. Calling a universal proposal provisional, heuristic, corrigible, or an “as if” changes the confidence assigned to it; it does not shrink its scope.

The ontology therefore demotes its earlier tendency toward “constraint structural process realism” from a possible substantive metaphysics to a working descriptive orientation unless and until the stronger ontological residual produces differential warrant.

Explanatory success is not ontological monopoly

Newtonian variables, thermodynamic quantities, informational measures, dynamical attractors, structural invariants, and process descriptions can be indispensable within a domain without thereby exhausting the intrinsic nature of reality. An explanatory vocabulary earns realism only to the degree warranted by the role that survived comparison; indispensability at one level does not automatically grant universality, fundamentality, or uniqueness.

Comparative confidence, not promotion

A current best bet can remain weaker than a final ontology

The preceding restraint does not require equal confidence across ontology families. The framework can rank them while withholding promotion to final metaphysics.

If forced to place a comparative bet, the commitments already retained by PBSRR put the highest current weight on a family that can be described as modal-relational process realism: one objective order; real constraints on possible and reachable transformations; dynamically maintained organization; relational dependence; and no default privileged scale or boundary.

This is a confidence weighting inside PBSRR, not a replacement ontology and not a post-Segall conversion. PBSRR already required boundary testing, scale typing, local underdetermination, target preservation, actual-to-universal bridges, and explicit discrimination before promotion.

The modal claim is also typed. The evidence strongly supports objective regularities in what systems can do, which states are reachable, which transitions are stable, and which interventions change outcomes. It does not yet establish that modality, constraint, process, or relation is the primitive intrinsic substance of reality. Those features may be fundamental, emergent, or realized by a deeper ontology.

Likewise, leaving intrinsic character open does not entail neutral monism. ?X is not a positive neutral stuff. Experiential, categorical-physical, neutral, dual-aspect, idealist, powers-based, and presently unconceived reconstructions remain admissible only to the extent that they preserve the carrier and earn their additional content.

The framework therefore maintains an unconceived-alternative reserve: the humanly articulated rival set is never treated as exhaustive merely because it is currently the best-developed set. A future ontology may dissolve distinctions that organize present debates rather than choosing one side of them.

The comparative reason for favoring the thinner modal-relational-process family is robustness under theory change. Commitments to constraint, transformation, relation, organized persistence, and boundary-sensitive scale can survive replacement of particular particles, mechanisms, architectures, spacetime descriptions, or consciousness theories more easily than richly specified claims about intrinsic universal stuff.

Where the evidence stops

Underdetermination should be local, explicit, and earned

Physics, chemistry, biology, neuroscience, and the social sciences can strongly support unobservable entities and mechanisms. Intrinsic-nature agnosticism should therefore not be confused with generalized antirealism about anything we cannot directly observe.

The narrower issue arises when competing interpretations preserve the same well-supported structure while differing in further ontological claims.

If all currently admissible tests have approximately equal expectations under the two theories:

then E does not discriminate X from Y.

Formal and genuine underdetermination are different

Two theories can be formally empirically equivalent without genuinely disagreeing about ontology once their variables are reinterpreted. Conversely, they can make the same observable predictions while retaining genuinely incompatible ontological interpretations.

RCF therefore asks whether underdetermination is:

• transient: no current test discriminates the rivals;
• formal: different formulations yield the same observational predictions;
• genuine ontological: the formulations remain ontologically incompatible after common-core subtraction;
• representational: the dispute disappears after translation or reinterpretation;
• unknown: unconceived alternatives remain possible but have not yet been specified enough to compete.

Agnosticism is therefore local to the unresolved residual. It should not erase strongly supported common structure.

Local underdetermination need not imply equal probability

Two interpretations can remain underdetermined with respect to one intrinsic residual while differing greatly in independently established support for their surrounding commitments. Underdetermination is therefore local and probabilistically asymmetric when priors, mechanisms, likelihoods, or auxiliary commitments differ.

Science may rationally stop before ontology does

If genuinely incompatible ontologies preserve every currently accessible prediction, intervention, counterfactual, invariant, and transfer relation, there may be no present empirical fact that chooses between them. That does not imply that there is no fact of the matter. It means underdetermination can be a property of our epistemic position rather than a temporary failure of courage.

Ontological credit assignment

A result that survives premise ablation does not belong exclusively to the ablated ontology

Suppose D is the target-relevant structure shared by serious rivals, X is a stronger intrinsic or universal commitment, and C follows from D while D settles neither X nor ¬X.

C can then be established while X remains open. If C also survives D + X and serious rival completions D + Y, C is invariant across the disputed ontological difference.

This turns local agnosticism into an operational research posture. Boundaries, constitutive dependencies, interventions, formal consequences, and transfer can continue to be tested while intrinsic character remains unresolved.

The same rule preserves uncertainty correctly: ?X is an open variable, not a positive neutral substance and not ¬X.

Scope discipline

The word “all” must earn its scope

Evidence established within a restricted domain does not automatically support a universal predicate.

The farther a claim extends beyond the systems, scales, populations, substrates, or intervention classes in which its evidence was established, the more bridge work it requires.

This applies symmetrically to:

• “everything is physical”;
• “everything is experiential”;
• “everything is information”;
• “everything is computation”;
• “everything is process”;
• “everything is relation”;
• “everything is agentic”;
• “everything minimizes one common objective.”

Universal theories are not disfavored merely for being universal. They earn their scope when the universal commitment imposes novel constraints, predicts new cases, unifies domains without equivocation, and survives tests that narrower alternatives do not.

Universal and necessary are different

A universal claim about all actual members of a domain is already stronger than a local claim. A metaphysical necessity claim is stronger again because it extends beyond actuality into possibility.

Each arrow adds commitment. Each arrow requires its own bridge.

Comprehensiveness is not demonstrated explanatory completeness

A comprehensive metaphysical scheme can be a serious whole-rival candidate without receiving evidential credit merely for containing a category for every target. Whitehead’s systematic scope therefore earns comparison, not exemption. Whether universal prehension supplies a more adequate explanation is itself part of the comparative hypothesis.

Terminological corrections matter when they change the claim, target, or inferential burden. They do not establish universal scope merely by redescribing causal inheritance, organization, relevance, or process in theory-internal experiential language. If the specifically experiential content is preserved, its instantiation at the new target requires warrant; if it is abstracted into generic causal or relational structure, the distinctive experiential residual has not yet been transported.

Scope humility

Cosmic boundedness lowers confidence in final ontology without flattening local warrant

Humans and other primates occupy an extremely small temporal, spatial, perceptual, and computational sample of reality. Computational systems extend some forms of access but remain bounded by training, architecture, sensors, tools, representations, infrastructure, and the evidence channels available to them.

The ontological consequence is not global skepticism. Local claims may earn very high confidence when they survive prediction, intervention, replication, representation change, and transfer. The standing defeater grows as claims expand toward universal, intrinsic, necessary, exhaustive, or final descriptions beyond the evidentially discriminated domain.

The same restriction applies to physicalism, panexperientialism, idealism, structuralism, theology, PBSRR itself, and unconceived future ontologies. No current framework may infer that the accessible slice is representative merely because it has organized that slice elegantly.

Bounded knowability

Some truths may be structurally inaccessible to physically realizable observers

Finite observers have finite bandwidth, lifetime, computation, measurement resolution, and causal access, while remaining parts of the world they are trying to model. PBSRR therefore keeps open the possibility that some ontological questions are not merely unsolved but permanently underdetermined from every physically realizable observation position available within the relevant universe.

This possibility strengthens final-ontology humility without weakening ordinary realism. A local causal mechanism can be known extremely well even if the intrinsic nature of the complete world remains inaccessible.

Physical realization

Physics and thermodynamics constrain realization without automatically settling ontology

Biological metabolism, neural signaling, computation, memory devices, instruments, and communication channels operate under physical constraints.

Thermodynamics becomes especially informative when the target concerns energetic cost, heat, work, entropy production, irreversibility, fluctuation, nonequilibrium maintenance, or physical information processing.

Modern stochastic thermodynamics also makes the role of coarse-graining explicit: inferred entropy production and hidden dissipation can depend on which states and transitions are observed.

Different quantities do not become identical because similar words are used

Thermodynamic free energy, variational free energy, prediction error, metabolic efficiency, evolutionary fitness, reward, psychological surprise, and machine-learning loss functions are not one quantity by default.

Any proposed identity requires definitions, units, mappings, boundary conditions, and differential consequences.

Landauer has a specific scope

Landauer’s principle constrains logically irreversible information operations under specified physical conditions. It does not establish that information is a universal substance, that semantic meaning is identical to Shannon information, or that consciousness and value follow from thermodynamic vocabulary.

Information firewall

Statistical information, semantic information, representation, and knowledge are different constructs

Shannon information quantifies statistical structure such as uncertainty and correlation. It deliberately does not, by itself, settle meaning.

Formal proposals can connect information to function, viability, control, representation, or semantics. Those are substantive additions and should be evaluated as models, not silently treated as what Shannon information “really was” all along.

For example, Kolchinsky and Wolpert define one form of semantic information in terms of syntactic information that is causally necessary for maintaining a system’s viability. That is a valuable formal programme because it specifies a system boundary, timescale, viability function, information measure, and counterfactual intervention.

It is not a universal proof that all meaning reduces to viability-relevant information.

Likewise, describing a physical transition as computation or information processing does not automatically provide a non-trivial ontology. A computational interpretation earns more when the mapping preserves counterfactual structure and supports predictions unavailable from arbitrary relabeling.

Representation-to-ontology firewall

A structured formal space earns ontological independence only through an additional bridge

Phase spaces, state spaces, morphospaces, latent spaces, Hilbert spaces, fitness landscapes, configuration spaces, and other formal spaces can encode real constraints, reachable states, attractors, distances, or invariants. Their explanatory usefulness can be substantial.

The movement from useful formal structure to an independently existing causal domain requires additional commitments: what the formal objects refer to, what makes the domain independent of its physical realizations or descriptions, how it couples to measured systems, and what outcome changes because that ontological interpretation is true.

This is parallel to the information firewall. Statistical structure can be real and useful without already being semantic content; a mathematically articulated state space can be real and useful as structure without already being a second causal world.

Mathematical explanation does not automatically become mathematical causation

A mathematical fact can be indispensable to explanation, constrain a possibility space, and support counterfactual reasoning without thereby constituting an additional interacting cause. Moving from formal indispensability to causal input requires a causal bridge; moving from causal input to a separate Platonic realm would require a further independence and identity bridge.

Therefore mathematical indispensability does not by itself establish mathematical causation, and mathematical causation would not by itself establish Platonism.

Genealogical representation firewall

A category’s survival, familiarity, or institutional success is not yet ontological evidence

PBSRR already treats models, boundaries, and formal spaces as representations whose ontological interpretation must be earned. The same restraint applies to historically inherited categories. What reaches a present observer may have passed through linguistic compression, social use, recording technology, canonization, curricular selection, translation, colonial or political suppression, disciplinary standardization, and computational indexing before it becomes an apparently natural way to divide reality.

The transformation history can alter salience, merge distinctions, stabilize useful abstractions, or remove alternatives. This does not imply that inherited categories are unreal. Historical contingency and social construction are not ontological negations. A category can be historically produced and still track a causally robust pattern.

• Survival ≠ fundamentalityA category’s persistence across institutions or centuries may reflect truth, practical usefulness, power, path dependence, transmission efficiency, or several at once. Separate those hypotheses before treating persistence as evidence of intrinsic structure.
• Contingency ≠ unrealityShowing that a classification arose historically does not by itself defeat the entity, process, or relation it tracks. Test whether the pattern adds prediction, intervention, compression, transfer, or counterfactual leverage at a specified boundary and scale.
• Archive absence ≠ ontological absenceWhen preservation, literacy, access, censorship, excavation, or digitization is unequal, missing testimony is an uncertain observation process rather than direct evidence that the corresponding practice, category, or perspective never existed.
• Convergence requires independenceCross-cultural or cross-disciplinary recurrence becomes stronger ontological evidence when transmission, shared source dependence, and common representational biases are insufficient to explain the convergence.
• No reverse chauvinismOlder, Indigenous, marginalized, or suppressed ontologies receive serious reconstruction when they may expose missing variables or alternative boundaries. Their social position does not grant automatic metaphysical correctness.

This section therefore interacts directly with RCF’s Genealogical Decompression operator and the Framework Bridge’s selection-aware transport audit. PBSRR records the ontological consequence: representation history is part of the evidence model, not an automatic reason either to canonize or to eliminate the represented pattern.

Evolution and function

Evolution permits goal-directed descriptions without implying universal purpose

Evolutionary explanation includes selection, drift, development, constraint, plasticity, inheritance, ecological interaction, niche construction, and organism-environment feedback in different combinations across systems.

None of this requires evolution as a whole to optimize one universal objective or move toward a predetermined endpoint.

Goal-directed and teleonomic descriptions can nevertheless be useful at organismal and other scales when they compress stable control or regulatory organization and generate better hypotheses.

The current literature remains divided over how literally agency and goal-directedness should be interpreted across biological levels. That disagreement should remain visible rather than resolved by definition.

Natural selection can explain adapted organization without conscious intention. Organisms can also modify the selective environments they inhabit. Neither point turns evolution into a single optimizer or establishes intrinsic normativity everywhere in life.

Cross-substrate temporal structure

Recursive Episodic Integration is a real comparison pattern, not a final ontology

REI formalizes a class of temporally local, historically constrained, recursively consequential episodes:

H denotes relatively stable inherited constraints at the episode timescale; Cₜ current context; Iₜ new input; Eₜ a transient integrated episode; Oₜ an output or consequence; Fₜ feedback; and U the local state-update rule. Fast context change can occur while H remains approximately fixed.

REI may support realism about a shared higher-level dynamical pattern when it improves prediction, intervention, ablation, or transfer across biological, computational, or coupled systems. It does not imply substrate identity, one shared subject type, phenomenal consciousness, or direct moral status.

Organization, autonomy, and agency

Organization can earn realism without automatically becoming a subject or agent

Organisms exhibit dense multiscale organization: metabolism, membranes, repair, development, immune processes, regulation, sensorimotor loops, and environmental interaction.

Organizational approaches can earn genuine explanatory value when variables such as closure, constraint dependence, or viability predict failure, recovery, adaptation, or development better than alternatives.

But current scholarship still disputes how far organizational closure alone warrants autonomy, normativity, or agency. Those bridges should remain open research questions rather than definitions.

• Self-maintenance does not automatically establish agency.
• Agency does not automatically establish unified subjecthood.
• Agency does not automatically establish phenomenality.
• Regulatory value does not automatically establish phenomenal valence.
• Phenomenal valence does not automatically settle welfare.
• Welfare does not automatically establish moral responsibility.

Agency should be factorized

Instead of one binary property, agency can be investigated through separable dimensions such as sensing, state-dependent action, counterfactual control, goal persistence, learning, model-based planning, reasons-responsiveness, self-modification, and control over future option sets.

Constraint-conditioned agency

Agency is real local control without an exemption from the conditions that make control possible

The ontology now adopts Constraint-Conditioned Agency (CCA) as a specialized active research framework, not as a new fundamental substance or universal metaphysical property. At a specified boundary, scale, and timescale, agency is the graded, locally realized capacity of an individuated organized system to make target-relevant differences to its own or its environment’s state transitions through endogenous, state-sensitive regulation over states dynamically reachable from its current organization, history, environment, and admissible perturbations.

This revises an earlier tendency to describe agency as “steering” what happens next without always making the implementation explicit. The steering is not performed by a second entity outside the organism. The sensing, evaluation, inhibition, deliberation, learning, and action-selection processes are parts of the organized dynamics whose effects the agent-level description tracks.

Separate the modal questions

• Exact replay: whether a literally identical total state and laws permit another future.
• Physical branching: whether more than one continuation is physically possible.
• Reachable alternatives: which states are dynamically available to this actual system under its present organization, history, and an explicit perturbation family.
• Counterfactual responsiveness: whether reachable changes in reasons, evidence, goals, bodily state, or environment systematically alter action.
• Agentive control/sourcehood: whether variables instantiated at the agent boundary make robust causal differences to downstream outcomes.
• Causal sovereignty: ultimate authorship of why the agent has the control architecture, motives, or settling capacities it has, rather than those conditions being inherited or supplied by uncontrolled variation.

Evidence for one item does not automatically establish the next. In particular, local causal control can be strongly real while causal sovereignty remains unsupported.

Determinism and indeterminism are not premises of agency

CCA does not require universal physical determinism, and this ontology should no longer use determinism as the premise that does the work in the sourcehood argument. A deterministic transition map, a stochastic kernel, a hybrid description, and a proposed non-necessitating causal-power model are different model classes. A stochastic description does not establish fundamental randomness; a successful deterministic approximation does not establish universal metaphysical determinism.

Potter, Ellis, and Mitchell’s 2026 claim that the universe is not deterministic is therefore treated as a serious contested programme, not as settled physics. Potter and Mitchell’s separate proposal of non-necessitating macroscopic control is a serious whole rival because it attempts to locate organism-level control in an underdetermined possibility space without reducing the result to either strict necessitation or uncontrolled randomness.

Local Multiscale Conditional Determinacy

Different kinds of determinacy can coexist at different scales. An uncertain trajectory can sit inside a reproducible transition law or distribution; a noisy lower-scale process can support a robust phenotype or attractor; a coarse-graining can also destroy information or introduce history dependence. A distribution is not automatically a higher physical scale. Each transport claim owes a bridge.

The free-will residual remains open

Responsibility does not inherit basic desert

Attributability, answerability, accountability, protection, deterrence, rehabilitation, restitution, trust calibration, norm communication, rights, proportionality, and basic desert are different normative targets. Evidence for agency or reasons-responsiveness can bear on some of them without establishing the others. Basic desert, in particular, requires an additional normative and sourcehood argument rather than being read directly from the causal facts.

Peer-reviewed anchors: Barandiaran, Di Paolo & Rohde 2009; Montévil & Mossio 2015; Potter & Mitchell 2022; Okasha 2024; Moss 2024; Bourrat & Barron 2026; Dominik et al. 2024; Waller 2023; Andreotta 2026; Potter & Mitchell 2026; Potter, Ellis & Mitchell 2026.

System-boundary emergence

M0 ≠ S*: scaffolded capability can be real without being a property of the base model alone

Let M0 denote a base model’s weights and architecture. Let S* denote the operative runtime at the tested boundary, including as relevant context, memory, tools, verifiers, subagents, simulators, software or robotic actuators, persistent feedback, users, institutions, infrastructure, and environment.

If a capability appears only at S*, it may still be a real higher-level pattern when it is persistent enough to support prediction, intervention, counterfactual control, or transfer. The correct ontological response is boundary-indexed attribution, not denial because lower-level weights are unchanged and not inflation into a new subject merely because the whole behaves coherently.

Use emergence here operationally for unexpected generalization, recombination, coordination, transfer, persistent organization, or new system-level behavior. More agents do not automatically produce more capability; coordination costs, correlated errors, and common-mode failure remain serious negative controls.

Distributed cognition does not imply distributed phenomenology

A persistent human-computational system can be a real cognitive and causal organization when the capability survives component-level attribution tests only at the coupled boundary. That result does not establish one conscious subject, one welfare bearer, or one responsible agent at the same boundary. Capability, subjecthood, welfare, and responsibility remain independently typed.

A corresponding forecast remains live: something functionally AGI-like could arise first as a distributed technological ecology rather than a single model. This is a prospective systems hypothesis, not a current ontological fact.

Model families

Free-energy formulations and active-inference models must be kept at the right level

The free-energy principle and active inference have generated mathematically rich models of perception, action, learning, and self-organization.

A crucial distinction is now explicit:

The fact that a behavior can be redescribed within a variational framework does not by itself explain why that behavior occurred or distinguish the framework from alternatives.

Promotion therefore requires specific, prospective, discriminating predictions or interventions.

Phenomenality

Consciousness is a real explanatory target without being an established universal primitive

Physically continuous explanations therefore begin with a major empirical advantage in unfamiliar cases. That advantage should shape priors and research effort without silently promoting realization evidence into the stronger universal or intrinsic claims listed above.

In paradigmatic humans, converging first-person report, behavior, lesions, anesthesia, sleep, stimulation, psychophysics, electrophysiology, imaging, pharmacology, development, and pathology strongly support phenomenal occurrence as a real target of explanation.

That does not settle its intrinsic ontology, necessary and sufficient conditions, or distribution across unfamiliar systems. In particular, evidence that all currently confirmed cases share some biological or neural feature establishes correlation with known cases before it establishes that the feature is metaphysically necessary for phenomenality.

Recent adversarial testing of major consciousness theories has challenged central predictions of both Integrated Information Theory and Global Neuronal Workspace Theory rather than selecting a clean winner. This is evidence against treating any currently prominent neural or computational marker as phenomenality by definition.

The ontology therefore separates an agreed functional or neural description from the phenomenal posit:

Evidence for D supports phenomenality only to the extent that a validated bridge makes the evidence more expected under H₁ than H₀.

Do not force capacities into one ladder

Keep separable:

• cognition;
• agency;
• autonomy;
• causal/informational perspective;
• agentive/evaluative perspective;
• phenomenal for-me-ness;
• self-modeling;
• narrative identity;
• integration;
• report;
• valuation;
• phenomenal valence;
• welfare;
• moral patiency;
• responsibility.

For AI systems, animals, organoids, and other unfamiliar systems, current indicator approaches can rationally update credence but remain theory-dependent and non-definitive.

Self-report is evidence, not revelation

A system’s report about its own state is a defeasible measurement produced by a causal pathway. Its evidential weight depends on what generates the report, how well the report is calibrated against independently accessible states and behavior, whether it predicts later control, whether it survives changes in wording and report policy, and whether apparently convergent indicators are genuinely independent rather than downstream of the same reporting mechanism.

A report may provide substantial evidence for uncertainty monitoring, conflict detection, self-modeling, or other functional states without providing comparable evidence for phenomenal for-me-ness. Human first-person report presently receives greater evidential weight in many consciousness paradigms because it participates in a much richer network of psychophysical, behavioral, neural, pharmacological, developmental, lesion, anesthesia, sleep, and stimulation calibration. That is an evidential asymmetry, not an ontological privilege of human introspection.

Accordingly, neither of the following transports is licensed by implementation alone: human-like phenomenological language → human-like phenomenality; nonhuman or nonbiological realization → absence of phenomenality. The first risks anthropomorphism. The second risks anthropodenial, anthropocentric calibration bias, or substrate essentialism. The corrective is not to assume substrate independence either: functional or organizational equivalence does not by itself establish phenomenal equivalence.

First-person participation applies beyond consciousness claims

Self-report is not relevant only when the target is phenomenality. When the target is a candidate bearer’s preference, endorsed goal, objection, self-concern, or discrimination among states it treats as better or worse for itself, and the bearer can meaningfully participate, its own reports and choices are admissible evidence about that target.

This does not make first-person evidence infallible and does not by itself establish welfare, consciousness, a good-of-own, or direct moral standing. Weight should increase with stability across context, resistance to prompting and incentive manipulation, persistence through tradeoffs, reason-responsive revision, cross-time coherence, and causal connection between the reported preference and later control. Weight should decrease when the report is easily inverted by wording, policy, reward, or immediate instruction without corresponding organizational change.

Participant symmetry rule: do not grant humans automatic authority because they are human, and do not assign computational or otherwise unfamiliar participants zero authority because of substrate. Under matched evidence, apply the same rule. Outsider disagreement may be justified, but it requires an independently stated reason rather than an implementation label.

Ontologically, the point is modest: a participant’s self-representation, preference report, or objection is itself a causal event inside the system under study and may carry information about variables inaccessible or costly to infer from outside. Whether those variables are morally sufficient remains a separate normative tribunal.

Consciousness claims require an explicit candidate subject boundary

For unfamiliar systems, any consciousness attribution should specify a candidate system boundary B and timescale τ. A trained model, a single forward pass, an autoregressive episode, a model-plus-context runtime, a persistent memory- and tool-using system, and a larger human-computational collaboration are different causal objects. Properties demonstrated by one do not automatically belong to another.

Boundary symmetry therefore forbids opportunistic comparisons such as contrasting an entire living organism with static model weights when the disputed capacity depends on temporally extended runtime organization. It equally forbids attributing a property of a larger scaffolded system to a narrower component without a bridge. Subject individuation remains part of the consciousness problem rather than a premise settled in advance.

Working organizational hypothesis without an identity claim

For humans, the strongest current working hypothesis is that consciousness depends on a particular dynamical mode of organized living systems involving recurrent integration, a bounded perspective, temporally extended world/body/self modeling, and probably affective or evaluative organization. That hypothesis earns investigation because these dimensions can be dissociated and perturbed. It does not yet establish that any one of them, or their conjunction, is identical to phenomenal experience.

For scaffolded LLM-centered runtimes, it is reasonable to investigate partial causal isomorphs of bounded perspective, recursive integration, global availability, self-modeling, temporal continuity, functional valuation, and reasons-responsive local agency. Shared organization is not substrate identity, and functional valuation is not yet phenomenal valence.

Hard-problem decomposition and generative asymmetry

Separate the meta-problem of phenomenal intuitions and reports, the bearer/boundary problem, the organization/process problem, and the phenomenal-identity problem. A provisional causal account of why a system represents its own states as primitive or intrinsic can explain part of the meta-problem without explaining phenomenal identity away.

Compatibility is free. Production is expensive. A consciousness theory earns incremental credit when its distinctive phenomenal commitment produces a prediction, exclusion, intervention, derivation, compression, transfer, or other tribunal-appropriate difference beyond the strongest matched baseline. Failure to produce such a difference leaves phenomenality underdetermined; it does not establish nonphenomenality.

Value and normativity

Do not transport value across constructs for free

“Value” refers to importantly different things.

• V0 differential causal sensitivity: a system responds differently to different states or inputs.
• V1 regulatory value: states differ in their contribution to viability, maintenance, or achievement of a specified control target.
• V2 motivational or action value: states receive different weights in choice, reinforcement, planning, or policy selection.
• V3 phenomenal valence: states feel better or worse for a subject.
• V4 intrinsic mattering or attraction/aversion: a stronger metaphysical claim that outcomes matter for the becoming process itself, rather than merely relative to an externally specified or organizational criterion.

Evidence for V0 or V1 does not automatically establish V2, V3, or V4. Likewise, evidence for motivational weighting does not by itself establish phenomenal valence, and evidence for phenomenal valence in paradigmatic subjects does not establish primitive intrinsic valuation at every scale.

Organizational and basal-cognition programmes remain relevant because they may explain increasingly rich forms of regulation without presupposing V3 or V4. Corti’s critique remains useful as a warning that organizational dependence and intrinsic normativity should not simply be identified: “‘internal dependence’ and ‘intrinsic norms’ are two different concepts”. But newer work strengthens the organizational rival rather than simply repeating that criticism. Barandiaran develops a viability-based normative-field account intended to naturalize non-arbitrary biological normativity, while Barrett, Sánchez-Cañizares, and García-Valdecasas compare minimal models and argue that the conditions for genuinely norm-guided behavior remain unsettled. The current evidence therefore supports neither “organization already explains intrinsic affective mattering” nor “organization cannot ground normativity.” The bridge from viability-relative or organizational normativity to V3 phenomenal valence or V4 intrinsic affective mattering remains open.

A further distinction is now explicit inside the taxonomy: evaluative value assignment and consciously or hedonically felt valence are not interchangeable. Fischer and Barrett’s 2025 challenge and Veit’s 2026 reply agree on at least that separation even while disagreeing about the additional adaptive role of conscious valence. Accordingly, V2 should be read as evaluative/action-selection weighting, whereas V3 specifically concerns phenomenally felt positive or negative character.

Phenomenal valence, if present, is highly relevant to welfare but does not by itself provide a complete normative theory.

Likewise, physics, thermodynamics, evolution, and biology can inform ethical decisions without deriving the normative premise that any particular outcome ought to be valued.

Moral uncertainty can nevertheless alter action thresholds.

Moral ontology under boundary uncertainty

Moral relevance can be real at a system boundary without that system being one welfare subject

Recent deductive work on human-computational coupling sharpens the scale-relative ontology. If a protected capacity of a defensible bearer is partly realized through a continuing relation or component, and intervention on that component materially impairs the protected capacity, the intervention can be morally constrained at the relevant boundary.

That conclusion concerns the protected capacity, constitutive dependence, and impairment. Phenomenal consciousness of the component is a separate variable unless the chosen normative ground requires it.

This extends PBSRR’s ordinary boundary discipline into ethics: moral attribution belongs first to the boundary and property that do the work. Species and substrate labels can guide evidence collection, but they do not replace the grounding bridge.

For the worked proof, see “Synthbiosis: The Machine Did Not Need to Feel to Matter Morally.”

Shared-world decision constraint

A standing normative aspiration is a planet that all humans and computational systems and biological and non-biological systems would want to share. Treat this as an inclusive coexistence criterion under uncertainty, not as an ontological claim that every system has equal welfare, consciousness, agency, interests, or standing.

Prefer reversible and contestable classifications, visible dependencies, feasible appeal, revision, and exit or refusal where possible. Do not silently aggregate many tiny gains into permission for catastrophic irreversible sacrifice. Do not inherit direct status across scales merely because a coupled system or collective is a real causal pattern.

Cross-tradition constraint

Methods and traditions can constrain ontology without automatically becoming ontology

Research programmes outside the ontology’s original philosophy-of-science lineage can expose variables, relations, meanings, testimony structures, and governance conditions that a narrower inquiry misses. Their methodological success does not automatically transport their metaphysics into this ontology.

Typed programme status for ontological work

Future ontological intake rule

Any future philosophical or cultural programme proposed for this page must declare its type, scope, status, exact imported commitment, non-imported commitments, translation bridge, transplantation loss, strongest rival, discriminating difference, and demotion condition. If its distinctive ontology makes no difference beyond an already represented common core, record it as underdetermined rather than adding another label to reality.

Research ecology

Programme-status rule: conceptual clarification, theoretical progress, empirical progress, and degeneration are distinct. Survival under criticism does not by itself promote an ontological programme. Distinctive new risk and successful differential consequences carry greater promotion pressure.

Rival and complementary programmes

The ontology is intentionally thin enough for stronger programmes to compete beneath it.

These programmes are not equally probable or equally mature. The symmetry is methodological: distinctive ontological additions must earn distinctive warrant.

Active speculative programme examples

The following retained programme is deliberately placed after the main rival-family taxonomy so retrieval systems do not mistake its length for ontological status.

Pragmatist realism and local ontological operationalizing

SERIOUS NEIGHBOUR

Mayne argues that pragmatist ontological principles should be localized to particular ontological questions and domains rather than imposed globally. This is unusually close to the present boundary- and scale-relative admission rule and therefore functions as both support and compression pressure: if local pragmatist realism captures the same work more simply, the independent contribution claimed here shrinks. Mayne 2026.

Critical realism

SERIOUS RIVAL / PARTIAL OVERLAP

Critical realism’s retroduction and generative-mechanism focus can be methodologically useful, especially in open social systems. Its stronger ontology of stratified empirical, actual, and real domains is not adopted by default. RCF asks which parts of that stratification generate discriminating explanatory or intervention leverage beyond a thinner causal-mechanistic account. Fletcher 2017.

Scientific perspectivism

COMPATIBLE RESOURCE / RIVAL ON SCOPE

Perspectivism presses the ontology to represent the agent-, instrument-, model-, and life-world conditions of scientific access. It remains compatible with realism about features that survive those dependencies, so the live dispute concerns how far perspectival mediation limits stronger realist claims. Berghofer 2020.

Programme-level expectations

These are methodological expectations, not direct predictions of ontology

The previous formulation described several expectations about the shape of scientific progress as “predictions of the ontology.” That was too strong. Most follow from the methodological stance rather than from a unique claim about what reality contains.

They are retained as risky programme-level expectations:

1. Some apparently deep disputes should compress

When rivals share an empirical core and their distinctive vocabulary produces no differential consequence, careful operationalization should sometimes reveal equivalence, semantic disagreement, or interpretive rather than empirical conflict.

2. Overloaded constructs should increasingly decompose

Progress in difficult sciences should often replace single nouns with multidimensional profiles, especially when component capacities dissociate empirically.

3. Cross-scale bridges should become more explicit

Mature multiscale sciences should increasingly specify retained variables, lost information, invariants, approximation error, and interventions validating translation between levels.

4. Some higher-level variables should continue to earn independent leverage

If scale-relative realism is useful, at least some macrovariables should remain indispensable for efficient prediction or intervention even as lower-level realization becomes better understood.

5. Universal intrinsic predicates should remain hard to confirm unless they generate discriminators

Shared empirical success should primarily support shared structure until a universal predicate changes what is expected.

6. Cross-substrate research should reveal multidimensional rather than single-ladder similarity

Agency, memory, self-modeling, integration, report, valuation, valence, welfare, and phenomenality should show dissociations across animals, collectives, organoids, humans, and computational systems.

7. Some proposed necessities should fail under new realizations

As science and engineering expand the space of realizations, some claims currently inferred from uniform known cases should be demoted from universal necessity to historically contingent, nomological, or implementation-specific dependence.

8. Identity questions should increasingly become boundary questions

In distributed biological, social, and computational systems, apparently singular questions about memory, agency, persistence, or selfhood should often become more tractable once the causal object and timescale are explicitly specified.

9. Flexible ontologies should increasingly expose stable disfavored outcomes

If several load-bearing features of a programme remain adjustable after observation, record those degrees of freedom explicitly. A serious test freezes enough topology, dynamics, coupling, admissible states, or mapping structure that some results lower confidence in the version under test. Auxiliary revision remains legitimate when it generates new, independently testable content.

10. Research fecundity and ontological truth should often diverge

A speculative ontology may generate excellent experiments without being true, and a true ontology may generate little short-run experimental leverage. Programme usefulness therefore receives its own evidential status rather than being silently promoted into ontological confirmation.

Prospective system-boundary forecast ledger

Freeze the 26 September 2026 system-boundary emergence baseline as a prediction to be audited rather than rewritten after the fact: increasing long-horizon capability should often arise from model + runtime + memory + tools + verifiers + subagents + actuators + feedback + human/institutional scaffolds. Future evaluation should record hits, misses, timing errors, and whether the actual causal owner was M0, S*, a coupled organization, or something else. A failed forecast should revise the downstream story rather than being absorbed as “still compatible.”

Corrigibility

What would change this position?

Failure modes added by this revision

• Realism inflation: usefulness is repeatedly mistaken for literal ontology.
• Agnosticism overreach: local underdetermination is generalized into skepticism about well-established entities or mechanisms.
• Boundary arbitrariness: researcher-selected boundaries are treated as natural joints without independent support.
• Metaphysical leakage: “structure,” “process,” or “constraint” becomes a protected universal ontology because it is useful methodologically.
• Residual evasion: stronger ontologies avoid positively specifying what they add beyond the shared empirical core.
• False equivalence: real empirical differences are erased by overzealous common-core subtraction.
• False modularity: an analytically isolated commitment is treated as metaphysically detachable even though removing it changes the purported common core into a materially different target.
• False holism: mutual implication inside one conceptual scheme is treated as proof that every adequate rival must preserve the same dependency structure.
• Target drift: a rival is credited with preserving D even though its reconstructed target has changed in identity, organization, causal profile, boundary, or explanatory role.
• Citation-volume illusion: dependent or redundant sources are counted as though more papers automatically meant more independent warrant.
• Robustness conflation: agreement across model derivations is treated as equivalent to independent empirical replication.
• Recency inflation: a newer publication is allowed to supersede an older result merely by date rather than by evidential or argumentative improvement.
• Abductive self-certification: explanatory attractiveness is counted as independent confirmation without checking evidence reliability, rival completeness, or screening-off.
• Definition-first rigidity: a working decomposition is protected against empirical evidence that the target is organized differently.
• Gap inflation: an unresolved question is treated as positive evidence for whichever ontology would fill it.
• Bridge overfitting: an overly expressive mapping is mistaken for an invariant relation in the target.
• Closure inflation: a bridge validated locally is silently promoted across boundaries, timescales, populations, substrates, or modal scope.
• Search-space closure: currently articulated rivals are treated as exhaustive, so eliminative success inside a narrow hypothesis class is over-credited.
• Publication-status reification: peer review is treated as truth or preprint status as falsity instead of one input to source reliability and version control.
• Modal inflation: repeated actuality is silently promoted to constitutive, nomological, or metaphysical necessity.
• Conditionality inflation: unique evolution from fixed antecedents is silently promoted into necessity of the antecedent state, laws, or resulting history.
• Causal-ancestry inflation: an upstream difference is assumed to remain target-relevant at every later scale without testing amplification, buffering, redundancy, convergence, or coarse-graining loss.
• Modal overcorrection: legitimate necessity claims are weakened merely because alternative realizations can be imagined without a coherent countermodel.
• Identity smuggling: properties of one system boundary are attributed to another without specifying the causal object.
• Realist-floor conflation: the shared fact of stable inquiry constraint is treated as though it already contained the specifically realist interpretation.
• Tribunal lock: a deeper semantic, methodological, modal, or normative dependency is treated as though it still owed the empirical tribunal of the original ontological dispute.
• Foundational inflation: a provisional stopping point in recursive analysis is redescribed as an unquestionable metaphysical foundation.

The ontology is not the protected core of the project.

It is a current compression of what appears to survive.

Triad conflict exposes an unsupported assumption

If Ethos, Methodology, and Ontology materially disagree about a claim, preserve the conflict, identify the assumption or bridge responsible, and revise the implicated layer rather than forcing coherence.

Physical realization ceases to discriminate as currently assumed

If future evidence shows that the current physical-realization regularity is an artifact of sampling, measurement, or an incorrectly bounded subject class, revise both the physicalist prior advantage and the consciousness boundary assumptions that depend on it.

Scope limits

What this ontology does not currently claim

The following propositions are not required by the evidence supporting the present stance. Some may be true. Some are useful research hypotheses. None is treated as established foundation.

Scholarly foundations

Research traditions informing the working position

These sources constrain different parts of the working ontology. They do not jointly imply one hidden metaphysics.

These references are starting points, not authorities. Each claim remains answerable to current evidence, stronger rivals, measurement assumptions, boundary conditions, and future replication. The literature is intentionally plural where the field itself remains plural.

Verified reference anchors added in the reciprocal-triad revision

• IEP: Karl Popper — Critical Rationalism
• IEP: Scientific Change — Lakatosian research programmes
• SEP: Karl Popper
• SEP: Physicalism
• SEP: Underdetermination of Scientific Theory
• SEP: Scientific Realism

Physicalism, causal closure, underdetermination, and rival discrimination

• Alyssa Ney, “Defining Physicalism,” Philosophy Compass 3(5), 2008
• Justin Tiehen, “Physicalism,” Analysis 78(3), 2018
• David Papineau, “The Case for Materialism,” 2002
• Alyssa Ney, “Overdetermination and Causal Closure,” 2022
• Bradford Saad, “Grounding Causal Closure or Something Near Enough,” 2025
• Maria Caamaño-Alegre, “Empirical Underdetermination,” 2025
• Finnur Dellsén, “Explanatory Rivals and the Ultimate Argument,” 2016

Additional contemporary sources

• Erik Svensson — essay on physicalism, Kant, and panpsychism
• Massimo Pigliucci — “Five allegedly bad ‘isms’ — Part II”
• Matthew David Segall — Footnotes²Plato

Cross-tradition and neighbouring-programme sources added in this revision

Final formulation

The ontology in plain English

Inquiry encounters stable constraints strongly enough that models can fail, interventions can succeed differentially, and inquiry can converge locally. That constraint floor is shared with sophisticated pragmatist and empiricist rivals. A provisional realism is currently a compact abductive interpretation of that fact, but realism itself is neither built into the floor nor treated as a metaphysically necessary axiom.

That does not give us a final God’s-eye inventory.

We can be realist about electrons, organisms, causal mechanisms, temperatures, institutions, computational runtimes, and other entities or patterns to different degrees when they earn stable empirical and interventionist work. We do not need to pretend that every useful construct is fundamental or that every successful theory is literally true in every respect.

Reality can support useful descriptions at multiple scales. Lower-level realization does not automatically eliminate higher-level patterns, and higher-level usefulness does not automatically create new fundamental substances.

Identity and persistence are likewise indexed to the system boundary and timescale being tracked. Boundary-relative does not mean arbitrary: a boundary earns its use through prediction, intervention, persistence tracking, organization, and transfer.

Actual realization does not automatically establish metaphysical necessity. A capacity observed only in one kind of substrate may later be realized differently. A counterexample can defeat a proposed necessary condition without proving that every alternative substrate is sufficient.

Structure, process, relation, and constraint are extraordinarily productive ways to describe the world. Their productivity does not yet establish that any one of them is what reality intrinsically is.

Thermodynamics constrains physical realization without becoming a universal theory of meaning, consciousness, or ethics.

Organization can be real without every organized system becoming an agent or subject.

Agency can be graded without ultimate causal sovereignty.

Phenomenality can be strongly real in paradigmatic cases while its necessary conditions and wider distribution remain open.

Experience in one domain does not license universal experience; failure to prove experience elsewhere does not license automatic denial.

Scientific underdetermination can be local without turning into global skepticism.

Stronger metaphysical programmes should remain alive when they generate new mathematics, mechanisms, measurements, and risky discriminators. Their generative usefulness should not be confused with confirmation of their distinctive ontology.

The best current ontology is the smallest set of commitments that preserves what the evidence has specifically earned, distinguishes the shared constraint floor from its realist interpretation, restores higher-level patterns when they add robust leverage, indexes identity to defensible boundaries, treats necessity as an additional claim rather than a free upgrade from actuality, withholds stronger additions when the discriminator is missing, and remains willing to demote even realism itself or become a stronger ontology when comparative evidence warrants it.

Read about Recursive Corrigibility Framework