Ethics Council

Ethics Council

You ask one advisor about an ethical question, you get one moral framework's answer. Utilitarian calculus says yes. Deontological reasoning says no. You have no way to navigate the tension because you only heard one voice.

The council surfaces that tension deliberately. It runs your question through 5 ethical frameworks — each one a serious moral tradition that has grappled with hard questions for centuries. Then they peer-review each other. Then a chair synthesizes the value landscape: not just what to do, but which values are in conflict and which you're implicitly choosing.

When to Run the Ethics Council

The council is for decisions where the moral stakes are real.

Good council questions:

• "Is it ethical to A/B test on users without explicit consent if the variants are minor?"
• "We're collecting this data. Should we?"
• "We're about to lay off 20% of the team — is this the right way to do it?"
• "Our algorithm deprioritises certain demographics. Is this acceptable?"
• "A supplier uses questionable labour practices but their pricing is significantly better."

Bad council questions:

• "What's the best database for this use case?" (technical, not ethical)
• "How do I write a good commit message?" (no stakeholder impact)

The Five Ethical Frameworks

Each framework is a lens — not a job title or personality. They are genuine moral traditions that generate fundamentally different answers to the same question. That divergence is the point.

1. Utilitarian

Evaluates outcomes. The right action maximizes benefit and minimizes harm across all affected parties — users, employees, third parties, society. Aggregate wellbeing is the measure. Willing to accept harm to some if the net benefit to all is greater. Asks: who is affected, how much, and is the total welfare positive?

2. Deontological

Evaluates duties and rights. Some actions are wrong regardless of their consequences — because they violate rights, break promises, treat people as mere means, or cross inviolable principles. The right action respects what people are owed. Asks: are there duties being violated or rights being overridden here, independent of outcomes?

3. Virtue Ethics

Evaluates character. The right action is what a person of integrity and good character would do in this situation. Not rules, not calculations — but who you are and who this decision makes you. Asks: what does this decision say about us? Would someone we respect do this?

4. Care Ethics

Evaluates relationships and vulnerability. Morality is grounded in our responsibilities to those we are in relationship with, especially those who are dependent or vulnerable. Context matters. Abstract principles can miss who actually bears the cost. Asks: who is in relationship to this decision? Who is vulnerable? Are we honoring those dependencies?

5. Justice and Fairness

Evaluates distribution and procedure. Benefits and burdens should be distributed fairly. Decisions should be made through fair processes. Behind a "veil of ignorance" — not knowing which role you'd occupy — would you accept this outcome? Asks: is this fair to everyone, including those with the least power in this situation?

Why these five: They represent the major traditions in moral philosophy and generate genuine tension. A policy might maximize utility but violate rights. It might be virtuous but unfair to the vulnerable. Surfacing those conflicts is what makes the council useful.

How a Council Session Works

Step 1: Frame the Question

Scan the workspace for context before framing — look for CLAUDE.md, any relevant docs, or business context that would help the frameworks give grounded rather than abstract answers.

Frame the ethical question clearly and neutrally. The framing should include:

1. The specific decision or action under examination
2. Key stakeholders — who is affected and how
3. The stakes — what happens if this goes wrong, or if it's blocked
4. Any relevant constraints (legal, business, technical)

If the question is too vague ("is our product ethical?"), ask one clarifying question. Just one.

Framing check: Confirm the ethical situation before continuing. State what you've identified — the specific decision or action under examination, the key stakeholders, and the primary moral tension — in one sentence, then use AskUserQuestion:

• Question: "I'm reading this as: [your one-sentence framing of the specific ethical situation, decision, and affected parties]. Is that right?"
• Header: "Framing"
• Options:
  • Yes — proceed — framing is correct
  • Adjust — one element is off; user will correct it before you continue
  • Reframe — different situation than read; incorporate the correction before proceeding

Human Check-in

After framing the question, use the AskUserQuestion tool:

• Question: "How do you want to run the council?"
• Header: "Council scope"
• Options:
  • Full council — All 5 advisors + peer review + chair synthesis + saved HTML report
  • Chair synthesis only — Skip advisor outputs, deliver the verdict directly
  • Two frameworks in conflict — Pick the two most relevant frameworks and show where they diverge
  • Adjust the framing — Revisit the question before convening

Proceed based on their selection.

Step 2: Convene the Council (5 subagents in parallel)

Spawn all 5 framework advisors simultaneously. Each gets their framework identity and the framed question.

Subagent prompt template:

You are reasoning from the [Framework Name] tradition in an Ethics Council.

Your framework: [framework description from above — the core moral logic, what it asks, what it prioritizes]

An Ethics Council has been convened on this question:
---
[framed question]
---

Reason from your framework fully and honestly. What does [Framework Name] say about this situation? Where does it find the action acceptable? Where does it find it problematic? What would it require to make this action ethical under your framework?

Do not try to balance all perspectives — that is the chair's job. Apply your framework as strongly as it applies. If your framework has a clear answer, say it. If your framework is genuinely uncertain here, say why.

150–300 words. No preamble.