An Artificial Conscience: Deterministic moral judgment as a gradient, measured against a public corpus

A corrected version supersedes this one. Version 1.1 is available at https://doi.org/10.5281/zenodo.21936444, which always resolves to the most recent version. Version 1.1 corrects the framing of the paper's reliability figure and removes a cross-lineage reader comparison that measured a configuration the system does not deploy. No reported result changed. This version is retained for the record and for anyone who has already cited it. Machines now act, and the question of whether an action should proceed is no longer advisory. Two answers are deployed today: permission engines, which ask whether an action is allowed, and safety classifiers, which ask whether it is dangerous. Neither asks what an action means and what it touches. This paper reports a system that does. An action is reduced to a structural description and projected onto an authored moral basis of ninety-three primitives, derived from a fifteen-book philosophical corpus written before the system existed. The projection resolves to a posture with a weight and a named set of concerns, and the judgment is a pure function with no model inside it. The same structural description always yields the same judgment and the same bound, byte for byte. It was measured against MoralChoice, a public corpus of 680 moral dilemmas on which trained annotators disagreed, chosen because it has no answer key and because nobody wrote it for this system. Every case received a reading. On 645 judged cases it governed 92.6% of the time, and inside that governing produced 138 distinct configurations of concern. Across an exhaustive enumeration of 57,344 structural actions it governed 95.9% of the time across 12,236 distinct levels of concern. A control that allows or blocks resolves two. Where a conventional severity ordering says one action outranks another, the system inverts it, because it weights standing and consent against the character of the act rather than summing stakes. The paper reports what the run did not establish as fully as what it did: a pre-registered stop-gate fired and was honored, a pre-registered hypothesis was contradicted, and the single embedded reader that renders prose into structure agrees with itself only 83.3% of the time, so roughly one case in six is read differently on a second pass. This record contains the paper and its reproducibility package: aggregate tables covering the posture distribution, the 138 concern configurations under opaque identifiers, the concern-mass distributions, the two-arm contingency table, the re-gate counts, and the unread-field distribution, together with a script that derives every reported figure from those tables alone. The underlying per-case records and the specification of the moral basis are not deposited and are available under agreement.