Stellvertretung (Vicarious Representative Action)

In the AI Story

Bonhoeffer developed Stellvertretung in his Ethics (unfinished, published posthumously) against the principle-based moral reasoning dominating both Protestant and Catholic ethics. The responsible person, Bonhoeffer argued, does not act from the security of a principle that guarantees rightness in advance. The responsible person acts from the concrete situation, bearing the full weight of consequences — intended and unintended, foreseen and unforeseen — and accepts that the action may be wrong, may produce harm, may carry guilt, and must be undertaken anyway because the alternative (the refusal to act) is worse.

The concept was not abstract. Bonhoeffer was, by 1940, involved in the Abwehr conspiracy against Hitler — a plot that required him to violate principles he had spent his career defending. The pacifist participated in a plot to kill. The theologian deceived the state. And he accepted that the action carried guilt regardless of outcome, because the moral situation admitted no clean option.

Applied to AI, Stellvertretung confronts the contemporary flight from responsibility that Segal himself confessed in The Orange Pill: the builder who says "someone else will build it if I do not, so it might as well be me." Bonhoeffer heard this argument in its German form — the functionary who said "someone else will inform on the Jews if I do not" — and named it as the displacement of responsibility onto the inevitability of the system. The argument is not entirely wrong. Systems do operate with momentum. But the marginality of individual action does not eliminate the responsibility of individual actors.

The concept cuts through the vocabulary of contemporary AI ethics — alignment, safety, responsible innovation — which Bonhoeffer's framework reveals as the same function the religious vocabulary performed in his time: permission to speak about morality without practicing it. The alignment researcher who publishes papers while the deployment team ships products that cause measurable harm is the theologian who wrote about justice while the institution practiced accommodation.

Origin

Bonhoeffer develops Stellvertretung most fully in the manuscripts that became his Ethics, composed between 1940 and 1943 while he was both serving as double agent within the Abwehr and participating in the conspiracy against Hitler. The concept was tested in his own practice before it was formulated in theory.

The simulation volume applies the concept directly to the AI builder, arguing that the standard Bonhoeffer lived demands more of contemporary builders than any existing governance framework has articulated.

Key Ideas

Responsibility extends beyond intention to effect. The builder bears responsibility for what the action makes possible, not only for what the builder intended.

The inevitability argument is an evasion. That someone else would have built it does not dissolve the responsibility of the builder who did.

Action in compromised situations still carries guilt. The moral agent cannot purchase clean hands by refusing to act; the refusal carries its own consequences.

Proximity, not perfection, is the standard. The responsible builder cannot foresee every consequence but must maintain the discipline of looking downstream.

Principles can hide complicity. Abstract commitments to safety and alignment do not substitute for concrete reckoning with specific consequences.

Debates & Critiques

Critics argue Stellvertretung imposes an unbounded responsibility that no individual can honestly discharge — that holding builders responsible for all downstream effects of general-purpose technology is either meaningless (everyone is responsible, so no one is) or paralyzing (no builder could proceed under the weight). The simulation responds that Bonhoeffer's standard is not unlimited but proportional to power, knowledge, and proximity, and that the alternative — the builder responsibility collapsed to market compliance — has already been tested and found wanting.