The moment a failure becomes a character trial, learning stops.
That’s what the Anti-Scapegoat is for. Girard’s scapegoat mechanism gave groups a way to discharge tension by sacrificing a person. The Anti-Scapegoat keeps the pressure release and removes the victim. It puts an object on trial instead: the prompt, the doctrine packet, the handoff, the escalation rule, the stale assumption.
This matters now because AI is multiplying action faster than most organizations can preserve coherence. An agentic workflow breaks, and the lazy conclusion is that someone “used AI wrong.” Often the real failure is that the system never made coherent use possible in the first place.
A group that can attack the artifact can stay honest without turning cruel. A group that can only attack the person will protect feelings, protect status, and lose the truth.
If you want faster learning, give blame a non-human target. Otherwise every mistake becomes a sacrifice.