refactor(topics): 멀티 에이전트용 지식 재편 — _Common(공통 기본기) + Domain_* 구조

에이전트 8종(대화형/프로그래머 C·S/디자이너/설계자/기획자/QA/PD/PM)에게 [공통 기본 능력 + 롤별 Specialty] 2층으로 지식을 주입하기 위한 재분류. 문서 내용·포맷은 무수정, 폴더 이동만 (6,372개 문서 수 보존 확인). - Topic_Programming → Domain_Programming (내부 구조 보존) - Topic_Graphic → Domain_Design - Topic_Business → Domain_Product - Topic_General → Domain_General - _Common 신설: Math(구 Topic_Math_Specialty), Reasoning(구 General/From_Thinking & Reasoning), Reasoning_Creativity(구 General/From_창의성), Communication(Poetic_Blog_Writing + From_writing) - 타 도메인의 From_* 폴더는 유지 (출처 표기일 뿐, 이미 도메인에 맞게 분류된 문서) - 빈 폴더 정리 (memory/procedures) - 에이전트→폴더 매핑은 workspace의 .astra/agent-knowledge-map.json (9개 에이전트) Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-07-11 11:05:56 +09:00
parent 6549ead309
commit c24165b8bc
6193 changed files with 1717 additions and 31 deletions
@@ -0,0 +1,164 @@
+---
+id: wiki-2026-0508-habit-formation
+title: Habit Formation
+category: 10_Wiki/Topics
+status: verified
+canonical_id: self
+aliases: [Habit Loop, Behavior Automation, Habituation]
+duplicate_of: none
+source_trust_level: A
+confidence_score: 0.9
+verification_status: applied
+tags: [psychology, behavior, neuroscience, habits]
+raw_sources: []
+last_reinforced: 2026-05-10
+github_commit: pending
+tech_stack:
+  language: en
+  framework: behavioral-psychology
+---
+
+# Habit Formation
+
+## 매 한 줄
+> **"매 habit = cue → routine → reward 의 basal-ganglia automatization"**. 매 21-day myth 의 false — Lally 2010 의 median 66 days (range 18-254). 매 2026 의 wearables + LLM coaching + JITAI (just-in-time adaptive intervention) 의 active.
+
+## 매 핵심
+
+### 매 habit loop (Duhigg / Wood)
+1. **Cue**: 매 trigger — time, place, emotional state, preceding action, people.
+2. **Routine**: 매 behavior 자체.
+3. **Reward**: 매 reinforcement — neural prediction error.
+4. **Craving** (Wood addition): 매 anticipation 의 cue→reward.
+
+### 매 neural substrate
+- **Goal-directed**: 매 prefrontal + dorsomedial striatum — early learning.
+- **Habitual**: 매 dorsolateral striatum (sensorimotor loop) — automatization.
+- **Switch**: 매 overtraining + stable context → habitual takeover.
+
+### 매 formation 의 핵심 levers
+- **Implementation intentions** (Gollwitzer): "When X, I will Y" — 매 효과 size large (d ≈ 0.65).
+- **Context stability**: 매 same time + place 의 consistency.
+- **Friction reduction**: 매 cue salience ↑, 매 obstacle ↓.
+- **Temptation bundling** (Milkman): 매 desired + pleasurable 결합.
+- **Identity-based**: 매 "I am someone who..." (Clear).
+
+### 매 응용
+1. Atomic Habits 의 4 laws (obvious, attractive, easy, satisfying).
+2. Health behavior change (exercise, medication adherence).
+3. Productivity (deep-work blocks).
+4. Habit-stacking (after-X-then-Y).
+
+## 💻 패턴
+
+### Habit tracker (streak + context)
+```python
+from dataclasses import dataclass, field
+from datetime import date, timedelta
+
+@dataclass
+class HabitLog:
+    name: str
+    cue_context: dict           # {time, location, preceding_action}
+    completions: list[date] = field(default_factory=list)
+
+    @property
+    def streak(self) -> int:
+        if not self.completions:
+            return 0
+        s, today = 1, max(self.completions)
+        for i in range(1, len(self.completions)):
+            if today - self.completions[-1 - i] == timedelta(days=i):
+                s += 1
+            else:
+                break
+        return s
+```
+
+### Implementation intention generator
+```python
+def implementation_intention(goal: str, cue: str, action: str) -> str:
+    return f"When {cue}, I will {action} in service of {goal}."
+
+# When I pour my morning coffee, I will do 10 push-ups in service of strength training.
+```
+
+### Habit-stacking chain
+```python
+def stack(anchor: str, new_habit: str, reward: str | None = None) -> dict:
+    return {
+        "anchor": anchor,
+        "new_habit": new_habit,
+        "rule": f"After {anchor}, I will {new_habit}.",
+        "immediate_reward": reward,
+    }
+```
+
+### JITAI delivery decision
+```python
+def jitai_should_deliver(state: dict) -> bool:
+    """Deliver intervention only when receptive + context-matched + low burden."""
+    return (state["stress"] < 0.7
+            and state["cognitive_load"] < 0.6
+            and state["context_match"] > 0.8
+            and state["recent_interventions_24h"] < 3)
+```
+
+### Lally formation curve
+```python
+import numpy as np
+
+def automaticity(day: int, asymptote: float = 0.95, k: float = 0.04) -> float:
+    """Asymptotic automaticity (Lally 2010 fit)."""
+    return asymptote * (1 - np.exp(-k * day))
+# day 21 → ~0.58, day 66 → ~0.88
+```
+
+### Context-cue salience score
+```python
+def cue_salience(cue: dict, history: list[dict]) -> float:
+    """Higher when cue co-occurs with successful routine."""
+    matches = [h for h in history if all(h.get(k) == v for k, v in cue.items())]
+    if not matches:
+        return 0.0
+    return sum(h["completed"] for h in matches) / len(matches)
+```
+
+## 매 결정 기준
+| 상황 | Strategy |
+|---|---|
+| Brand-new habit | implementation intention + context stability |
+| Existing routine + new addition | habit stacking (anchor) |
+| High-friction habit | reduce friction first, then add cue |
+| Reward-poor habit | temptation bundling |
+| Identity-level change | identity-based ("I am the kind of person who...") |
+| Relapse prevention | re-stabilize context, restore cue |
+
+**기본값**: 매 implementation intention + same context daily + 60-90 day window. 매 21-day promise X.
+
+## 🔗 Graph
+- 부모: [[Operant_Conditioning|Operant Conditioning]]
+- 변형: [[Habit Stacking]]
+- 응용: [[CBT]]
+- Adjacent: [[Basal Ganglia]] · [[Self-Determination Theory]]
+
+## 🤖 LLM 활용
+**언제**: 매 implementation intention drafting, 매 habit-stack anchor 의 brainstorm, 매 obstacle anticipation, 매 daily reflection scaffold.
+**언제 X**: 매 individual psychological diagnosis (e.g., compulsion vs habit) — 매 clinical professional 필수.
+
+## ❌ 안티패턴
+- **21-day promise**: 매 individual variance 무시 — 매 18-254 day range.
+- **Willpower 만 의 의존**: 매 ego depletion + decision fatigue — 매 environment design.
+- **Multiple new habits 의 동시**: 매 cognitive bandwidth 초과 — 매 1-2 habits 의 sequence.
+- **No cue specification**: 매 vague intention ("eat better") — 매 specific cue + action.
+- **Punishing missed days excessively**: 매 self-shame spiral — 매 "miss once, never twice" rule.
+
+## 🧪 검증 / 중복
+- Verified (Lally et al. 2010 EJSP, Wood "Good Habits, Bad Habits" 2019, Duhigg "Power of Habit", Clear "Atomic Habits", Gollwitzer 1999 meta-analysis).
+- 신뢰도 A.
+
+## 🕓 Changelog
+| 날짜 | 변경 |
+|---|---|
+| 2026-05-08 | Phase 1 |
+| 2026-05-10 | Manual cleanup — habit loop, Lally curve, JITAI, implementation intentions 추가 |