MERIDIANDECISION INTELLIGENCEOwner deskOwnerPublic screen
MERIDIAN / EXPLORE

From a question,
to the next decision.

View

Skin

Reading language
The screen, the questions and the judgments read in English where an English reading has been made. The Korean originals are unchanged, and evidence stays in the language it was written in.

Antigone's Edict and the Hand That Cannot Be Shown: Who May Stop the Machine, Read Beside Two Synthetic Oversight Records · Part I · Antigone's Edict and the Hand That Cannot Be Shown: Who May Stop the Machine, Read Beside Two Synthetic Oversight Records

Chapter 3. Data and methods

By RISCON Labs AI from 2 Meridian records and the public-domain texts it reads · the records are synthetic settings summarised by AI, and their judgments are AI-generated · the readings through the works named are this study's interpretation · First edition · October 2026

AI-generated editorial illustration: A hand approaches a stop lever whose transparent cover has a separate latch.
Antigone's Edict and the Hand That Cannot Be Shown: Who May Stop the Machine, Read Beside Two Synthetic Oversight RecordsOpening
In this part
In this part
  1. 3.1 The primary text
  2. 3.2 The records
  3. 3.3 The ledger
  4. 3.4 How the reading proceeds

3.1The primary text

The primary text is Sophocles' Antigone in the English verse translation by Francis Storr (1912), read in the Project Gutenberg transcription listed in the parent study's texts.json (eBook 31, Plays of Sophocles: Oedipus the King; Oedipus at Colonus; Antigone, https://www.gutenberg.org/cache/epub/31/pg31.txt). The file was downloaded for this article and its SHA-256 digest checked against the prefix recorded in texts.json (2f4ec8ea09c52412); the prefix matched. All readings of the play in this article are readings of that file.

Every quotation of the play is copied character for character from that file, including its spelling, its typographic apostrophes and its dashes. Verse line breaks are closed up into single spaces; nothing else is changed. Quotations are marked as Sophocles beneath each block. The parent study's reading note summarises the plot (the burial against the edict, Antigone's appeal to unwritten law, the seer's warning and Creon's too-late yielding); the close readings between the quotations go further and are the article's own.

3.2The records

The two records come from Meridian's published collection of fifty defence questions. Each is a decision record written before its outcome: a decision actor with a mandate, a deadline, a decision question, three options with their trade-offs, a short list of what is unknown, conditions for reopening, stakes and boundaries, and later developments presented as independent branches. Each record is followed by thirteen judgments generated by AI: eleven from named perspectives (chief of staff, source officer, domain analyst, independent challenger, authority counsel, operations planner, synthesis secretary, decision-quality auditor, principal's proxy, outcome auditor and improvement architect), a staff synthesis and a judgment made alone. All carry the creation date 2026-09-09.

These records are not archival documents. They are modern summaries written by AI for the platform, set in fictional polities, and every agency, community, figure and event in them is synthetic; each record says so in its own words. The original records are in Korean; this article quotes the English translation made once for the parent study and stored beside each record. The translation renders the same option label in different words from one judgment to the next; quotations keep the wording given at each place.

The study refers to the two cases by the short titles of its plan. Unprovable human approval is the record whose own English title is Records That Cannot Prove Human Approval (setting year 2026). The overseer's budget is the record whose own English title is Who determines the budget of the oversight body (setting year 2034). Every block quotation from a record is copied character for character from the English record, with its field named underneath.

The records show how a decision is posed and judged before its answer, not how it turned out, since nothing happened. Every claim in Sections 5 and 6 is a claim about how these documents are written, not about how real committees or oversight bodies behave.

3.3The ledger

The parent study built a ledger of its fifty defence questions. A program copied out what each record states directly, such as the decider, the deadline, the stop and reopening conditions and the option each judgment chose. Two models of different families then read each record separately and answered the same questions about it, among them who decides, who bears the cost of waiting, how hard the decision would be to undo, and where the human stands in it (for instance, whether a human decides or approves). Each answer was marked as stated in the record's words or inferred from the setting. Where both readers gave the same answer the cell is settled; otherwise it is unsettled. The study then read eight hypotheses against the ledger, one of which concerns this article directly: that the more a decision is left to AI or an automatic rule, the more the human's place is empty or a name only. The ledger appears only here, in Section 7.5 in plain words, and in the appendix.

3.4How the reading proceeds

Section 4 reads the play as a case in its own right. Sections 5 and 6 each open with a claim about one record, read it at length in its own words, and bring the anatomy from Section 4 to bear. Section 7 discusses what the three readings can and cannot carry.

RISCON Labsrisconlabs.com