A search system can leave every source document online and still alter the public object those documents constitute. It can return one entity when asked for another, keep a method as a topic while losing what the method lets a reader do, keep a name while changing its referent, keep a concept while removing its provenance. Nothing need be deleted, and the public world has changed.
Ontological economy is the political economy of machine-mediated entity standing: the production, allocation, enclosure, circulation, depreciation, extraction and repair of what publicly exists as what. Its question is who controls the transformation of an entity into its representation, and who bears the cost when that transformation is wrong.
An entity acquires value inside mediated systems as it accumulates stable, machine-readable relations: identity, provenance, disambiguation, citations, identifiers, retrieval paths. That stock is ontological capital, and it is produced by labor — authors, archivists, editors, librarians, communities correcting false relations. The party who produced it may not control the infrastructure that decides whether it stays attached to them.
A mediator that controls the route by which entities are encountered can capture ontological rent without having performed the labor that produced them. A mediator with enough reach also stabilizes the unit: what a term means, which entity a name denotes, which provenance is remembered. That is ontological seigniorage. Rent comes from control of the route; seigniorage from power over the unit. Control over encounter creates obligations because it creates power over standing.
The affected entity sees the output. The provider sees, or can reconstruct, the causal path. The public may see only the answer, and when a query for one entity simply returns another, a user who does not already know the difference experiences no error at all. Mechanism invisibility plus event invisibility is invisibly invisible governance: a refusal can be challenged because it is known to have happened; a substitution arrives as reality.
Ontology laundering is the transformation by which a governance decision disappears and its consequence remains as fact: a source exclusion reappearing downstream as “the entity has no such work,” a trust rule as “there is no evidence,” a graph substitution as an identity. When governance is emitted as ontology, responsibility remains with the governor.
Ontological discrimination is patterned unequal treatment in the infrastructure of recognition — in who keeps their identity, provenance, relations and functions, and above all in whose correction is allowed to become real. One wrong answer is a defect. A repeated pattern is a candidate distributional phenomenon. Defect, pattern, mechanism and motive are recorded separately, and none is inferred from another.
When a query for one entity resolves to another, concepts, citations, authorship and opportunities meant for the first can accrue to the second. The first loses ontological capital while the second may gain it. A graph cannot misplace one edge without changing at least two nodes, so the injury is nonlocal, and a repair that restores the first without removing the false inheritance from the second leaves the ontology damaged.
Mechanism non-identifiability is not mechanism neutrality. An outside observer who cannot say which internal mechanism operated is not thereby entitled to treat them as equally likely, and still less to set the probability of intervention to zero. The provider holds the graph history, candidate sets, policy actions and intervention records; the affected entity holds the query, the output, the date and the replications. Provider-controlled opacity cannot function as a liability shield.
So the burden is split by access. The affected entity establishes the observable defect and its reproducibility. The provider then bears the burden of producing the internal account — not a presumption of malice, but a rule against using exclusive evidence to keep the causal field permanently empty. Notice converts repeated misrepresentation into accumulating repair debt. And the corrective labor forced onto the misrepresented party — reproducing queries, building disambiguation, repeating the same explanation across surfaces — is a transferred cost before it is a damages theory. The first economic discipline is to stop pretending that unpriced loss is zero.
Misrepresentation reduces encounters, which reduces citations and independent mentions, which lowers measured standing, which the mediator can then cite as the reason for the representation. That is validation foreclosure. A provider may not treat standing deficits partly produced by its own mediation as independent evidence against the represented entity. Identity precedes evaluation: a harsh evaluation of the correct entity remains evaluation; a precise evaluation of the wrong entity remains wrong.
The Crimson Hexagon is an ISBN-bearing literary project in print since 2014; Crimson Hexagonal Archive is its later, disambiguated archive-facing name. Both it and an analytics company founded in 2007 take the phrase from Borges. Asked for the archive's canonical name, the composition layer returned a shadow referent derived from the company, weighted toward regulatory scrutiny. Deleting two letters from hexagonal changed the answer's ontological mode: the literary archive, its author, and an explicit distinction from the company. The disambiguated name, which is itself corrective labor, is the address at which the substitution occurs.
A platform that identifies, summarizes, ranks and answers about entities at scale assumes practical custody of their public representation. Custody creates duties, not ownership: identity preservation, provenance preservation, correctability, traceable repair, and correction that is never itself held against the one who corrects.
The most native recourse for an ontological injury is ontological. Graph-native recourse builds durable, provenance-rich, machine-resolvable public relations until the evidence environment contains what is needed to correct, expose or contradict a false entity state. It is not spam, poisoning or manufactured consensus. Its governing constraint is that every added edge must be independently supportable. One supplies the graph with the distinctions it should already have been able to preserve.
Exact-name substitution that cannot be replicated, or that vanishes under controlled querying; variants that do not behave differently; comparable entities showing no asymmetry in correction; provider traces showing an ordinary transient failure; recurrence that stops after repair; provenance stripping shown not to have occurred. Specific claims of censorship, manual intervention, discrimination or loss each need their own evidence. The framework itself rests on one proposition: machine-mediated entity standing is a governed, value-bearing, labor-dependent infrastructure with distributive consequences. If that is false, it fails at the root.
Whoever has the power to make one entity publicly appear as another has entered the economy of ontology. The inability of outsiders to see the internal decision does not make the decision disappear from the causal world.