2nd/10_Wiki/Topics/AI_and_ML/Homeostasis (항상성).md

---
id: wiki-2026-0508-homeostasis-항상성
title: Homeostasis (항상성)
category: 10_Wiki/Topics
status: verified
canonical_id: self
aliases: [homeostasis, 항상성, allostasis, set-point, regulation]
duplicate_of: none
source_trust_level: A
confidence_score: 0.92
verification_status: applied
tags: [biology, physiology, homeostasis, control-systems, allostasis]
raw_sources: []
last_reinforced: 2026-05-10
github_commit: pending
tech_stack:
  language: Biology / Physiology
  applicable_to: [Biology, Cybernetics, Control Theory]
---

# Homeostasis (항상성)

## 매 한 줄
> **"매 internal environment 의 의 의 stable 의 maintain"**. Cannon 1929. 매 negative feedback loop. 매 응용: 매 body temperature, glucose, pH, blood pressure. 매 modern: 매 allostasis (Sterling) — 매 anticipatory regulation.

## 매 핵심

### 매 mechanism
- **Sensor** → **comparator** → **effector** → **feedback**.
- **Negative feedback**: 매 dominant (95%).
- **Positive feedback**: 매 amplification (childbirth, blood clotting).

### 매 example
- **Body temp**: 36.5-37.5°C.
- **Blood glucose**: 70-100 mg/dL.
- **Blood pH**: 7.35-7.45.
- **Osmolality**.
- **Calcium**.

### 매 vs allostasis
- **Homeostasis**: 매 fixed set-point.
- **Allostasis** (Sterling 1988): 매 변화 의 anticipate.

### 매 응용
1. 매 medical diagnostics.
2. 매 control system design.
3. 매 robot adaptation.
4. 매 RL (intrinsic motivation).
5. 매 organization theory.

## 💻 패턴

### Negative feedback (PID)
```python
class PIDController:
    def __init__(self, kp, ki, kd, setpoint):
        self.kp, self.ki, self.kd = kp, ki, kd
        self.setpoint = setpoint
        self.integral = 0; self.prev_error = 0
    
    def update(self, current, dt):
        error = self.setpoint - current
        self.integral += error * dt
        derivative = (error - self.prev_error) / dt
        self.prev_error = error
        return self.kp * error + self.ki * self.integral + self.kd * derivative
```

### Glucose homeostasis (simplified)
```python
def glucose_regulation(glucose, insulin_secretion=True):
    if glucose > 100:
        insulin = (glucose - 100) * 0.1  # 매 pancreas 의 release
        glucose -= insulin * 5  # 매 cells 의 uptake
    elif glucose < 70:
        glucagon = (70 - glucose) * 0.1
        glucose += glucagon * 3  # 매 liver 의 release
    return glucose
```

### Body temperature
```python
def thermoregulation(core_temp):
    if core_temp > 37.5:
        return {'sweat': True, 'vasodilation': True, 'shiver': False}
    if core_temp < 36.5:
        return {'sweat': False, 'vasoconstriction': True, 'shiver': True}
    return {'normal': True}
```

### Set-point + tolerance
```python
class HomeostaticVar:
    def __init__(self, name, set_point, tolerance):
        self.name = name; self.set_point = set_point; self.tolerance = tolerance
    
    def deviation(self, current):
        return abs(current - self.set_point) / self.tolerance
    
    def is_stable(self, current):
        return abs(current - self.set_point) <= self.tolerance
```

### Allostasis (anticipatory)
```python
def allostatic_adjust(predicted_demand, current_state):
    """매 매 demand 의 의 의 의 의 의 adjust."""
    # 매 e.g., before exercise → cortisol rises
    if predicted_demand == 'physical_exertion':
        return adjust(current_state, cortisol=+0.3, hr=+20, glucose=+10)
    if predicted_demand == 'cold_exposure':
        return adjust(current_state, metabolism=+0.2, thyroid=+0.1)
    return current_state
```

### Robot adaptive control
```python
class HomeostatRobot:
    def __init__(self, target_battery=80):
        self.target = target_battery
    
    def step(self, battery, env):
        if battery < self.target * 0.3:
            return 'return_to_charge'
        if battery < self.target * 0.7:
            return 'reduce_power'
        return 'normal'
```

### Intrinsic motivation (RL)
```python
def homeostatic_reward(state, target_state, weights):
    """매 매 deviation 의 의 의 의 의 reward."""
    dev = sum(w * abs(state[k] - target_state[k]) for k, w in weights.items())
    return -dev
```

### Organizational metaphor
```yaml
org_homeostasis:
  set_points:
    - growth_rate: 25%
    - margin: 30%
    - employee_satisfaction: 7/10
  feedback_mechanisms:
    - quarterly_review
    - employee_survey
    - financial_audit
  effectors:
    - hiring / firing
    - investment
    - culture initiative
```

### Allostatic load
```python
def allostatic_load(biomarkers):
    """매 cumulative wear from chronic stress."""
    score = 0
    if biomarkers.cortisol_pm > 8: score += 1
    if biomarkers.crp > 3: score += 1  # 매 inflammation
    if biomarkers.hba1c > 5.7: score += 1  # 매 glucose
    if biomarkers.sbp > 130: score += 1  # 매 hypertension
    if biomarkers.hdl < 40: score += 1  # 매 cholesterol
    return score  # 매 0-5 (more = more wear)
```

## 매 결정 기준
| 상황 | Concept |
|---|---|
| Static target | Homeostasis |
| Anticipatory | Allostasis |
| Engineering | PID controller |
| Biology medicine | Set-point + tolerance |
| Long-term stress | Allostatic load |

**기본값**: 매 fixed-point system = homeostasis (PID). 매 anticipatory = allostasis. 매 chronic = allostatic load monitor.

## 🔗 Graph
- 부모: [[Cybernetics Foundations|Cybernetics]]
- 변형: [[Allostasis]]
- Adjacent: [[Cybernetics Foundations|Cybernetics]] · [[Free-Energy-Principle]]

## 🤖 LLM 활용
**언제**: 매 medical / control. 매 biology.
**언제 X**: 매 simple stateless.

## ❌ 안티패턴
- **Fixed set-point in dynamic env**: 매 allostasis 의 ignore.
- **No allostatic load monitor**: 매 chronic stress invisible.
- **Homeostasis without feedback**: 매 open-loop.

## 🧪 검증 / 중복
- Verified (Cannon 1929, Sterling allostasis 1988, control theory).
- 신뢰도 A.

## 🕓 Changelog
| 날짜 | 변경 |
|---|---|
| 2026-05-08 | Phase 1 |
| 2026-05-10 | Manual cleanup — homeostasis + allostasis + 매 PID / glucose / load code |