Machine Resolution

Machine Resolution #

Premise: Machine Resolution is a Machine Relations term for the observable result of asking an AI engine about a category, problem, or buying intent and recording whether and how it identifies a named entity. It describes the output that can be measured. It does not expose why the engine produced that output.

The term was coined by Jaxon Parrott as part of the Machine Relations framework. Under this definition, an entity can be present, absent, ambiguous, mentioned, cited, or recommended in a captured answer. Those are observations, not evidence of the engine's hidden source weights, confidence, or internal decision path.

The DNS Analogy #

The name borrows an analogy from DNS resolution: a human-readable request produces a machine-readable result. The analogy stops there. DNS follows a published lookup protocol; AI answer engines generally do not publish a complete causal account of how a particular brand or source was selected.

Premise: Machine Resolution measures the answer, not an assumed internal mechanism. A result can change across engines, model versions, query wording, dates, or account contexts. The protocol records those conditions instead of treating one answer as a universal property of the entity.

What Entity-Resolution Research Establishes #

Research on knowledge-graph-enhanced LLM entity disambiguation reports that structured knowledge such as entity types, class taxonomies, and descriptions can improve identification of the entity to which an ambiguous mention refers. The reported method uses hierarchical entity representations to prune candidates and improve disambiguation accuracy within its evaluated setting (Knowledge Graphs for Entity Disambiguation, 2025).

Premise: Machine Relations applies those findings only to the defined experimental system. It does not treat them as evidence that commercial AI answer engines use the same architecture, that any particular signal causes a brand recommendation, or that an entity with more records will necessarily appear in an answer.

Observable Measurement Protocol #

Premise: A Machine Resolution study should define the measurement before collecting answers:

1. Entity identity - record the entity name and any aliases that count as the same entity.
2. Query set - freeze the category, problem, comparison, or intent queries being tested.
3. Engine context - record the engine, model or product label when available, account state, locale, and collection date.
4. Answer capture - preserve the complete answer and its displayed citations or source links.
5. Outcome labels - label the entity as present, absent, or ambiguous; separately record whether it is mentioned, cited, or recommended.
6. Repeat observations - collect the same query under the same stated protocol across dates before reporting change.
7. Report denominators - publish the number of eligible answers behind every rate and keep engines, query groups, and collection windows separate.

Premise: The protocol does not combine a mention, citation, and recommendation into one event. It also does not infer causality from a change. If an entity appears more often after a website update, publication, or data correction, the result is an observed association until a design capable of testing causality says otherwise.

Machine Resolution and Search Ranking #

Premise: Search position and Machine Resolution are different observations. A search study records where a URL appears in a defined result set. A Machine Resolution study records whether and how an entity appears in a captured AI answer. Neither measurement can be substituted for the other.

Absence from one AI answer means only that the entity was absent from that recorded answer. It does not prove that the engine cannot identify the entity, that the entity is absent from training or retrieval data, or that a specific authority signal failed.

Interpretation Limits #

Machine Resolution can answer questions such as:

• How often was the entity present in the eligible answers?
• In how many answers was it mentioned, cited, or recommended?
• Did the observed rate differ by engine, query group, or collection window?
• When the name was ambiguous, which entity did the answer describe?

Premise: Machine Resolution alone cannot determine the engine's private ranking factors, source weights, retrieval path, or causal reason for an output. It also cannot prove that earned media, schema, citation velocity, or cross-domain references caused the result. Those explanations remain hypotheses unless they are tested separately.

Relationship to Entity Resolution Rate #

Premise: Entity Resolution Rate is a reported rate produced by a declared observation protocol. Machine Resolution names the broader observation construct; Entity Resolution Rate reports a defined numerator over eligible answers. The rate must state what counted as a resolved entity and what was excluded.

Origin #

Jaxon Parrott coined Machine Resolution within the Machine Relations framework. The term is published here as an MR measurement convention, not as a description of a universally documented engine subsystem.