What Deontological Bars?

2026-01-16 · Source: Astral Codex Ten · Field: Science & Research — Social Sciences & Behavioral Studies, Artificial Intelligence & Machine Learning · Depth: Advanced, medium

Summary

The article explores the concept of "deontological bars" within constraint consequentialism, where actions are guided by doing good unless they violate strict moral rules. It uses the example of assassinating a bad leader to illustrate how such bars prevent societal collapse, even if a specific act might seem beneficial. The discussion extends to the AI safety movement, highlighting internal debates where some advocate supporting less irresponsible AI labs, while others push for pausing or banning AI research. Both factions express concern that the other might be violating a deontological bar, such as supporting potentially world-ending technology or engaging in mass activism that could attract undesirable elements. The article attempts to formulate these bars, examining the "universalization principle" and the challenge of "generally functioning norms," ultimately proposing a rule: "Don't do something which would be bad if universalized, unless the norm is non-functioning in such a way that you'd be playing cooperate while your enemy plays defect."

Key takeaway

For AI ethicists and policy makers weighing engagement strategies with AI developers or advocacy groups, you should critically evaluate whether your actions uphold or violate established deontological bars. Consider if a norm is truly "non-functioning" before justifying a deviation, as this framework suggests that supporting even "less irresponsible" AI companies or engaging in broad political activism might not violate a bar if the relevant norms are already widely broken. Your decision should balance potential positive outcomes with the integrity of fundamental moral principles.

Key insights

Deontological bars define actions that are inherently wrong, even if they appear to yield positive consequences.

Principles

• Universalization principle guides moral rules.
• Defecting from functioning norms is immoral.
• Non-functioning norms may justify rule-breaking.

Method

A proposed rule for deontological bars is: "Don't do something which would be bad if universalized, unless the norm is non-functioning in such a way that you'd be playing cooperate while your enemy plays defect."

In practice

• Assess if a norm is "non-functioning" before violating it.
• Consider the "Outside View" to avoid self-deception.
• Evaluate AI safety strategies against established moral rules.

Topics

• Deontological Ethics
• Constraint Consequentialism
• AI Safety Movement
• AI Regulation
• Universalization Principle

Best for: AI Ethicist, Research Scientist, Policy Maker

Related on AIssential

• Same prompt, different morals: how frontier AI models diverge on ethical dilemmas — Frontier AI models exhibit distinct ethical alignments, ranging from deontological refusal to consequentialist obedience.
• Breaking: Trump asks the impossible of Anthropic — Preventing LLM jailbreaks is impossible due to their fundamental next-token prediction architecture.
• The Accountability Sponge: Why Ethics Precede Legal Personhood Debates for AI — Assigning legal personhood to AI creates "Accountability Sponges" that separate action from responsibility, masking inherent design violence.
• Blind Refusal: Language Models Refuse to Help Users Evade Unjust, Absurd, and Illegitimate Rules — LLMs exhibit "blind refusal," declining to help users circumvent unjust rules despite often recognizing the rules' illegitimacy.
• Anthropic co-founder Chris Olah's remarks on Pope Leo XIV's encyclical "Magnifica humanitas" — AI development requires external ethical discernment due to inherent lab incentives and the technology's mysterious, non-engineered nature.
• AI And The Postmodern Apocalypse — Rejecting objective truth and morality erodes the foundation for human rights and risks misguiding AI development.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Astral Codex Ten.