{"id":"be616d08-ffce-445c-8237-1ac18f00ac93","arxiv_id":"2505.14940","paper_version":1,"verdict":"REJECT","confidence":"HIGH","novelty_score":4.0,"correctness_risk":"high","formal_verification":"none","parameter_count":5,"one_line_summary":"A position paper argues that existing formal ontologies are not formal in Husserl's sense and proposes vector spaces as a truly formal ontology to express foundational ontology concepts.","lead":"A philosophy-of-ontology paper argues that existing formal ontologies are not truly formal in Husserl's sense and proposes that vector spaces, the mathematical framework behind many machine learning systems, should serve as the formal ontology. The idea is that representing concepts and objects as vectors could create a shared ontology for humans and AI, but the paper offers no experiments and only illustrative examples.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Vector-space axioms alone cannot express existence: 'V_reality' is an extra subset that is not closed under vector operations, so the claimed a priori formalism is not actually a vector space and the central representation claim is unproven.","rationale":"The reader's weakest_assumption identifies both the finite-dimensional representation claim and the V_reality closure problem. I agree that the closure/existence issue is the more decisive internal inconsistency, because it affects the paper's titular notion of 'to be or not to be' and the claim that vector-space axioms alone constitute a formal ontology. The finite-dimensional assumption is also load-bearing, but it is at least acknowledged as an asymptotic approximation in Sections 5.2.2 and 8; the V_reality problem is not acknowledged and cannot be patched without adding structure outside the vector space. My concrete test would settle whether the existence predicate is derivable from the axioms; if it is not, the central claim of a priori formal status fails. The paper has genuine value as a discussion piece, and the author's engagement with Gardenfors and Guizzardi is useful, but the formal framework as specified is not coherent enough to support the abstract's 'concrete proof' claim. Hence the reader's REJECT verdict stands unchanged.","tokens_in":15772,"tokens_out":2890,"duration_ms":28506,"concrete_test":"Formalize Section 5 in a proof assistant such as Lean with only the eight vector-space axioms and a declared basis B_D. Attempt to derive the existence statement 'blue rectangle exists' (i.e., [4,0,0,255] ∈ V_reality) from those axioms alone. The derivation will require adding an uninterpreted membership predicate or an extra sparse-subset axiom; if so, the existence predicate is not a consequence of the vector-space formalism. As a second check, take V = R^2 with B = {x1, x2} and V_reality = {(0,0), (1,0)}; the linear automorphism swapping x1 and x2 preserves all vector-space axioms but maps V_reality to {(0,0), (0,1)}, so membership is not invariant under vector-space structure and must be an externally imposed notion.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is that a class of formal ontology using the axioms of vector spaces can express most foundational-ontology concepts, including existence, as a priori occurrences independent of perception. The load-bearing step is in Section 5.2.2: existence is defined as membership in a sparse subset V_reality of the full vector space, via the expression v ∈ V_reality. But the eight vector-space axioms in Section 5 define only a vector space V = span{x1,...,xn}; by closure, every linear combination of basis vectors is already a member of V. The axioms alone cannot distinguish 'actually exists' from 'theoretically possible', because both are elements of the same vector space. The paper explicitly says the vector space is 'continuous or densely populated' while the ontology is interested in a 'sparse population' representing reality. That sparse population is not a vector subspace: it is not closed under addition or scalar multiplication, and the paper gives no axioms for which vectors belong to V_reality. Therefore the formal framework is not a vector space once existence is introduced; it is a vector space plus an unexplained external predicate. The claim that existence is 'the only information we can directly extract from a vector ontology' is thus not justified by the vector-space axioms. This is not merely a missing detail: it undercuts the assertion that vector ontologies are truly formal in the Husserlian sense, because the central ontological notion of being is smuggled in as extra, non-vector-space structure. Section 11 compounds the problem by listing as future empirical tests hypotheses that the abstract presents as demonstrated, including the very representability of domains as vector ontologies.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper argues that previous \"formal ontologies\" (Husserl, Guizzardi, BFO, GFO, DOLCE) are not formal in Husserl's sense because they are built by empirical categorization, and proposes instead that a vector space, called a vector ontology (Vont), can serve as a truly formal ontology. It maps central foundational-ontology concepts to vector-space notions: existence as membership in a sparse subset V_reality, endurants as continuous functions of existence, mereology as convex regions, causation as linear dependence, correlation as probabilistic existence, and similarity as Minkowski or reconstruction distance. The paper claims this shows vector ontologies are a priori, independent of perception, and already used by AI systems and humans.","tokens_in":16100,"tokens_out":6703,"duration_ms":61154,"significance":"The proposal is ambitious and potentially valuable: if rigorously established, a vector-space foundation for ontology could offer a principled, interpretable bridge between human conceptualization and machine-learned representations, and would provide a fresh perspective on the long-standing distinction between formal and foundational ontology. The paper deserves credit for engaging with the philosophical literature, for using concrete examples, and for explicitly listing its limitations and future empirical tests in Section 11. However, the central expressivity claim is not demonstrated: the bookkeeping is done by a handful of selected examples, the existence predicate is left formally undefined, and key mathematical descriptions contain errors. The current manuscript is a programmatic essay rather than a formal contribution, and its load-bearing claims are not supported as written.","major_comments":[{"comment":"Existence is the central notion of the paper, but it is not defined by the vector-space axioms. Section 5.2.1 defines VontD = span{x1,...,xn}, and the axioms in Section 5 guarantee closure under addition and scalar multiplication, so every linear combination of basis vectors is already an element of VontD. Section 5.2.2 then defines existence as membership in a sparse subset V_reality (via \"v ∈ V_domain reality\"), but this subset is not a vector subspace, is not closed under the vector operations, and is given no axioms or rules specifying which vectors belong to it. The paper explicitly notes that the full vector space is \"continuous or densely populated\" while the ontology cares about a sparse population, yet no formal characterization of this sparse population is provided. Consequently, existence is an external, unexplained predicate, and the claim that vector ontologies are a priori formal frameworks from which existence can be \"deducted\" is not supported; the formalism is a vector space plus an unformalized additional structure.","section":"Section 5.2.2 and Section 5.2.1"},{"comment":"The paper asserts that \"any object of interest can be fully described as a vector in a vector space with a finite number of basis vectors representing quality dimensions.\" This is the load-bearing assumption of the entire framework, but it is merely stated, not argued for. The paper offers no reason why all ontological concepts—including normative, modal, intentional, or qualitative phenomena—should admit finite-dimensional vector-space representation. The analogy to Fourier transforms in the same section is suggestive, but it is not a derivation. Since this assumption is exactly what must be established for the expressivity claim, its status as an unproven premise is a major gap.","section":"Section 5.2.2, \"key assumption\""},{"comment":"The abstract characterizes the showcase as \"concrete proof\" that a class of formal ontology based on vector-space axioms can express most foundational-ontology concepts, and the conclusion states that the author \"clearly demonstrated\" this. Yet no formal proofs are given. The mappings in Section 5.2.3 are stipulated: endurants are defined as continuous functions of existence, causation as linear dependence, correlation as probabilistic existence, and parthood as convex regions. These are redefinitions of ontological concepts in vector-space terms, not consequences derived from the axioms. Section 11 itself says \"since this Paper is theoretical,\" and lists future empirical tests. The discrepancy between the abstract's \"concrete proof\" and the actual content is not merely a matter of wording; it obscures the epistemic status of the central claim, which is at best a conjecture supported by examples.","section":"Abstract and Section 11"},{"comment":"Several mathematical statements are incorrect or inconsistent with the formal framework. In Section 5.2.1, the paper suggests using N (natural numbers) for discrete properties and gives an example vector [4,0,0,255] with integer components, but a vector space is defined only over a field, and N is not a field (it lacks additive inverses). The shelf example uses R, but the colored-shapes example uses integer coordinates without specifying a field. In Section 5.2.3, \"Functions of Existence\" are introduced as linear maps, yet the worked example fe(t) is a piecewise function that is not linear; the term \"microcontinuity\" is used without a formal definition; and the \"Convex subspaces\" definition actually defines a convex subset, not a subspace, since it is not closed under addition and scalar multiplication. These inaccuracies undermine the paper's stated goal of providing a rigorous, formal framework.","section":"Section 5.2.1 and Section 5.2.3"}],"minor_comments":[{"comment":"There are multiple typographical and formatting errors, including \"V ector\" in the Section 5.1 heading, \"Existance\" in the mathematical comment, \"Preconstrcution\" in the reconstruction path formula, and \"sucessfully\" in Section 8.","section":"Throughout"},{"comment":"The text refers to \"section 4.1.3\" when discussing functions of existence, but the relevant discussion appears in Section 5.2.3.","section":"Section 9.3"},{"comment":"The notation \"Blue Rectangle exists = ⇒ [4, 0, 0, 255] ∈ Vcolored-shapes\" is malformed; the symbol \"= ⇒\" is not standard, and the distinction among Vcolored-shapes, Vdomain, and V_reality is used inconsistently throughout the section.","section":"Section 5.2.2"},{"comment":"The name \"Gärdenfors\" is consistently misspelled as \"Gardenfors\"; please correct all occurrences.","section":"References"}],"recommendation":"reject","confidential_remarks":"This is fundamentally a programmatic essay rather than a formal research paper. The author admits in Section 11 that the work is theoretical and requires future empirical testing, yet the abstract and conclusion claim \"concrete proof\" and \"clearly demonstrated.\" The central formal gap—the undefined existence subset V_reality—cannot be repaired by minor edits; it requires rethinking what the framework actually is and what it can claim. The paper may be suitable for a venue that publishes speculative, interdisciplinary position pieces, but for this journal, the load-bearing technical and philosophical issues are too severe. I recommend rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nRead the Rothenfusser paper. It is a clear, ambitious essay arguing that vector spaces can serve as a truly formal ontology in Husserl's sense, and that current foundational ontologies (BFO, DOLCE, GFO) are not formal but foundational. The paper does a real service by articulating that distinction and by giving a readable mapping of endurants, mereology, causation, correlation, similarity, and identity into vector-space terms. It also cites Gardenfors honestly, acknowledges the earlier conceptual-spaces work, and takes Guizzardi's identity criticism seriously. The examples (color spaces, shelves, apples) are helpful.\n\nThe soft spot is load-bearing. Existence, the key ontological notion, is introduced as membership in a subset V_reality. But the vector-space axioms in Section 5 define the whole space by closure under addition and scalar multiplication; every linear combination of basis vectors is already in the space. The axioms alone cannot distinguish 'exists' from 'merely possible.' So the formal framework is not a vector space once existence is introduced; it is a vector space plus an unexplained, unaxiomatized selection predicate. The paper's own phrase 'sparse population of the vector space' makes this clear, but no rules are given for which vectors populate reality. That undercuts the claim that everything is derived a priori from the axioms.\n\nSecond, the expressivity claim is demonstrated only by selected examples, not derived. The abstract says 'concrete proof'; Section 11 says the paper is theoretical and future empirical tests are needed. That is a real inconsistency.\n\nThird, the 'total absence of content' claim is weakened by the paper's own admission that the choice of basis vectors is domain-specific and perceptually biased. The formal structure itself may be content-free, but the moment you pick a basis for 'shelves' or 'colored shapes,' you are importing empirical content. The paper tries to rebut this in Section 9.1 by saying the axioms precede the basis, but that only works if existence is also axiomatically defined, which it is not.\n\nWhat the paper does well: it is an honest discussion piece. It predicts criticism and rebuts it, it lists concrete hypotheses for future work, and it engages with the relevant literature. The Husserlian framing is interesting even if the execution is incomplete.\n\nI would not cite this in my own work yet, and I would bring it to a reading group as a discussion prompt. As a submission, I would not trust it as a formal result, but it deserves a serious referee rather than a desk reject. The author should be told to either formalize the existence predicate or soften the claims from 'proof' to 'proposal.' With that revision, the paper could be a useful contribution. My honest verdict: reject in current form, but encourage resubmission.","headline":"The paper is a readable philosophical proposal, but the existence predicate is extra structure that breaks the vector-space formalism, so the central formal claim fails on its own terms.","tokens_in":16635,"tokens_out":3068,"would_cite":false,"duration_ms":27907,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Vector spaces, not categories, make ontology truly formal","keywords":["vector ontology","formal ontology","foundational ontology","vector spaces","a priori","functions of existence","mereology","AI interpretability"],"falsifier":"Exhibit a single object or concept that cannot be encoded by any finite set of quality dimensions with coordinate values—a normative statement, a universal generalization, or a self-referential thought—and show that no enlargement of the basis repairs the gap; if such a case exists, the finite-basis assumption fails and the expressivity claim collapses.","tokens_in":15537,"feed_emoji":"📐","tokens_out":14241,"duration_ms":109396,"temperature":0.7,"pith_summary":"This paper argues that the ontologies usually called formal—categorical systems built by abstracting from human perception—do not meet the strict criterion of formality set out in the Logical Investigations: validity independent of perception and total absence of content. The author proposes instead that a formal ontology should be a mathematical structure chosen before any ontological content is added, and offers vector spaces as the concrete candidate. A vector ontology defines a domain by a finite set of basis vectors that name quality dimensions; existence is then the membership of a point in the sparse set of real things, and the classic concepts of foundational ontology are recovered as geometric or algebraic features of that space. If the proposal holds, ontology becomes an a priori, content-free framework into which human and machine worldviews can both be poured, making machine representations interpretable and human knowledge machine-readable. The stakes are practical as well as philosophical: many AI systems already compute in high-dimensional vector spaces, so an ontology with vector form could serve as a common interface between human and machine understanding.","feed_headline":"Vector spaces, not categories, make ontology truly formal","feed_subtitle":"Existence, part, cause, and similarity all fall out of one a priori vector structure humans and machines can share.","key_machinery":"The load-bearing object is the vector ontology $V_{\\mathrm{ont}}$, defined as the span of a finite set of basis vectors $\\{x_1,\\dots,x_n\\}$ over a field, where each basis vector names a quality dimension and the field supplies possible values. Existence is encoded by the membership test $v \\in V_{\\mathrm{reality}}$ for the sparse subset of the space corresponding to actual things, and the main analytical tool is the function of existence $f_e$—a (multi)linear map from the ontology into a binary existence space—which compresses scattered vectors into concepts. Around these sit the derived mechanisms: convex regions model parthood, continuous functions in the time basis model endurance, linear dependence models causation, probabilistic functions model correlation, and reconstruction paths through a small set of interpretable dimensions model similarity and metaphor. The entire argument depends on the choice of basis being interpretable, so that the vector space is navigable rather than a black-box feature space.","core_discovery":"On the paper's own terms, the central claim is that the axioms of a vector space constitute a formal ontology in the strict sense, and that nearly all categories of existing foundational ontologies reappear as structural facts about such a space. The basis vectors are universals—quality dimensions—and the field supplies the quale, so an object is a vector; to be is to belong to the sparse subset of the space that records what actually exists, while possible existence is modal membership. Endurants become functions of existence that are continuous in the time dimension, perdurants are the complement; mereological parthood becomes containment of convex regions; causation is linear dependence among vectors; correlation is a learned probabilistic function of existence; similarity is distance; and identity across change is a function of existence that stays continuous or constant in the relevant dimensions. The paper further claims that this structure is not merely convenient: it is already the implicit ontology of neural-network computations, and there is cited empirical evidence that human perception operates in the same geometric manner.","pith_inferences":["A natural next test is to build a bounded vector ontology, train a network on that domain, and check whether the network's learned functions of existence align with the human-selected basis; the author lists this as future work, but it is already well-specified enough to run.","The finite-basis assumption suggests an approximation agenda: if domains behave like Fourier spectra, a small dominant set of quality dimensions should account for most of the structure, and identifying that set for real domains becomes an empirical research program.","The identification of causation with strict linear dependence predicts that genuinely causal relations are exact algebraic dependencies, which would separate deterministic causation from probabilistic correlation in a sharper way than most current accounts.","The formal status of the framework is limited to the bare vector axioms; fixing a domain's basis is a perceptual choice, so the content-free guarantee does not extend to any populated ontology."],"forward_implications":["If vector ontologies are genuinely formal, then existing categorical systems built by recursive abstraction should be reclassified as foundational rather than formal, and their categories become derived phenomena rather than primitive axioms.","An ontology populated inside a vector space is internally consistent by construction, because the axioms themselves cannot generate contradiction; errors surface only as mismatches with observed reality.","Because neural networks compute in high-dimensional vector spaces, a vector ontology with interpretable basis vectors would make it possible to extract or align a network's learned ontology, turning the interpretability problem into a design requirement.","Interoperability between information artifacts becomes a property of form: any two systems that populate the same vector space can share and navigate each other's content before any domain-specific vocabulary is fixed.","Search and analogical reasoning reduce to navigation along a small set of quality dimensions, since reconstruction distance—not raw coordinate distance—is the proposed measure of relatedness."],"supporting_citations":[{"why":"It supplies the strict definition of formal ontology as a priori and entirely free of content, the criterion against which all later systems are judged.","marker":"Husserl, 2001b"},{"why":"It supplies the account of logic as an a priori formal science that the paper transfers to ontology.","marker":"Husserl, 2001a"},{"why":"It supplies the foundational-ontology vocabulary—universals, qualia, endurants, perdurants, parthood—that the vector ontology re-expresses.","marker":"Guizzardi, 2005"},{"why":"It offers a leading categorical system whose recursive-abstraction method the paper classifies as foundational rather than formal.","marker":"Arp et al., 2015"},{"why":"It provides a second categorical ontology whose endurant and perdurant distinctions are reproduced as vector-space phenomena.","marker":"Herre et al., 2006"},{"why":"It gives empirical evidence for geometric conceptual spaces in human perception and the transfer-learning idea the paper adapts.","marker":"Gardenfors, 2004"},{"why":"It raises the identity-and-types objection to vector-style spaces that the paper answers with functions of existence.","marker":"Guizzardi, 2015"},{"why":"It argues that vector representations define only clusters, prompting the paper's answer that types are classes of functions.","marker":"Lopes, 2023"},{"why":"It defines the FAIR interoperability principles that formal-by-design vector ontologies are claimed to satisfy.","marker":"Wilkinson et al., 2016"}],"fun_headline_variants":["True formal ontology is just vector spaces","Husserl's formal ontology is linear algebra","Vector ontologies: the only truly formal ones","Formal ontology reduces to vector space axioms"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The entire framework rests on the claim that any object of interest can be fully described as a vector in a vector space with a finite number of basis vectors representing quality dimensions; if some objects or concepts—norms, modalities, intentional states—resist finite-dimensional vector representation, the framework cannot express them, and the central claim collapses.","fun_headline_variants_meta":{"raw":{"variants":["True formal ontology is just vector spaces","Husserl's formal ontology is linear algebra","Vector ontologies: the only truly formal ones","Formal ontology reduces to vector space axioms"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000372,"raw_usage":{"total_tokens":2043,"prompt_tokens":1056,"completion_tokens":987,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":672,"completion_tokens_details":{"reasoning_tokens":931}},"tokens_in":672,"tokens_out":987,"duration_ms":7853,"temperature":1.0,"reasoning_tokens":931,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T15:26:33.645959+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Exhibit a single object or concept that cannot be encoded by any finite set of quality dimensions with coordinate values—a normative statement, a universal generalization, or a self-referential thought—and show that no enlargement of the basis repairs the gap; if such a case exists, the finite-basis assumption fails and the expressivity claim collapses.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"It offers a leading categorical system whose recursive-abstraction method the paper classifies as foundational rather than formal."}],"review_version":1}