Just ethical theories as update rules, not action rankings

A new essay on LessWrong argues that ethical theories should be treated as online learning algorithms, not fixed rankings of actions. The author compares utilitarianism, deontology, and virtue ethics to different update dynamics — deontology offers stability but can be brittle in unfamiliar situations, while consequentialism is flexible but vulnerable to noisy estimates. The essay applies this lens to the AGI transition, noting that Anthropic appears to be giving Claude deontological constraints while still justifying them consequentially. The takeaway is that we should judge ethical theories by how they adapt to new information, not just their initial recommendations.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton