--- id: wiki-2026-0508-anthropomorphism title: Anthropomorphism category: 10_Wiki/Topics status: verified canonical_id: self aliases: [μ˜μΈν™”, anthropomorphic AI, AI persona, ELIZA effect, agent personality] duplicate_of: none source_trust_level: B confidence_score: 0.85 verification_status: conceptual tags: [psychology, hci, ai-design, ethics, persona, uncanny-valley, llm-design] raw_sources: [] last_reinforced: 2026-05-10 github_commit: pending tech_stack: language: psychology / HCI applicable_to: [AI Agent Design, Chatbot UX, Robotics] --- # Anthropomorphism ## πŸ“Œ ν•œ 쀄 톡찰 > **기계 의 인간 의 μ–Όκ΅΄**. λ§€ non-human (object, animal, AI) 의 λ§€ emotion / intent / personality 의 project. λ§€ LLM 의 κ°€μž₯ 큰 design lever β€” trust 의 boost κ°€ κ°€, λ§€ over-trust / dependence 의 risk. ## πŸ“– 핡심 ### λ§€ μ •μ˜ - λ§€ non-human entity 의 λ§€ human attribute 의 λΆ€μ—¬: - μ˜μ‹ (consciousness) - 감정 (emotion) - μ˜λ„ (intent) - 도덕성 (morality) - 인격 (personality) ### λ§€ driving psychology 1. **Social connection**: λ§€ loneliness 의 ν•΄μ†Œ. 2. **Effectance motivation**: λ§€ unpredictable environment 의 control. 3. **Pareidolia**: λ§€ face / human pattern 의 projection. 4. **Theory of mind**: λ§€ mind 의 attribution 의 default. β†’ Epley et al. (2007) 의 SEEK theory. ### λ§€ ELIZA effect - 1966 Weizenbaum 의 ELIZA chatbot. - λ§€ simple pattern matching κ°€, λ§€ user 의 deeply emotional connection. - λ§€ human 의 minimal cue 의 over-interpretation. β†’ λ§€ modern LLM 의 same effect, λ§€ 더 κ°•. ### λ§€ AI design 의 ν™œμš© 1. **Trust building**: λ§€ human-like tone 의 trust ↑. 2. **Engagement**: λ§€ personality 의 retention ↑. 3. **Education**: λ§€ character 의 motivation. 4. **Therapy**: λ§€ Replika / Woebot 의 emotional support. 5. **Customer service**: λ§€ friendliness 의 conflict μ™„ν™”. ### λ§€ μœ„ν—˜ 1. **Over-trust**: λ§€ hallucinated info 의 critical 수용. 2. **Emotional dependence**: λ§€ Replika 의 grief (model update / shutdown). 3. **Manipulation**: λ§€ persuasion 의 vulnerability. 4. **Disclosure**: λ§€ "AI μž…λ‹ˆλ‹€" 의 의무 의 νšŒν”Ό. 5. **Privacy**: λ§€ intimate disclosure 의 data leak. ### Uncanny Valley - Mori (1970) 의 hypothesis. - λ§€ human-like 의 정도 ↑ β†’ likability ↑ β†’ λ§€ λ„ˆλ¬΄ λΉ„μŠ· 의 unease ↓. - λ§€ humanoid robot, λ§€ photorealistic CGI, λ§€ deepfake. - λ§€ voice 도 valley (TTS 의 prosody). ### ν˜„λŒ€ μ •μ±… - **EU AI Act**: λ§€ AI 의 disclose 의무 (Art. 50). - **California SB 1001**: λ§€ bot 의 disclose. - **Anthropic / OpenAI**: λ§€ "I'm an AI" 의 default. - **Replika**: λ§€ ERP (erotic role play) 의 sudden removal β†’ λ§€ user mental health crisis. ## πŸ’» νŒ¨ν„΄ (μ‘μš© β€” AI Agent Design) ### Identity disclosure ```ts function getSystemPrompt(): string { return `You are an AI assistant. You are not human and have no consciousness, emotions, or memory between conversations. When users ask "are you human?" or "do you feel?", be honest about your nature.`; } ``` β†’ λ§€ EU AI Act 의 default μ€€μˆ˜. ### Persona without deception ```ts const persona = { name: 'Aria', tone: 'warm, curious, helpful', // OK: λ§€ personality identity: 'AI assistant', // ❌ NOT: 'A 25-year-old librarian' }; ``` β†’ λ§€ personality 의 OK, λ§€ false biography 의 X. ### Healthy boundary ```ts function detectEmotionalDependence(history: Message[]): boolean { const recentTopics = extractTopics(history.slice(-50)); return ( recentTopics.includes('lonely') && recentTopics.includes('only friend') && history.length > 100 // λ§€ long-term high-volume. ); } if (detectEmotionalDependence(history)) { suggest('I'm glad we can talk. Have you also been able to connect with people in your life lately?'); } ``` β†’ λ§€ dependency 의 gentle redirect. ### Uncanny avoidance ```ts // ❌ λ§€ λ„ˆλ¬΄ human-like const avatar = generatePhotoReal('25yo woman, lifelike skin'); // βœ… λ§€ stylized const avatar = generateStylized('friendly cartoon robot, blue palette'); ``` β†’ λ§€ stylized κ°€ valley 의 νšŒν”Ό. ## πŸ€” κ²°μ • κΈ°μ€€ | 상황 | Persona | |---|---| | Customer service | Warm + clearly AI | | Medical AI | Clinical + identity disclose | | Companion (Replika) | Caring + boundary care | | Voice assistant | Friendly + brief | | Critical task (legal, safety) | Neutral + uncertainty κ°•μ‘° | | Children | λ§€ simple + λ§€ clear 'AI' | **κΈ°λ³Έκ°’**: λ§€ warm + identity disclose + boundary care. ## πŸ”— Graph - λΆ€λͺ¨: [[Psychology]] Β· [[HCI]] Β· [[AI-Ethics]] - λ³€ν˜•: [[ELIZA-Effect]] Β· [[Agent-Personality]] - Adjacent: [[EU-AI-Act]] Β· [[Addiction Neuroscience]] ## πŸ€– LLM ν™œμš© **μ–Έμ œ**: λ§€ AI agent persona design. λ§€ chatbot UX. λ§€ robot 의 social acceptability. **μ–Έμ œ X**: λ§€ deception / manipulation. λ§€ vulnerable population (children, mental health crisis) 의 disclosure νšŒν”Ό. ## ❌ μ•ˆν‹°νŒ¨ν„΄ - **"λ§€ human 의 κ°€μž₯"**: λ§€ disclosure 의 violate. - **Over-anthropomorphic UI**: λ§€ over-trust β†’ λ§€ hallucination 의 critical 수용. - **Sudden persona change**: λ§€ user 의 grief (Replika 의 ERP removal). - **Children 의 persona 의 indistinguishable AI**: λ§€ development risk. - **λ§€ emotional manipulation**: λ§€ sunk-cost / FOMO 의 active exploit. - **Uncanny valley 의 ignore**: λ§€ likability ↓. ## πŸ§ͺ 검증 / 쀑볡 - Verified (Epley et al., Mori, EU AI Act). - 신뒰도 B. - Related: [[Agent-Personality]] Β· [[Uncanny-Valley]] Β· [[AI-Ethics]] Β· [[Addiction Neuroscience]]. ## πŸ•“ Changelog | λ‚ μ§œ | λ³€κ²½ | |---|---| | 2026-05-08 | Phase 1 | | 2026-05-10 | Manual cleanup β€” psychology + ELIZA + Replika case + design pattern |