The mechanism is subtler than outright suppression. Suppression requires someone to recognize the moral content and decide to ignore it. Conversion requires only that the institutional processes transform the moral input into a technical output, with no specific person needing to make a decision about the transformation. The input is ethical; the output is optimization-shaped; the transformation is invisible because it was performed by the institutional language itself.
The critical feature of the efficiency language is its apparent neutrality. Efficiency does not claim to override ethics; it simply does not accommodate ethics in its vocabulary. A concern that cannot be stated efficiently appears, inside the efficiency frame, as a concern that has not yet been properly specified — as if the fault were in the formulation rather than in the frame's incapacity to register the content.
Alford's framework identifies this as one of the most damaging features of total-institutional logic. The institution is not refusing to hear moral concerns. It is sincerely trying to hear them, and the sincerity of the attempt produces the transformation. The engineer who proposes the redesign is received as a team member contributing input; the input is processed through the institutional frame; the frame produces an output that fails to retain what the engineer most wanted to convey. She has been heard, in a certain sense; and the hearing has disappeared her testimony.
In the AI transition, efficiency-as-eraser operates at every scale. Safety concerns become cost-benefit analyses. Fairness concerns become optimization targets. Displacement concerns become retraining budgets. Each translation retains something of the original and loses something essential. The cumulative loss — the systematic evacuation of ethical content from institutional deliberation — is the structural condition against which protective structures must be built.
The framework draws on Max Weber's analysis of instrumental rationality and its tendency to colonize domains that resist rational calculation. Alford's distinctive contribution is the empirical demonstration of the mechanism in contemporary organizational life and the identification of its specific operation in the conversion of whistleblower testimony into optimization failures.
The application to AI ethics has been developed by scholars working on the limits of techno-solutionism and the specific institutional pathologies of frontier labs.
Conversion, not suppression. Moral concerns are transformed into technical concerns, which are then disposed of without engaging the ethics.
Invisible operation. No specific person decides to erase the ethics; the institutional language performs the erasure.
Apparent neutrality. The efficiency frame does not claim to override ethics; it simply lacks grammar for ethics.
Sincere dismissal. The witness is heard, converted, and disappeared — all in good faith.
Cumulative loss. The repeated small conversions constitute a systematic evacuation of ethical content from institutional deliberation.