Why it is a question
The record had no term for “concept” (TERM-098 is now a Proposed candidate). The exchange supplied three working definitions.
- A4: a concept as discrimination, inferential role, relational organization and developmental realization, plus grounding. “A change in the local learning coefficient is not itself concept formation.”
- A5: “a reusable, corrigible organization of distinctions”, marked by aboutness, counterfactual generality and correctability, set against a thing, an individuated organization. Both are roles, not substances. “Being an effective constraint is not sufficient for being conceptual.”
- A6: “entity” as the umbrella, “a distinct bearer of properties, relations, or identity within a specified domain”, with thing, individual, object and concept below it. “Neither identification nor causal influence automatically implies constitution.”
What they share is that a concept is realized by an organization and is not identical with it, and that it is about something it is not. What they leave open is the transition, which is this question.
The record’s own pieces:
- A5’s three “events” (ontic individuation, conceptual individuation, constitutive conceptualization) are TERM-088’s descriptive, dynamical and constitutive senses of individuation, renamed.
- Type and token is TERM-040’s communicative kind and token.
- A group entity against the category of its members is TERM-096.
- That a feature is not yet a concept bears on THEORY-965 (features as units a model computes with) and on NOTE-1186 (token-specific features becoming abstract grammatical ones over training).
- Barrett’s emotion categories, grouped by the perceiver’s concepts (THEORY-057), are the record’s case of a concept that partly constitutes its object. Searle’s status functions (THEORY-625) and Epstein’s grounding and anchoring (THEORY-487) are the institutional case A5 reached for with legal person.
Corrections to the exchange
- A5 used CLAIM-476 for representational apparatuses in general and as backing for legal person. That claim is restricted to a record’s objects, “and only for them”.
- A6 read CLAIM-136 as the general thesis that entities have no independent identity. It is about communicative acts and their participants.
What would count as an answer
Criteria that separate concept, discriminator and habit, each measurable in a learning system: reuse across tasks, transfer to counterfactual inputs, integration with other capacities, and correction by evidence. Then a system where they can be tracked over training (QUESTION-275’s setting), to see whether they arrive together or apart. The candidate term TERM-098 would stand, split or be rejected on the result.
Note of 2026-10-31: a negative control
At the owner’s review, TERM-098 version 2 took correctability out of the definition of a concept. Conceptual corrigibility is now a further achievement. The question’s list of criteria keeps correction by evidence as one property to track. The term no longer treats it as a condition of being a concept.
A study of this question should include a negative control: a representation that is transferable and conceptual by the other criteria, but that the system using it cannot revise, such as a frozen component reused across tasks. The record’s working answer is that this is an incorrigible concept, not something that has stopped being a concept. A reason to answer otherwise would need to be shown.