Operationalization, executability, automation, and autonomy

Commissioned: 2026-10-31 by the operator. Revised: 2026-10-31 following operator feedback on dimensionality, automation/autonomy, source ingestion, and locating application-area generalization in task definition rather than operationalization. Revised again: 2026-10-31 after merge, review, and the ingestion of five candidate sources. Checked: 2026-11-01; the two open checks below were run and recorded, no library file changed. Revised: 2026-11-02 after ingesting Suchman's 1985 thesis and rewriting the Suchman case from it.
Status: scoped conceptual investigation, not adopted vocabulary or an empirical test.
Repository baseline: d19065b3aef33ffca92fa2f72f5203251fcae3b0 (main when opened).

Purpose

Commonplace aims to make open-ended text-mediated knowledge work progressively executable by LLM-based systems. Investigate how operationalizing a methodology, executing it, automating functions, and delegating control are related but different.

The operator's revision questions are binding on this workshop:

  1. Generalizing a methodology and constraining it are different changes, but they are not the only dimensions. Distinguish the method's domain and coverage from the executor's operational support and actual reach.
  2. Do not collapse automation (performed work) with autonomy (discretion, authority, and unattended continuity). Test whether operationalization suffices to make a task fully executable and fully automated by LLMs.
  3. Verify which literature is already ingested and whether full sources can be ingested; do not create duplicates, assume an abstract is a full source, or assume public retrieval licenses redistribution.
  4. Revise architectural implications using the multidimensional account rather than a one-way chain from methodology to code.
  5. Task definition fixes the application area, objective, admitted cases and success criteria. Operationalization concerns how knowledge K becomes usable for this task T by executor E under specified conditions; generalizing to T' is a separate task-scope or method-scope investigation, not part of operationalization for T.
  6. The same task-relative operation may work a theory into methodology or existing methodology into an executor-facing skill. Name source and output when discussing provenance; these are not compulsory successive stages. Preserve the narrower operationalized-from artifact-lineage rule.
  7. Revisit vocabulary and extraction recommendations only after those distinctions are settled.

Scope and evidence requirements

  • Compare human checklists, pen-and-paper algorithms, natural-language LLM skills, deterministic scripts, agent workflows, and human–machine arrangements.
  • Reuse current actionable methodology, methodological/computational closure, constraining, codification, and behavioral authority before minting vocabulary.
  • Respect the existing operationalized-from lineage meaning in ADR 054.
  • Read sources using the actual discipline's meaning; evaluate transfer rather than simply borrowing terminology. A check for the availability of original texts is not a completed source ingest.
  • Do not implement machinery, modify shipped instructions or definitions, or claim measured comparative improvement without a trial.

Present findings and limitations

The revised exploration treats task T as supplying the application area and success boundary, while operationalization concerns making existing knowledge K usable by executor E to perform T in setting S. Generalizing the knowledge or broadening its task area requires a separate scope claim and, where appropriate, a revised task T'. A task-specific operationalization can entail constraining, method elaboration, method selection, tools or integration, but does not require a particular output format or a fixed sequence of steps. This is not an intrinsic rank of an artifact. Theory → methodology and methodology → skill are two possible source-to-task paths; neither is a necessary intermediate stage. The accepted operationalized-from label remains the narrower methodology-to-procedure lineage claim. The exploration also separates actual task execution, automation, discretionary autonomy, unattended completion, and warranted outcome quality. No new technical definition has been adopted.

It also distinguishes full bounded task automation by computation from decision autonomy. An LLM can supply decisions not settled by the retained method; a fully written procedure alone cannot confer required capabilities, authority or reliability.

The second revision rests on nine new ingests: Norman, the Stanford Encyclopedia entries on the Church–Turing thesis and operationalism, Plotkin, CMMN 1.1, Haynes and three secondary readings of Suchman (Dittrich, Duguid, and a 1997 course summary), beside the three reports already in the KB. Section 6 of the exploration turns them into worked cases. Section 0 adds three conditions to the provisional relation from the operationalism literature. The Suchman book itself is not ingested; on 2026-11-02 its earlier version, the 1985 PARC thesis, was ingested in full and section 6.6 was rewritten from it, with book-only material still marked secondhand.

The review of the merged draft found three faults, now corrected in the exploration: the findings restated the commission, the actionable-methodology relation was described as narrower than it is, and five workshop-local coordinates were introduced without saying which are new. Only generality and operational integration are new; the others reduce to existing definitions.

Answers to the binding questions, 2026-10-31

  1. Answered and adopted. The method's domain is the task's application area, which the actionable-methodology definition now carries as the task element; the executor's support is the operator, operations, target and setting elements; coverage is closure without the settling condition; reach is the relation holding over the admitted cases. Exploration sections 0 and 2.
  2. Answered in the negative by cases, with an executor-relative sharpening. Operationalization coincides with automatability only for an executor permitted no insight (the effective method, 6.1). Haynes was operationalized in 2009 with nothing automated, and the same checklist is now a candidate for LLM execution of its administering function without any change to the artifact (6.5): automatability is a property of the artifact and executor pair, and it changed when the executor changed. What became automatable is the reader's function, not the surgical role, which is Norman's distinction (6.2); whether the outcome gain survives that transfer is untested, because the study identified only the bundle. CMMN is executable with no function automated and no engine discretion (6.3). So operationalization does not suffice: it makes an artifact consumable by whichever executors can read it, and which functions then move to computation is set by the task boundary and the executor's operations, not by the artifact. Automation, decision autonomy and unattended continuity are kept apart in section 3 and already in the closure note.
  3. Done. Ten ingests, no duplicates, abstracts distinguished from full sources, snapshots kept ignored. The Suchman 1985 thesis now stands in for the 1987 book, which is the only edition gap left. The primary-source check of 2026-11-02 found nothing else free: Orr's Talking About Machines is a licensed ebook, his 1986 and 1990 papers are paywalled, and the purchase is deferred because the Orr candidate accepts an agent-runtime instance instead. Turing 1936 and Bridgman 1927 are free but no candidate waits on them.
  4. Answered in the workshop; extraction checked 2026-11-01 and declined. Section 4 replaces the one-way chain with a map of independent changes and a diagnosis-first rule. The check placed each of the ten failure mechanisms in the library: applicability and unavailable operations are the task and operations elements of the actionable definition; wrong interpretation and weak competence are the underspecification and interpreter questions of three-way diagnosis; a missing method decision is methodological closure and an inappropriate handoff is per-function reporting in the closure note; failed activation is storage without activation; context drift is in determinability; poor warrant is oracle-bounded autonomy. Only inadequate feedback has no note, and it is already a section 8 candidate waiting on an agent-runtime instance. The table adds a routing index across those notes, not a mechanism, so it stays here under the closure rule's path (b); the relaxing-candidate signals cover the codify-or-relax branch of the same rule.
  5. Answered and adopted, as item 1: the task is an input to the relation, and widening it is a separate claim.
  6. Answered; the pending sentence withdrawn 2026-11-01. Theory to methodology and methodology to skill are two instances of one task-relative operation, not stages. operationalized-from stays the narrow lineage edge. A sentence in the actionable definition saying that "operationalize" is ordinary English whose technical content is the actionable relation was considered. The vocabulary collision ledger row for operational already fixes the writing decision: the workshop sense is not adopted, bare "operational" does not carry it, and the row escalates to a definition only when a library note needs that sense. No library note does, so the sentence is not written.
  7. Answered. The distinctions are settled enough to revisit vocabulary: the recommendation is no new term, and items 4 and 6 above confirm that the two remaining candidates for a definition change are covered by existing notes. Extraction candidates are in exploration section 8; each names the case or source it waits for.

Closure rule

Close only after either (a) making an evidence-backed, nonduplicative theory extraction or (b) documenting why existing definitions suffice and the additional distinction stays in workshop analysis. Current position, 2026-11-01: both conditions are met. The task-relative claim is extracted on path (a) as an amendment to the actionable-methodology definition (disposition item 7); the remaining distinctions are documented on path (b) in disposition item 2 and in the item 4 check above. The two desk checks in that list were run on 2026-11-02: the closure-artifact claim went into the closure note's scope section as a second path (a) extraction, and the Haynes-transfer claim is covered by the representational-form and actionable-methodology definitions with the closure note. What keeps the workshop open is the rest of the section 8 candidate list, whose entries wait for cases the KB does not yet hold. On close, move that list to kb/log.md as one entry so the waiting candidates survive the workshop's deletion. Strong causal claims about the benefit of operationalization require comparisons for the same T, E and objective, with documented changes to K, instructions or environment. Claims about generality require separate comparison on T' with its application area declared. The proposed tests are in the exploration; none were run.

Deliverable

Complete file listing (generated at build time)