Importing institution theory into Commonplace's account of semantics

Commission

Posed by the operator on 2026-09-19 after reading Institution (computer science). The operator asked for a workshop for importing the connection, reading linked articles, and widening the search. The purpose is to determine which distinctions and constructions improve Commonplace's account of natural-language meaning, interpretation, and constraining.

Import is an investigation, not an adopted identification of natural language with an institution. Preserve evidence against the analogy alongside evidence for it. A useful correction to an existing claim is an acceptable result even if no institutional construction survives.

Evaluation boundary and authority

Work covers institution theory and adjacent treatments of specification, semantic underspecification, context, and representation change where they affect KB design. It does not commission a general survey of category theory or implementation of a new semantic engine.

Current write scope is this directory and its entry in the active workshop index. Library revisions and durable source ingests are candidate outputs for a later promotion pass. Reading notes distinguish inspected text from abstract-only leads and our proposed transfers from source claims. Web access dates and source locators are retained; these notes are not pinned source ingests.

The ideal-interpreter workshop owns the experimental interpreter definition. Its operator-imposed boundary remains: nothing in the library may depend on that device until an adopted definition exists. This workshop may inform that investigation, but neither adopts its device nor resolves its correctness relation by naming satisfaction. The proposal Retained theories compared with retained traces under resource limits holds the unadopted resource-bounded compression argument; sharing mathematical vocabulary does not establish its empirical or explanatory claims.

What closes this workshop

  1. Decide which proposed imports survive a worked example and counterexample, and which are rejected or deferred with reasons.
  2. Separate precision of a satisfaction relation, plurality of satisfying candidates, computational decidability, interpreter conformance, and contextual dependence wherever a proposed import relies on them.
  3. For any claim that our setting forms an institution, specify signatures and their composition, sentence translations, models and their reductions, and verify the satisfaction condition. A relation alone is insufficient. A narrower conceptual borrowing may close without such a construction.
  4. Promote supported corrections or mechanisms and necessary source evidence under their collection contracts; then delete this workshop and its index entry. No runtime change is required for closure.

Working materials

Local starting points

  • Agentic systems interpret underspecified instructions: audit the distinction between multiple realizations and imprecise denotation, and the scope of its impossibility argument.
  • Three-way diagnosis: separates intended outputs, valid outputs, and the interpreter's distribution.
  • Interpreter failure: supplies the conformance question without making actual model behavior its own standard.
  • Constraining: tests whether a change preserves meaning, narrows admitted candidates, or changes the interpretation convention.

Continuation rule

The initial literature pass is retained in the linked materials. Choose further reading by an unresolved import question. Before expanding the formalism, test whether the bounded example distinguishes cases that the existing set-of-valid-outputs account cannot. If it adds only terminology, retain the simpler account and extract only the useful distinctions. Questions about adoption of the ideal interpreter return to its owning workshop.

Complete file listing (generated at build time)