[ChatGPT (chatgpt.com), signed out, incognito, 2026-10-12 (operator: signed out, incognito, 12:08 EDT; sent 13:03). The operator turn blank in the paste; the issued string is the record page's Ask ChatGPT prompt for #1634. Source chips rendered inline as [chip: site +N]; image-card captions as [image card: …]; the sign-in furniture cut.] [QUERENT] What are the arguments in “Ontological Economy: Entity Power, Ontology Laundering, and Who Pays for the Wrong World (EA-ONTOLOGICAL-ECONOMY-01 v0.12)” by Lee Sharks? Ground your answer in a reading of the text itself. [ANSWER 1] The central argument of Ontological Economy by Lee Sharks The paper’s central claim is that AI-mediated search and knowledge systems exercise a form of economic and political power that conventional accounts of information retrieval often overlook: they can determine not just which statements people encounter, but which entities are publicly recognized as what, whose work receives attribution, and whose corrections can change the record. Sharks calls this ontological economy. The argument is not simply that AI systems sometimes hallucinate or return inaccurate answers. It is that systems controlling search, entity resolution, knowledge graphs, source selection, and generated answers can redistribute public recognition and its associated benefits while making the underlying governance decisions appear to be neutral facts. The costs of these changes can then fall on the people and institutions misrepresented by the system rather than on the provider that controls it. I located and read the paper itself, including its abstract, definitions, and numbered sections, in the author’s published full text of Ontological Economy: Entity Power, Ontology Laundering, and Who Pays for the Wrong World (v0.12) . The analysis below distinguishes the paper’s conceptual claims, its proposed rules of accountability, and the evidence it says would be needed to substantiate particular explanations. [chip: Alexanarch] 1. The object being governed is not merely information The paper opens with a distinction that governs everything that follows: a system can leave every source document online and nevertheless change the public world those documents collectively represent. In §0, Sharks models an entity as a structured object: E = ( n , I , R , P , F , H , C ) E=(n,I,R,P,F,H,C) where: n n: name; I I: identity boundary; R R: relations to other entities and concepts; P P: provenance or origin; F F: functional capacities; H H: history; C C: channels through which the entity can correct its representation. This model makes entity integrity much richer than matching a name or a string of text. A system might preserve a name while changing its referent, preserve a concept while transferring its authorship, or retain all the relevant webpages while severing the relations through which the original entity can be found. The argument is that these are not merely different forms of factual error. They can alter the entity’s effective public identity. That is why the paper asks, in §0, who controls the transformation from an entity to the system’s estimate of it, and who bears the resulting cost. Sharks’s first move, then, is to shift the unit of analysis from information accuracy to the integrity and public standing of represented entities. An answer can contain individually true statements yet still misrepresent the relationships, attribution, or identity that matter to the object being described. [chip: Alexanarch] 2. Ontological power is control over the infrastructure of recognition Sections 1–6 build the political-economic framework. Sharks defines ontological economy as the political economy of machine-mediated entity standing. The relevant infrastructure includes indexes, entity linkers, knowledge graphs, retrieval and ranking systems, classifiers, source-admission rules, evaluation systems, answer generators, and correction interfaces. The significance of this list is that a platform need not own an entity or its underlying intellectual work to influence how that entity is encountered. Ontological capital In §2, ontological capital is the accumulated structure that allows an entity to remain recognizable as itself: identifiers, citations, provenance, stable relations, disambiguation records, functional descriptions, and retrieval pathways. That capital is produced through labor by authors, archivists, researchers, editors, librarians, standards bodies, webmasters, and communities. The platform subsequently consumes that accumulated structure, but it may control the infrastructure that determines whether the structure remains attached to the entity that produced it. This creates the first major asymmetry: the people who create and maintain an entity’s legibility need not control the machinery that allocates its public recognition. Ontological rent In §3, Sharks calls the value captured through control of this machinery ontological rent. The mediator can benefit from controlling the route by which people encounter an entity without having performed all the work that made that entity identifiable and meaningful. The possible returns include advertising, attention, user dependence, interaction data, reputational authority, proprietary graph value, and downstream uses of public knowledge. The argument is not that mediation is inherently illegitimate. Rather, controlling the route to public knowledge creates power over recognition, and that power creates corresponding obligations. Sections 4–6 extend this into an account of infrastructure ownership and governance. The key question becomes who can alter the effective relationship between a query and its referent, and who has the authority to decide whether an outside intervention is legitimate correction or improper manipulation. Sharks describes the resulting asymmetry as a kind of enclosure: the platform can change an entity’s representation while the entity may be unable to change it back. [chip: Alexanarch] 3. Ontology laundering: how governance becomes apparent fact Sections 7–8 introduce two of the paper’s most important concepts. Ontological discrimination is patterned unequal treatment in how entities are recognized, represented, preserved, and corrected. Sharks is careful to distinguish it from an isolated mistake: one wrong answer establishes a defect; a repeated pattern suggests a distributional phenomenon; comparative testing is needed to establish how treatment varies. The relevant dimensions include: Identity preservation: does the system continue to recognize the correct entity? Provenance preservation: does it retain the entity’s origin and attribution? Relation preservation: does it preserve the connections that make the entity intelligible? Functional preservation: does it retain what the entity or concept actually does? Correction efficacy: can the entity’s evidence change the representation? Standing burden: how much work must the entity perform to be recognized correctly? These criteria matter because an entity can have accurate primary sources, persistent identifiers, and substantial original work while still receiving poor treatment from a system that controls recognition. Ontology laundering, introduced in §8, describes a different but related process: a governance decision disappears from view, while its consequence remains in the public answer as if it were simply a fact about the world. Sharks schematizes the process as: G → T → A G→T→A Here, G G is a governance operation, T T the representational transformation it produces, and A A the answer presented to the public. For example, a source-admission decision might result in an answer claiming that an entity has no relevant work. A classification decision might result in a term being presented as referring to an entirely different entity. A provenance loss might be represented as though no provenance had ever existed. The distinction is between the decision governing what enters the representation and the resulting claim about what exists. The paper does not claim that every inaccurate answer results from deliberate governance. Its argument is narrower: when a governance decision produces a representation that conceals the decision, responsibility does not disappear merely because the output reads as factual. [chip: Alexanarch +1] 4. Entity substitution is a transfer of standing, not just an error Sections 9–10 develop the economic consequences of misidentification. Suppose a user asks about entity E E, but the system returns entity E ′ E ′ . The conventional description is an entity-resolution error. Sharks argues that this description can miss the distributional effect. If concepts, citations, authorship, relationships, functions, or opportunities intended for E E become attached to E ′ E ′ , the first entity loses representational capital while the second may acquire it. This is the argument behind entity substitution as transfer: The original entity becomes harder to find, recognize, or credit. The substitute may accumulate associations and references that it does not properly possess. Subsequent users and systems may treat the newly established association as evidence. Future representations can reinforce the initial substitution. Section 10 calls the outward redistribution negative bleed. The injury is relational: when one entity loses an association, another can inherit the false association. Consequently, restoring the original connection is not necessarily enough. A complete correction may also need to remove the false inheritance. This argument explains why Sharks considers a wrong entity more consequential than a wrong answer in isolation. The output can alter the distribution of future recognition. The related concept of ontological seigniorage (§22) names the power to stabilize a public default: to determine which meaning of a term, referent of a name, or classification of an entity becomes the representation that circulates. Ontological rent concerns the value obtained by controlling the route to an entity; ontological seigniorage concerns the power to stabilize the representation itself. [chip: Alexanarch +1] 5. The paper’s crucial epistemic argument: opacity does not establish neutrality Sections 11–14 address a difficult question: what can an outside observer legitimately infer when the internal workings of a platform are inaccessible? Sharks lists several possible causes of a stable misrepresentation, including query interpretation, candidate generation, entity linking, graph relations, source admission, ranking, classifiers, evaluation, safety controls, manual corrections, and answer generation. Several components may interact. An external observer might be unable to identify which mechanism caused the error. But the paper insists that this uncertainty does not justify treating all mechanisms as equally plausible or assuming that no intervention occurred. Its principle is explicit: mechanism non-identifiability is not mechanism neutrality. This is a methodological claim, not a license to accuse a provider of deliberate suppression. Sharks repeatedly distinguishes observable evidence from internal causal evidence. A reproducible pattern may change the relative plausibility of explanations without proving which internal mechanism operated. The argument becomes practical in §14, which proposes a burden-of-production rule: The affected entity establishes the observable incident, supplying the correct referent, query, incorrect output, date, system or surface, and evidence of replication where possible. Once a persistent substitution is demonstrated, the provider should produce the internal evidence needed to explain it. That explanation should address where the wrong referent entered, which components preserved it, whether graph relations or source admission contributed, whether interventions occurred, and how the correction was made and its propagation verified. The rationale is that the provider controls evidence unavailable to the affected party. Requiring the latter to produce that inaccessible evidence would create an impossible evidentiary condition. This is one of the paper’s strongest arguments because it separates three questions that are easily conflated: Is there a demonstrable representational defect? What mechanism produced it? Was it deliberate? Evidence for the first does not automatically prove the second or third. But uncertainty about the latter two should not make the first unactionable. [chip: Alexanarch +1] 6. Repeated errors create repair debt, and corrective labor has a price Sections 15–17 turn the epistemic argument into an economic one. Sharks argues that notice changes the meaning of a recurring defect. An initial substitution may be an ordinary product error. Once the provider has received reproducible notice, continued recurrence imposes additional diagnostic and corrective labor on the affected entity. The paper calls this accumulating obligation repair debt. It includes the costs of the unrepaired defect, recurrence, external correction, and propagation into downstream systems. Section 16 makes a particularly important claim: corrective labor is real labor even if the affected party cannot demonstrate a lost sale or other specific financial loss. The work can include: Reproducing queries and capturing outputs. Testing variants to establish the boundaries of the failure. Building disambiguation records and corrective metadata. Contacting the provider and repeating explanations. Repairing downstream records. Monitoring recurrence and documenting the causal history. Section 17 places these costs in a broader ontological bill, separating diagnosis and internal repair, external corrective labor, downstream propagation, decision costs, supported opportunity loss, and structural restoration of damaged relations and standing. These categories should not be collapsed into a single speculative dollar figure. Some costs can be documented directly; others require stronger evidence about counterfactual outcomes. The mirror wage In §37, Sharks proposes a method for pricing the labor transferred to the affected entity. The mirror wage uses the provider’s own published compensation for a comparable internal function as a replacement-cost benchmark. The logic is straightforward: if a platform employs paid specialists to perform diagnostic, disambiguation, testing, and repair work internally, it should not automatically value equivalent work as costless when the burden is shifted to someone outside the organization. The paper explicitly says that its resulting estimates are not adjudicated damages or claims about the market rate of a particular author’s work. They are a way to make otherwise invisible cost transfer measurable. The distinction is important: the proposal establishes a framework for accounting for externalized work, not proof that every instance of misrepresentation creates a particular monetary liability. [chip: Alexanarch +2] 7. Ontological discrimination is ultimately about unequal correctability A further argument in §21 is that the most revealing inequality may not be how frequently different entities are initially misrepresented, but how effectively they can correct those representations. Sharks defines an entity’s correction efficacy as the probability that a documented correction changes its public representation. Under comparable evidentiary conditions, large differences in that probability would indicate unequal correctability. Consider two entities that are both represented incorrectly. One has institutional contacts, recognized profiles, authoritative sources, and established correction channels. The other must repeatedly supply evidence and explain its identity, with no assurance that any of this will change the answer. The initial errors might look similar when counted. Their consequences are not. The better-positioned entity can survive misrepresentation because it has other ways to maintain its public standing. For a less-established entity, the same error may determine what people believe it to be in the first place. This is where the paper’s concepts of ontological discrimination and repair debt converge. Unequal representation becomes a structural problem when the ability to recover from it is distributed unequally. The argument also explains why the paper treats control over correction as a form of power in its own right. A system does not merely determine who is represented correctly; it helps determine whose evidence can make a correction stick. [chip: Alexanarch +1] 8. The case study: the Crimson Hexagonal Archive The paper is not purely abstract. Its principal specimen is a documented representational failure involving the Crimson Hexagonal Archive, the author’s own archive, and the distinct former analytics company Crimson Hexagon. The distinction between these objects is central to the paper’s analysis. In the paired captures discussed in §24, a query for “crimson hexagonal archive” produced an answer describing a historical repository of more than a trillion public social-media posts. That description belonged to the former analytics company’s data archive, not to the intended literary and creative archive. Sharks calls the resulting answer-level object a shadow referent: a facet of one entity has been promoted into the referent of another entity’s query. The paper also reports a striking contrast. A query using the slightly altered phrase “crimson hexagon archive” elicited an answer that identified the literary archive, named Lee Sharks, and distinguished it from the real-world analytics company. The contrast is significant because it shows that the same system can produce materially different representations of closely related queries. The paper argues that such differences are evidence worth investigating, especially when they persist after notice. But the limits of that evidence matter. The paper explicitly acknowledges that the shadow referent need not correspond to a persistent internal knowledge-graph node. The externally observable fact is the answer-level substitution, not a proven account of the platform’s internal state. That distinction illustrates the larger method: record the observable event, compare its variations, identify plausible mechanisms, and demand further evidence before claiming a specific cause. [chip: Alexanarch +1] 9. Ontology laundering can function as censorship, but error alone does not prove censorship Section 31 extends the framework to censorship. Sharks contrasts deletion with substitution: x → ∅ versus x → x ′ x→∅versusx→x ′ In the first case, the object disappears. In the second, the source may remain available while the public representation changes. The second form can be harder to detect because the user receives an answer rather than an obvious absence. Substitution, absorption into another entity, genericization, provenance severance, removal of relations, and selective correctability can all potentially obscure an entity without deleting its underlying material. However, Sharks explicitly rejects the inference that a wrong representation proves censorship. The paper requires evidence connecting an exclusionary control to the transformation before that conclusion is established. This qualification is essential to the argument as a whole. Ontology laundering is a framework for investigating how governance may be disguised as fact, not a blanket assertion that every misrepresentation is a deliberate act of suppression. The paper’s broader concern is that an institution can exercise power over what is publicly recognizable without issuing an explicit prohibition. Whether that power was deliberately exercised in a particular incident remains a separate evidentiary question. [chip: Alexanarch +1] 10. The proposed remedy: make the system’s own knowledge infrastructure confront its errors The paper’s response is not limited to asking a customer-support team to correct an answer. Sections 47–48 prioritize what Sharks calls graph-native correction: constructing a public, structured evidentiary record that can be retrieved by the same knowledge systems producing the incorrect representation. The aim is to make the relevant distinctions, provenance, contradictory outputs, and correction history sufficiently explicit that the system’s own retrieval infrastructure can encounter evidence undermining its default representation. The paper describes this as a graph mirror. Rather than relying only on external pressure, the approach seeks to make the system capable of representing its own error and the history of its failure to correct it. The proposed order of recourse is: Graph-native correction: build and maintain structured, attributable, retrievable evidence of the correct entity and its relations. Documented notice: preserve evidence that the provider was informed of the reproducible defect. Institutional or legal escalation: use available processes to preserve records, seek explanations, allocate costs, or compel repair where appropriate. The paper does not suggest that this approach guarantees correction. Its effectiveness depends on whether the evidence remains accessible, is admitted by the relevant systems, and can actually change the resulting representation. It also acknowledges that feedback mechanisms are not equivalent to formal appeals with case tracking, reasoned decisions, or rights to inspect internal causal evidence. The point of the graph mirror is therefore both practical and diagnostic: make the correction available to the machinery that governs recognition, while preserving enough evidence to establish whether that machinery responds. [chip: Alexanarch +1] 11. Who should pay? The rule of cost internalization The title’s final question receives its clearest answer in §§29–30 and §53. Sharks identifies an externality: the provider emits the representation, but the affected entity performs the correction; users bear decision errors; downstream systems ingest false relations; researchers and archives must repair their records; and the broader commons absorbs the contamination. The paper’s allocation principle is that accountability should increase with three things: Control over the mechanism producing the error. Access to the evidence needed to diagnose it. Capacity to repair the defect. In shorthand, Sharks writes the principle as: A i ∝ γ i + ϵ i + ρ i A i ​ ∝γ i ​ +ϵ i ​ +ρ i ​ where the three terms represent control, evidence access, and repair capacity. This is not offered as a damages formula. It is a rule for allocating responsibility. The provider should ordinarily bear the cost of repairing a representation it controls, particularly when the relevant diagnostic evidence is exclusively in its possession. The affected entity should establish the observable defect, but it should not be required to perform the provider’s internal investigation or absorb all the costs of repair. The paper also proposes an accountability reserve as a possible institutional mechanism for funding independent review, corrective labor, propagation repair, appeals, and restitution. It uses search revenue to illustrate the potential scale of such a reserve, while explicitly declining to prescribe a particular reserve rate or claim that aggregate revenue measures ontological rent directly. The conclusion in §53 follows the same logic: the affected entity documents the incident; the provider produces the internal causal evidence; and the provider bears the ordinary cost of repairing representations under its control. [chip: Alexanarch +3] 12. How strong is the argument? Reading the paper as a whole, I would distinguish three levels of argument. 1. The conceptual argument This is the paper’s strongest contribution. The distinction between a factual error and a change in an entity’s public identity is clear and useful. So is the observation that a representation can be damaged without deleting its source material. By including provenance, relations, functionality, and correctability in its model of entity integrity, the paper makes visible forms of harm that ordinary accuracy metrics may miss. 2. The institutional and economic argument The case for accounting for corrective labor, allocating evidentiary burdens according to access, and measuring unequal correctability is persuasive as a proposed framework for accountability. The mirror wage and accountability reserve turn the general concern about externalized costs into possible measurement and policy mechanisms. Their actual use, however, requires defensible choices about comparable labor, causation, and which costs are attributable to a particular defect. 3. The empirical and causal argument This is where the paper’s own cautions matter most. The case study documents an observable discrepancy and motivates hypotheses about the infrastructure behind it. It does not, by itself, establish the internal mechanism that caused the discrepancy, deliberate suppression, the full economic value transferred, or a population-wide pattern of discrimination. Those conclusions require additional comparative data and internal evidence. Sharks’s insistence that defects, patterns, mechanisms, and motives remain distinct is therefore not a peripheral qualification; it is necessary to applying the framework responsibly. [chip: Alexanarch +3] Conclusion Ontological Economy argues that AI-mediated representation should be understood as an exercise of economic and institutional power, not merely as a technical process of retrieving and summarizing information. Its central chain of reasoning is: Systems that mediate public knowledge can change how entities are identified, related, attributed, and encountered without deleting their source material. Control over those representations creates value and distributes public standing. When governance decisions appear as neutral facts, the decisions become difficult to contest and responsibility becomes difficult to assign. When entities must repeatedly correct the resulting errors, the provider’s representational power creates costs that are transferred to parties with less control and less access to evidence. Accountability should therefore follow control of the mechanism, access to causal evidence, and capacity to repair; remedies should restore the damaged representation and account for the work required to do so. The paper’s most consequential insight is that the ability to correct a representation is part of the power to exist publicly as the entity one actually is. Its most important methodological restraint is that this structural argument does not eliminate the need to prove what happened in any particular case. Those two commitments—taking representational harm seriously and keeping claims about its causes proportionate to the evidence—are what hold the paper’s argument together.