Luke F. Walton
EN 日本語
Luke F. Walton

The Answerability Quartet

Answerability and Authored AI Systems

Four philosophical works. A reference implementation, set apart.

The Answerability Quartet is a four-paper project on answerability, authorship, and machine-mediated action. It moves from diagnosis, to live market form, to invariant, to construction. The papers stand on their own; Answer Engine — technical note, repository, and live site — is one runnable shape of the builder pattern, offered separately for engineers who want to read and run the contract in code.

DOI records prove the works.
ORCID proves the person.
lukefwalton.com explains the system.

Jump to section
  1. The Answerability Quartet
  2. The papers
  3. Technical implementation
  4. Reading order
  5. Lexicon
  6. Citations

The Answerability Quartet

Four philosophical works — the project's public-facing name for P1 through P4. Each paper remains canonical on its own page.

  1. P1 diagnoses the special failure: no one authored the decision.
  2. P2 shows the live market form: the answer channel gets captured.
  3. P3 states the invariant: the owing survives the route.
  4. P4 gives the builder response: build systems where error stays owned.

The papers

Four papers, P1 → P4. DOI records prove the works; site pages are the canonical surfaces here.

The Decision No One Authored: The Answerability Gap in Generative AI

P1 · Preprint

Diagnoses the special failure: no one authored the decision.

Read when you are working on responsibility gaps, human oversight, meaningful human control, or the difference between control and authorship.

Page DOI PhilPapers CC BY-NC-ND 4.0

The Captured Oracle: Authorship and Agency in the Ethics of Answer-Engine Optimization

P2 · Preprint

Shows the live market form: the answer channel gets captured.

Read when you are working on answer engines, AEO, AI-mediated markets, retrieval, or covert authorship.

Page DOI PhilPapers CC BY-NC-ND 4.0

Building Answerable AI: Why Automation Needs Owned Error

P4 · Working paper

Gives the builder response: build systems where error stays owned.

Read when you are building systems and want the constructive pattern: framed automation, owned error, and answerability as an enabling condition for automation.

Page DOI CC BY-NC-ND 4.0

Technical implementation

Set apart from the four papers: a technical note, an open-source repository, and a live deployment. Answer Engine is one runnable shape of the pattern P4 describes — not a fifth philosophical work, but a reference implementation engineers can clone, run, and push against.

answer-engine

Technical implementation · software artifact

Clone-and-run repository — the runnable implementation the note describes.

Read when you want to clone, run, and read the implementation in code.

Page DOI GitHub Apache-2.0

Reading order

The quartet, P1 → P4.

  1. Start with P1 for the core problem.
  2. Read P2 for answer engines and AEO.
  3. Read P3 for the general invariant.
  4. Read P4 for the builder response.

The quartet moves from diagnosis, to live market form, to invariant, to construction.

For engineers

The quartet stands on its own. If you build systems, here is one way to implement the philosophy — retrieval held outside the model, citations grounded, refusals tested.

  1. Read the technical note for the design contract — what may enter a prompt, what must stay out, and when to decline.
  2. Clone answer-engine, run the example corpus, and read the source in one sitting.
  3. Inspect Ask the Archive on this site for the live deployment behind the search.

Issues, PRs, and disagreement are welcome. This is a discussion, not a lecture — the repo is small on purpose so the contract stays readable; production lessons belong in the conversation too.

Lexicon

The Answerability Lexicon — defined terms used across the quartet.

Answerability
the obligation of a person or institution to give reasons for a decision and stand behind its consequences. (Japanese: 答責性 / tōsekisei, as distinct from 説明責任 / explanation-responsibility.)
Answerability gap
when an automated system produces consequential outputs but no responsible actor can fully explain, own, or correct the decision.
Captured oracle
an answer system whose supposedly neutral outputs are shaped by hidden optimization, incentives, or control over the frame of retrieval.
Decision laundering
using automation to make a human or institutional choice appear neutral, inevitable, or machine-derived.
Scapegoat layer
the lowest-power human or team left to absorb blame for an automated decision they did not meaningfully author.
Authored decision
a decision whose responsible actor, evaluative frame, reasons, and consequences can be identified and answered for.

Citations

Pick a style and copy — APA, Chicago, BibTeX, RIS (for Zotero, EndNote, Mendeley), or a compact prose line with DOI. Per-paper pages carry the same toggle under Cite.

P1 · Preprint

Walton, L. F. (2026). The Decision No One Authored: The Answerability Gap in Generative AI [Preprint]. Zenodo. https://doi.org/10.5281/zenodo.20614374

P2 · Preprint

Walton, L. F. (2026). The Captured Oracle: Authorship and Agency in the Ethics of Answer-Engine Optimization [Preprint]. Zenodo. https://doi.org/10.5281/zenodo.20676327

P3 · Working paper

Walton, L. F. (2026). The Invariant of Answerability (Version 1) [Working paper]. Zenodo. https://doi.org/10.5281/zenodo.20606493

P4 · Working paper

Walton, L. F. (2026). Building Answerable AI: Why Automation Needs Owned Error (Version 1.0) [Working paper]. Zenodo. https://doi.org/10.5281/zenodo.20682307

Technical implementation

Technical implementation · technical note

Walton, L. F. (2026). Answer Engine: A Small Reference Implementation for Citation-Grounded AI Answers (Version 1.1) [Technical note]. Zenodo. https://doi.org/10.5281/zenodo.20686053

Technical implementation · software artifact

Walton, L. F. (2026). answer-engine [Computer software]. Zenodo. https://doi.org/10.5281/zenodo.20676773

One question, one sourced answer. Try: