2nd/10_Wiki/Topics/AI_and_ML/AI Humanism.md

---
id: wiki-2026-0508-ai-humanism
title: AI Humanism
category: 10_Wiki/Topics
status: verified
canonical_id: self
aliases: [AI 휴머니즘, human-centric AI, augmented intelligence, co-intelligence, AI for humans]
duplicate_of: none
source_trust_level: B
confidence_score: 0.85
verification_status: conceptual
tags: [ai-ethics, philosophy, human-agency, augmentation, co-intelligence, ux-design, ai-policy]
raw_sources: []
last_reinforced: 2026-05-09
github_commit: pending
inferred_by: Claude Opus 4.7 (manual cleanup 2026-05-09)
tech_stack:
  language: philosophy / process
  applicable_to: [Product Design, UX, Policy, Ethics]
---

# AI Humanism

## 📌 한 줄 통찰 (The Karpathy Summary)
> **AI = 인간 의 replacement X, 인간 의 augmentation**. 매 design / policy 의 인간 dignity 의 center. **Human agency + augmentation + empathy** 의 3 axis. Mollick 의 "Co-intelligence", IBM 의 "Augmented Intelligence" 의 modern formulation.

## 📖 구조화된 지식 (Synthesized Content)

### 핵심 가치 (3 axis)

#### 1. Human Agency (인간 결정권)
- 매 critical decision 의 human final.
- 매 user 의 control / choice.
- 매 AI 의 advisory role.
- "Algorithm 이 결정 했다" 의 excuse 의 reject.

#### 2. Augmentation > Replacement
- 매 task 의 AI 의 partner.
- 매 human capability 의 amplification.
- 매 unique human strength (creativity, judgment, empathy) 의 preserve.
- "Centaur" model (chess: human + AI > AI alone, 2010+ 의 evidence).

#### 3. Empathy & Morality
- 매 AI design 의 affective awareness.
- 매 ethical limit 의 explicit.
- 매 user wellbeing 의 priority.
- 매 vulnerable group (minor, elder) 의 special care.

### 매 인접 movement
- **Humane Tech** (Tristan Harris): 매 user wellbeing 의 design.
- **Slow AI** (analogous to slow food): 매 thoughtful adoption.
- **Critical AI** (academic): 매 power dynamic 의 critique.
- **Human-Centered AI** (Stanford HAI): research framework.

### 매 application area

#### Education
- AI 가 personalized tutor (Khan Academy 의 Khanmigo).
- 매 student 의 own pace.
- Teacher 의 augment X replace.
- Critical thinking 의 emphasize.

#### Healthcare
- AI 의 diagnosis aid (final = doctor).
- Empathic communication.
- Patient autonomy.
- "AI 의 medical advice" 의 disclaimer.

#### Creative work
- 매 artist 의 tool (no replace).
- 매 originality 의 attribution.
- Human-curated dataset.
- 매 AI-generated 의 transparency.

#### Workplace
- 매 employee 의 augment (boring task automation).
- 매 high-judgment work 의 human.
- Reskilling support.
- 매 AI surveillance 의 limit.

### 함정 / antithesis

#### "AI Solutionism"
매 problem 의 AI 답 의 가정.
- 매 social problem 의 root cause (X technology).
- 매 AI 의 새 problem 도 (energy, bias).

#### "Job replacement narrative"
- 매 worker 의 fear / motivation 의 weaponize.
- 매 vendor 의 hype + customer 의 anxiety.
- Reality: replacement 가 task, not job (Acemoglu).

#### "Anthropomorphism"
- 매 AI 의 emotion / consciousness 의 attribution.
- 매 user 의 over-trust.
- "AI 가 sentient" 의 myth.

#### "Techno-optimism"
- 매 AI 의 unconditional adoption.
- 매 critical reflection 의 dismissal.
- "Move fast and break things" 의 limit.

### 매 design principle (HCD - Human-Centered Design)

1. **User-first**: 매 feature 의 매 user 의 real benefit.
2. **Inclusive**: 매 disability / age / culture 의 access.
3. **Transparent**: 매 AI 의 use 의 disclosure.
4. **Reversible**: 매 action 의 undo.
5. **Predictable**: 매 user 의 expectation.
6. **Respectful**: 매 user 의 attention / time.
7. **Educational**: 매 user 의 understanding 의 support.

### Co-intelligence (Mollick 의 modern formulation)
> "Think of AI as a person, not as software."

매 AI 의 collaborator.
- 매 task 의 jointly tackle.
- 매 strength 의 complementary.
- 매 weakness 의 mutual cover.
- "Always invite AI to the table".

### 매 industry shift

#### Past (industrial)
- AI = automation tool.
- Replace > augment.
- Efficiency > meaning.
- 매 worker 의 cost.

#### Modern (AI Humanism)
- AI = partner.
- Augment > replace.
- Productivity + meaning.
- 매 worker 의 capability ↑.

### 매 critique
- "AI Humanism 가 corporate-friendly buzzword".
- 매 large vendor 의 marketing 의 위장.
- 매 worker displacement 의 reality.
- 매 power asymmetry (vendor vs user).

→ Humanism 가 sincere policy + design + accountability 가 매 유의미.

## 💻 패턴 (HCD 의 implementation)

### Human-in-the-loop UX
```tsx
// Bad: 매 action 의 AI 의 자동.
async function processOrder(order) {
  await aiClassify(order);
  await aiPrioritize(order);
  await aiRoute(order);   // 매 step 의 silent
}

// Good: 매 critical step 의 review.
async function processOrder(order) {
  const classification = await aiClassify(order);
  
  if (classification.confidence < 0.9) {
    await requestHumanReview(classification);
  }
  
  // ...
}
```

### Transparent AI use
```tsx
function AIBadge({ feature }: { feature: string }) {
  return (
    <div className="ai-badge">
      🤖 AI-assisted: {feature}
      <button onClick={() => showInfo()}>How?</button>
    </div>
  );
}

// 매 AI feature 의 visible.
```

### Reversible action
```ts
async function applyAISuggestion(suggestion: Suggestion) {
  // Save undo state
  const undoSnapshot = currentState.snapshot();
  
  await applyChange(suggestion);
  
  showToast('Applied AI suggestion. Undo?', {
    action: { label: 'Undo', onClick: () => undoSnapshot.restore() },
    duration: 5000,
  });
}
```

### Empathic prompt design
```ts
// AI assistant 의 emotion-aware
const prompt = `
You are a helpful assistant. The user just shared bad news.

User: My dog passed away yesterday.

Respond with:
1. Acknowledge the emotion (don't dismiss).
2. Brief empathy.
3. Don't immediately offer solutions.
4. Ask if they want to talk or need practical help.
`;
```

### Vulnerable user safeguards
```ts
async function ageGate(user: User) {
  if (user.estimatedAge < 13) {
    redirect('/youth-portal');   // COPPA-compliant
    return false;
  }
  return true;
}

// 매 mental health discussion
async function detectCrisis(message: string) {
  if (containsCrisisIndicator(message)) {
    return {
      response: getCrisisResources(),
      escalateToHuman: true,
    };
  }
}
```

### Augmentation pattern (centaur model)
```ts
// AI 가 draft, human 의 edit
async function writeArticle(topic: string, author: User) {
  const aiDraft = await ai.draft(topic);
  
  return {
    draft: aiDraft,
    suggestion: 'Review and personalize. Your voice is unique.',
    nextStep: 'edit',
  };
}

// 매 final = human.
```

### Education (augment teacher)
```ts
// Khanmigo 식
class AITutor {
  async respondToStudent(question: string) {
    // Don't give answer immediately
    const guidance = await ai.complete({
      prompt: `Student asked: ${question}. Don't give answer. Ask Socratic question to help them think.`,
    });
    return guidance;
  }
}
```

## 🤔 의사결정 기준 (Decision Criteria)

| 상황 | 추천 |
|---|---|
| Critical decision (medical, legal) | Human final + AI advisory |
| Creative work | Human-led + AI assist |
| Repetitive task | AI-led + human override |
| Customer-facing | Transparent + reversible |
| Vulnerable user | Extra safeguard + human escalation |
| Educational | Augment teacher / student, no replacement |
| Workplace automation | Augment worker, reskilling support |

**기본값**: AI = advisor + augment. Human = final + meaning.

## ⚠️ 모순 및 업데이트 (Contradictions & Updates)
- **Efficiency vs humanism**: 매 user 의 augment 가 efficiency ↓ (slower than full automation). 매 trade-off.
- **"Human in the loop" 의 fatigue**: 매 review 의 over-load → rubber stamp.
- **Vendor 의 humanism marketing vs reality**: 매 large vendor 의 humanism rhetoric + 매 worker displacement reality.
- **Inclusivity 의 cost**: 매 vulnerable user 의 safeguard 가 development cost.
- **Power dynamic**: 매 design choice 의 매 user 의 가 vendor / employer 의 가 결정.
- **AI 의 emergent capability**: 매 augmentation tool 가 replacement 가 될 수.

## 🔗 지식 연결 (Graph)
- 부모: [[AI-Ethics]]
- 관련: [[AI Literacy]] · [[AI Accountability]] · [[AI 거버넌스 정책(AI Usage Policy)|AI-Governance-Policy]] · [[AI Safety and Alignment]]
- Adjacent: [[Universal Basic Income (UBI)]]

## 🤖 LLM 활용 힌트 (How to Use This Knowledge)

**언제 이 지식을 쓰는가:**
- AI product 의 design philosophy 결정.
- 매 AI feature 의 UX review.
- 매 corporate AI strategy 의 ethical assessment.
- "AI 가 인간 대체?" 의 nuanced 답.
- AI policy 의 framing.

**언제 쓰면 안 되는가:**
- Specific technical implementation (engineer).
- Hard regulation (legal counsel).
- Crisis response (incident team).
- Marketing 의 buzzword 의 generation.

## ❌ 안티패턴 (Anti-Patterns)
- **"AI 가 모든 거 답"**: solutionism.
- **매 user 의 silent automation**: agency 위반.
- **Reversibility 없음**: 매 user 의 trapped.
- **Vulnerable user 의 same UX**: harm 위험.
- **"Human in the loop" 의 fatigue + rubber stamp**: 가짜 oversight.
- **Anthropomorphism + over-trust**: 매 AI 의 limitation 의 ignore.
- **Marketing humanism + reality replacement**: hypocrisy.

## 🧪 검증 상태 (Validation)
- **정보 상태:** verified (concept-level).
- **출처 신뢰도:** B (Stanford HAI, IBM Augmented Intelligence framework, Mollick "Co-Intelligence", Tristan Harris Humane Tech, MIT Sherry Turkle).
- **검토 이유:** Manual cleanup. Philosophy 가 long-term 안정. 매 application 가 evolving.

## 🧬 중복 검사 (Duplicate Check)
- **기존 유사 문서:** [[AI-Ethics]] (parent), [[AI Literacy]] (subset), [[AI for Social Good (AI4SG)]] (related).
- **처리 방식:** KEEP (philosophical / design framework).
- **처리 이유:** AI Humanism 가 distinct philosophy + design discipline.

## 🕓 변경 이력 (Changelog)
| 날짜 | 변경 내용 | 처리 방식 | 신뢰도 |
|------|-----------|-----------|--------|
| 2026-05-08 | P-Reinforce Phase 1 정규화 | UPDATE | A |
| 2026-05-09 | Manual cleanup — 3 axis + design pattern + 비판 + 안티패턴 + Co-intelligence 추가 | UPDATE | B |