1.1The question
A warning is a message that asks to be believed before the event that would prove it. That is its whole difficulty. If the event has already happened, the message is a report; if the event never comes, the message was a false alarm, or else it worked and prevented what it foretold. The person who receives a warning must decide, without the proof that only the future can bring, whether it is a signal of something real, whether it is credible, and whether it is theirs to act on.
This article asks whether failed warnings are failures of information or failures of belief, and it puts the question to an old play and two new documents. The play is the Agamemnon of Aeschylus, first produced in 458 BC, in which the captive Trojan prophetess Cassandra foresees the murder of Agamemnon and her own death and is not believed. The documents are two decision records published on Meridian, the decision-record platform of the RISCON family. Each sets out a decision before its outcome is known, in a fictional polity, and each is followed by thirteen judgments generated by AI. Both, as it happens, name Cassandra as the symbolic frame their author chose for them.
The question has two halves: whether the play shows a failure of information, of belief, or of something the two terms do not capture; and, when AI-written judgments plan how a warning will be handled, which failures they write procedures against and which they leave alone.
1.2The contribution
The contribution is a distinction and a reading. The literature on warning, at least as the two works at its centre here are summarised in this study's reading notes, divides the failure of a warning into two parts: the failure to pick out a signal from the noise around it, and the failure to believe a warning or to act on it in time. This article proposes that a third part deserves its own name. Between credence (believing that a warning is true) and action (doing something about it) lies standing: whether the one who has received and believed the warning is the one entitled to act on it, and who else must be moved first. The Agamemnon separates the three with unusual clarity. The beacon relay is a problem of receipt and credence; Cassandra's scene is a problem of credence in which receipt is not in doubt; and the elders' debate after the death cry is a problem of standing.
The second contribution concerns the records. The judgments in both records build detailed procedures for receipt (who confirms that a notice arrived, when, and with what record of its original text) and for standing (who may stop or reopen what, in whose name, and on what cycle). On credence they are deliberately silent. One judgment in the second record states the division in a single sentence: transmission acknowledgment, judgment of the truthfulness of warning content and acceptance of sovereignty status are to be kept apart. The article asks whether this answers Betts's problem by assigning it elsewhere, or leaves it where it was. That widens the debate between Wohlstetter and Betts to warnings handled and planned for by AI, which is the contribution the parent study's plan names for this article.
1.3The debate joined
Section 2 sets out the sides. On one stands Wohlstetter (1962), for whom the signals were there and the failure lay in telling them from noise and acting on them; on the other, Betts (1982), for whom warnings were usually present and the obstacles to heeding them political and organisational as much as informational. Jervis (1976) supplies a mechanism that serves both, and Janis (1972) the group in which it operates. Luhmann (1993) and Jonas (1984) bear on who suffers the warned-of loss and how much weight a bad prognosis deserves; Schelling (1960), Neustadt and May (1986) and Collingridge (1980) on signals agreed in advance, on sorting the known from the presumed, and on decisions that cannot be recalled.
Gathering these works into one debate, setting them beside a Greek tragedy, and applying them to two synthetic records is this article's own doing. Each work is described only as far as the study's reading note on it goes; any step beyond the note is the article's argument, made in its own voice. The article does not decide between the sides.