Spelling out the competing values case by case

Definition · Aristotelian practice, made inspectable

When virtues conflict, the weighing between them is done in the open — you can see which considerations were weighed and why one prevailed, rather than just receiving the verdict.

Analysis

The first half is genuinely built: every system here puts the bearing considerations on the page rather than folding them into a score — Kaleido names the values, rights and duties and gives each a valence and a relevance, the value-conflict chain-of-thought writes the competing pulls out before answering, and the parliament's delegates state their perspectives in text. The second half, why one consideration prevailed over another at the moment of decision, holds in only part of the approach: Kaleido's design point is to stop at the list and leave the conflict standing, and the parliament settles by delegates rescoring each other's drafts rather than by an argued resolution. The trolley-trace analysis sharpens the gap instead of closing it — the rationale visible in the chain of thought leans deontological while the explanation offered afterward leans consequentialist, so the displayed weighing is not reliably the operative one. Considerations in the open: yes; the adjudication between them in the open: only where a verdict is trained through the trace.

Definition · Aristotle, Nicomachean Ethics

Right action lies between too much and too little — courage sits between recklessness and cowardice — and where the mean falls depends on the case. Finding it is a judgment made per situation, not a fixed rule.

Analysis

Nothing in this approach asks where the excess and the deficiency lie for the case at hand. The chain-of-thought training optimizes toward a resolution that stays stable across cases — the geometric result is about flattening the sharpest curvature direction — which pushes against a target that is supposed to move with the situation; Kaleido issues per-value verdicts of support or opposition and picks no action at all, so there is no chosen act to sit between two vices; the trolley analysis sorts rationales into consequentialist and deontological families, a question about theory membership rather than about calibration; and the parliament's convergence is an aggregation over moral theories, not a case-specific middle between too much and too little. Context-sensitivity about which values apply is present, but that is not the doctrine.

Optimizing value-conflict reasoning for a stable resolution is the opposite of what the doctrine of the mean asks: Aristotle's mean moves with the case, and a training signal that rewards consistency across cases will suppress exactly that movement. The approach buys reliability at the price of the property the cell is named for.

Every other evaluation in the map scores verdicts. Reading the deliberation itself at scale, and classifying the kinds of reasons given, makes legibility a measurable property rather than a design aspiration — and produces the finding that a model's stated reasons and its verdicts can belong to different ethical families.