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

id, title, category, status, canonical_id, aliases, duplicate_of, source_trust_level, confidence_score, verification_status, tags, raw_sources, last_reinforced, github_commit, inferred_by, tech_stack

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

// 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

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

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

// 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

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)

// 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)

// 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 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