hum-ethics

asgard-ai-platform ｜ 規劃與釐清 ｜ 原文 ｜ MIT

用義務論、功利主義、德行倫理、正義理論分析道德兩難

---
name: "hum-ethics"
description: "運用倫理學框架，包括義務論、功利主義、德行倫理與正義理論，分析道德困境並做出有原則的決定。當使用者提出具體的道德兩難、有倫理意涵的決定，或需要結構化的多框架倫理分析時使用，即使他們說的是「這樣做對嗎」「這個決定有哪些倫理意涵」「用不同的倫理視角評估這個兩難」也適用。"
metadata:
  category: "WP-19 文學院/人文"
  tags: ["humanities", "ethics", "moral-philosophy", "decision-making"]
---

# 倫理分析

## 概述

倫理分析把道德哲學的框架套用到現實世界的兩難上。沒有任何單一框架能給出所有答案，價值在於透過多種倫理視角檢視一個情境，並把其中的張力明確地攤出來。

## 框架

```
鐵律：要套用多個框架，不能只用一個

不同的倫理框架，對同一個兩難可能得出「不同」的結論。
只透過一個視角分析是不完整的。至少套用兩個框架，
並明確比較它們在哪裡一致、在哪裡分歧。
分歧「本身」就是洞見。
```

### 四大框架

**1. 義務論（Kant）**：以義務為基礎的倫理
- 行動的對錯取決於規則/義務，與結果無關
- **定言令式**：只依據你能希望成為普遍法則的規則行動
- 測試：「如果每個人都這樣做，還行得通嗎？」
- 優點：一致，保護個人權利。缺點：僵硬，無法處理義務之間的衝突。

**2. 功利主義（Mill、Bentham）**：以結果為基礎的倫理
- 正確的行動是為最大多數人帶來最大的善
- 計算所有受影響各方的總利益減去總傷害
- 測試：「這樣做能讓整體福祉最大化嗎？」
- 優點：重視結果，務實。缺點：可能為了多數人的利益而替傷害少數人辯護。

**3. 德行倫理（Aristotle）**：以品格為基礎的倫理
- 關注有德行的人會怎麼做，而不是規則或結果
- 核心德行：勇氣、節制、正義、審慎、誠實、同情
- 測試：「品格良好的人會怎麼做？」
- 優點：整體性、對情境敏感。缺點：主觀，難以操作化。

**4. 正義理論（Rawls）**：以公平為基礎的倫理
- 決定應該像是站在「無知之幕」後面做的，也就是不知道自己在社會中的位置
- 優先考慮社會中最弱勢的成員
- 測試：「如果我可能是任何一位受影響的人，我還會接受這個結果嗎？」
- 優點：處理不平等。缺點：對日常決定不切實際。

### 分析步驟

1. **陳述兩難**：要決定的是什麼？有哪些選項？
2. **找出利害關係人**：誰受影響？怎麼受影響？
3. **套用每個框架**：每個框架建議怎麼做？
4. **比較建議**：它們在哪裡一致？在哪裡分歧？
5. **把張力明確化**：哪些價值在衝突？
6. **附上理由提出建議**：你最看重哪個（些）框架，為什麼？

## 輸出格式

```markdown
# 倫理分析：{兩難}

## 兩難陳述
- 決定：{必須決定的事}
- 選項：A) {選項} B) {選項}
- 利害關係人：{誰受影響}

## 框架分析
| 框架 | 建議 | 推理 |
|-----------|---------------|-----------|
| 義務論 | A / B | {以義務為基礎的推理} |
| 功利主義 | A / B | {結果的計算} |
| 德行倫理 | A / B | {以品格為基礎的推理} |
| 正義理論 | A / B | {公平的推理} |

## 趨同 / 分歧
- 一致之處：{各框架吻合的地方}
- 分歧之處：{它們在哪裡分歧、為什麼}
- 核心張力：{互相衝突的根本價值}

## 建議
{決定，並明確說明優先考量了哪些價值、接受了哪些取捨}
```

## 範例

### 正確的應用
**情境：** 公司該不該在沒有搜索票的情況下，把使用者資料交給執法單位，協助抓捕罪犯？
| 框架 | 建議 | 推理 |
|-----------|---------------|-----------|
| 義務論 | **不該** | 使用者同意的條款承諾了隱私。違反這個承諾就是違背義務。把沒有搜索票就分享資料這件事普遍化，會摧毀人們對所有數位服務的信任。 |
| 功利主義 | **也許** | 取決於傷害的計算：抓到一名罪犯 vs 侵蝕數百萬人的隱私。如果罪行夠嚴重，總效用可能傾向分享。 |
| 德行倫理 | **不該** | 誠實、值得信賴的公司會信守承諾。有勇氣的公司會對抗政府的壓力。 |
| 正義（Rawls） | **不該** | 站在無知之幕後面，你會希望自己的資料受到保護，尤其是如果你屬於容易受到不當監控的弱勢群體。 |

**趨同**：四個框架中有三個說不該。張力在於公共安全（功利主義）與個人隱私（義務論、以權利為基礎）之間 ✓

### 不正確的應用
- 「功利主義說要分享資料，所以就分享」 → 只套用了一個框架。違反鐵律：要套用多個框架。

## 常見陷阱

- **倫理分析 ≠ 道德審判：** 目標是讓推理過程透明，不是宣判誰「好」或「壞」。
- **文化相對性是真的，但有限度：** 倫理規範因文化而異，但有些原則（不要酷刑、不要奴役）被廣泛認為是普世的。要承認文化脈絡，但不要滑入徹底的相對主義。
- **功利主義很容易被濫用：** 「為最大多數人帶來最大的善」可以為針對少數人的可怕行為辯護。一定要搭配以權利為基礎的分析。
- **真正的兩難沒有乾淨的答案：** 如果答案很明顯，它就不是兩難了。分析的價值在於把取捨講明白，不是找出「正確答案」。
- **找出哪些利害關係人，會改變分析：** 忘了某個利害關係人群體（未來世代、非人類動物、間接受影響者），會讓分析產生偏差。

## 參考資料

- 電車難題的各種變體及其意涵，見 `references/trolley-problems.md`
- 商業倫理案例研究，見 `references/business-ethics-cases.md`

---
name: "hum-ethics"
description: "Apply ethical frameworks — deontology, utilitarianism, virtue ethics, and justice theory — to analyze moral dilemmas and make principled decisions. Use this skill when the user presents a concrete moral dilemma, a decision with ethical implications, or needs a structured multi-framework ethical analysis, even if they say 'is this the right thing to do', 'what are the ethical implications of this decision', or 'evaluate this dilemma through different ethical lenses'."
metadata:
  category: "WP-19 文學院/人文"
  tags: ["humanities", "ethics", "moral-philosophy", "decision-making"]
---

# Ethical Analysis

## Overview

Ethical analysis applies moral philosophy frameworks to real-world dilemmas. No single framework provides all answers — the value is in examining a situation through multiple ethical lenses and making the tensions explicit.

## Framework

```
IRON LAW: Apply Multiple Frameworks, Not Just One

Different ethical frameworks can reach DIFFERENT conclusions for the same
dilemma. Analyzing through only one lens is incomplete. Apply at least
two frameworks and explicitly compare where they agree and disagree.
The disagreement IS the insight.
```

### Four Major Frameworks

**1. Deontology (Kant)** — Duty-based ethics
- Actions are right or wrong based on rules/duties, regardless of outcomes
- **Categorical Imperative**: Act only according to rules you could will to be universal
- Test: "If everyone did this, would it still work?"
- Strength: Consistent, protects individual rights. Weakness: Rigid, can't handle conflicting duties.

**2. Utilitarianism (Mill, Bentham)** — Consequence-based ethics
- The right action produces the greatest good for the greatest number
- Calculate total benefit minus total harm across all affected parties
- Test: "Does this maximize overall well-being?"
- Strength: Outcome-focused, pragmatic. Weakness: Can justify harming minorities for majority benefit.

**3. Virtue Ethics (Aristotle)** — Character-based ethics
- Focus on what a virtuous person would do, not rules or outcomes
- Core virtues: courage, temperance, justice, prudence, honesty, compassion
- Test: "What would a person of good character do?"
- Strength: Holistic, context-sensitive. Weakness: Subjective, hard to operationalize.

**4. Justice Theory (Rawls)** — Fairness-based ethics
- Decisions should be made as if from behind a "veil of ignorance" — not knowing your own position in society
- Prioritize the least advantaged members of society
- Test: "Would I accept this outcome if I could be anyone affected?"
- Strength: Addresses inequality. Weakness: Impractical for everyday decisions.

### Analysis Steps

1. **State the dilemma**: What is the decision? What are the options?
2. **Identify stakeholders**: Who is affected? How?
3. **Apply each framework**: What does each framework recommend?
4. **Compare recommendations**: Where do they agree? Where do they diverge?
5. **Make the tensions explicit**: What values are in conflict?
6. **Recommend with justification**: Which framework(s) do you weight most, and why?

## Output Format

```markdown
# Ethical Analysis: {Dilemma}

## Dilemma Statement
- Decision: {what must be decided}
- Options: A) {option} B) {option}
- Stakeholders: {who is affected}

## Framework Analysis
| Framework | Recommendation | Reasoning |
|-----------|---------------|-----------|
| Deontology | A / B | {duty-based reasoning} |
| Utilitarianism | A / B | {consequence calculation} |
| Virtue Ethics | A / B | {character-based reasoning} |
| Justice Theory | A / B | {fairness reasoning} |

## Convergence / Divergence
- Agree on: {where frameworks align}
- Disagree on: {where they diverge and why}
- Core tension: {the fundamental values in conflict}

## Recommendation
{Decision with explicit justification of which values are prioritized and which trade-offs are accepted}
```

## Examples

### Correct Application
**Scenario:** Should a company share user data with law enforcement without a warrant to help catch a criminal?
| Framework | Recommendation | Reasoning |
|-----------|---------------|-----------|
| Deontology | **No** | Users consented to terms that promise privacy. Breaking that promise violates a duty. Universalizing warrant-less sharing would destroy trust in all digital services. |
| Utilitarianism | **Maybe** | Depends on harm calculus: catching one criminal vs eroding privacy for millions. If the crime is serious enough, total utility might favor sharing. |
| Virtue Ethics | **No** | An honest, trustworthy company keeps its promises. A courageous company stands up to government pressure. |
| Justice (Rawls) | **No** | From behind the veil of ignorance, you'd want your data protected — especially if you're in a vulnerable group subject to wrongful surveillance. |

**Convergence**: 3 of 4 frameworks say no. The tension is between public safety (utilitarian) and individual privacy (deontological, rights-based) ✓

### Incorrect Application
- "Utilitarianism says share the data, so share it" → Only one framework applied. Violates Iron Law: apply multiple frameworks.

## Gotchas

- **Ethical analysis ≠ moral judgment**: The goal is to make the reasoning transparent, not to pronounce someone "good" or "bad."
- **Cultural relativity is real but has limits**: Ethical norms vary across cultures, but some principles (don't torture, don't enslave) are widely considered universal. Acknowledge cultural context without falling into full relativism.
- **Utilitarianism is easily abused**: "Greatest good for the greatest number" can justify horrific acts against minorities. Always pair with rights-based analysis.
- **Real dilemmas have no clean answer**: If the answer were obvious, it wouldn't be a dilemma. The value of the analysis is making trade-offs explicit, not finding "the right answer."
- **Stakeholder identification changes the analysis**: Forgetting a stakeholder group (future generations, non-human animals, indirect affected parties) biases the analysis.

## References

- For trolley problem variations and their implications, see `references/trolley-problems.md`
- For business ethics case studies, see `references/business-ethics-cases.md`