soc-social-network

asgard-ai-platform ｜ 資料與分析 ｜ 原文 ｜ MIT

用中心性和結構洞畫出組織裡誰真正有影響力、資訊卡在哪

---
name: "soc-social-network"
description: "運用社會網絡分析的概念（節點、連結、中心性、結構洞、強弱連結）來描繪並分析關係結構。使用者需要了解組織內的影響力模式、找出關鍵的連結者、分析資訊流動，或描繪利害關係人的關係時使用，就算他們說的是「誰是意見領袖」、「資訊在這裡是怎麼傳開的」或「畫出我們團隊裡的關係」也一樣。"
metadata:
  category: "WP-18 社會科學院"
  tags: ["social-science", "social-network-analysis", "organizational-behavior"]
---

# 社會網絡分析

## 概述

社會網絡分析檢視行動者（節點）之間的關係（連結），揭露組織圖上看不見的結構。它能找出誰真正握有影響力、資訊瓶頸在哪裡，以及想法如何在社群中擴散。

## 框架

```
IRON LAW: Structure Determines Influence, Not Just Position

A mid-level manager who bridges two disconnected departments may have more
real influence than a VP who sits in a dense, well-connected cluster.
Network position (centrality, brokerage) determines influence more than
formal hierarchy.
```

### 核心概念

| 概念 | 定義 | 為什麼重要 |
|---------|-----------|---------------|
| **節點（Node）** | 一個行動者（人、組織、實體） | 誰在網絡裡 |
| **連結（Tie）** | 節點之間的關係 | 節點是怎麼連在一起的 |
| **強連結（Strong tie）** | 頻繁、有情感、互惠的關係 | 信任、支持、可靠的資訊 |
| **弱連結（Weak tie）**（Granovetter） | 不頻繁、隨意、橋接的關係 | 取得「新」資訊和機會的管道 |
| **度中心性（Degree centrality）** | 直接連結的數量 | 受歡迎程度、活躍度 |
| **中介中心性（Betweenness centrality）** | 一個節點出現在其他節點之間最短路徑上的頻率 | 經紀角色、把關、資訊控制 |
| **接近中心性（Closeness centrality）** | 到所有其他節點的平均距離 | 資訊觸及的速度 |
| **結構洞（Structural hole）**（Burt） | 兩個群集之間的缺口，由一位經紀人搭橋 | 競爭優勢的來源：經紀人控制資訊流動 |

### 分析步驟

1. **定義網絡**：節點是誰？什麼算是連結？（溝通、信任、諮詢、合作）
2. **蒐集資料**：問卷（「你會找誰討論意見？」）、email／Slack 資料、會議共同出席
3. **描繪網絡**：把節點和連結視覺化
4. **計算中心性指標**：每個節點的度、中介、接近中心性
5. **找出結構模式**：群集、橋接、孤立者、結構洞
6. **詮釋並轉為行動**：誰是關鍵連結者？瓶頸在哪裡？

## 輸出格式

```markdown
# Network Analysis: {Context}

## Network Definition
- Nodes: {who} (N = {count})
- Tie definition: {what constitutes a connection}
- Data source: {survey / communication data / observation}

## Key Metrics
| Node | Degree | Betweenness | Role |
|------|--------|-------------|------|
| {person} | {N connections} | {score} | Hub / Bridge / Isolate |

## Structural Findings
- Clusters: {identified groups}
- Bridges: {who connects clusters}
- Structural holes: {where gaps exist}
- Isolates: {disconnected nodes}

## Implications
1. {finding → action}
```

## 範例

### 正確的應用
**情境：** 一家 50 人新創的諮詢網絡
- 中介中心性最高的節點：產品經理（不是 CEO），她橋接了工程、設計和商務團隊
- 結構洞：行銷團隊跟工程完全沒有直接連結，所有溝通都經過產品經理
- 影響：如果產品經理離職，三個團隊之間的資訊流動就會崩潰。需要建立直接的跨職能連結 ✓

### 錯誤的應用
- 「CEO 的連結最多，所以他最有影響力」→ CEO 的度中心性高（連結多），但中介中心性可能很低（每個人彼此之間也有連結，不需要經過 CEO）。違反鐵律：決定影響力的是結構，不只是連結的數量。

## 注意事項

- **Granovetter 的弱連結的力量**：弱連結對於取得新資訊和機會「更」有價值，因為它們橋接了不同的社交圈。強連結分享的是重複的資訊。
- **網絡資料很敏感**：畫出誰跟誰說話，可能讓人覺得像監控。要對目的保持透明，並盡可能匿名化。
- **網絡會改變**：關係會演變。網絡圖只是一張快照。要定期重新量測。
- **中心性依賴情境**：在諮詢網絡裡中心性高，不等於在友誼網絡裡中心性高。要小心定義連結的類型。
- **不要把相關當成因果**：中心的人可能因為所處的位置而表現更好，也可能因為表現好而變成中心。要分清楚很困難。

## 參考資料

- 網絡視覺化的工具與方法請看 `references/network-tools.md`

---
name: "soc-social-network"
description: "Apply social network analysis concepts including nodes, ties, centrality, structural holes, and strong/weak ties to map and analyze relationship structures. Use this skill when the user needs to understand influence patterns in an organization, identify key connectors, analyze information flow, or map stakeholder relationships — even if they say 'who are the influencers', 'how does information spread here', or 'map the relationships in our team'."
metadata:
  category: "WP-18 社會科學院"
  tags: ["social-science", "social-network-analysis", "organizational-behavior"]
---

# Social Network Analysis

## Overview

Social network analysis examines relationships (ties) between actors (nodes) to reveal structure invisible in org charts. It identifies who really holds influence, where information bottlenecks exist, and how ideas spread through a community.

## Framework

```
IRON LAW: Structure Determines Influence, Not Just Position

A mid-level manager who bridges two disconnected departments may have more
real influence than a VP who sits in a dense, well-connected cluster.
Network position (centrality, brokerage) determines influence more than
formal hierarchy.
```

### Core Concepts

| Concept | Definition | Why It Matters |
|---------|-----------|---------------|
| **Node** | An actor (person, org, entity) | Who's in the network |
| **Tie** | A relationship between nodes | How nodes are connected |
| **Strong tie** | Frequent, emotional, reciprocal relationship | Trust, support, reliable info |
| **Weak tie** (Granovetter) | Infrequent, casual, bridging relationship | Access to NEW information and opportunities |
| **Degree centrality** | Number of direct connections | Popularity, activity |
| **Betweenness centrality** | How often a node sits on shortest paths between others | Brokerage, gatekeeping, information control |
| **Closeness centrality** | Average distance to all other nodes | Speed of information reach |
| **Structural hole** (Burt) | Gap between two clusters, bridged by a broker | Source of competitive advantage — the broker controls information flow |

### Analysis Steps

1. **Define the network**: Who are the nodes? What constitutes a tie? (communication, trust, advice, collaboration)
2. **Collect data**: Surveys ("who do you go to for advice?"), email/Slack data, meeting co-attendance
3. **Map the network**: Visualize nodes and ties
4. **Calculate centrality metrics**: Degree, betweenness, closeness for each node
5. **Identify structural patterns**: Clusters, bridges, isolates, structural holes
6. **Interpret for action**: Who are the key connectors? Where are the bottlenecks?

## Output Format

```markdown
# Network Analysis: {Context}

## Network Definition
- Nodes: {who} (N = {count})
- Tie definition: {what constitutes a connection}
- Data source: {survey / communication data / observation}

## Key Metrics
| Node | Degree | Betweenness | Role |
|------|--------|-------------|------|
| {person} | {N connections} | {score} | Hub / Bridge / Isolate |

## Structural Findings
- Clusters: {identified groups}
- Bridges: {who connects clusters}
- Structural holes: {where gaps exist}
- Isolates: {disconnected nodes}

## Implications
1. {finding → action}
```

## Examples

### Correct Application
**Scenario:** Advice network in a 50-person startup
- Node with highest betweenness centrality: Product Manager (not the CEO) — she bridges engineering, design, and business teams
- Structural hole: Marketing team has zero direct ties to engineering — all communication goes through PM
- Implication: If PM leaves, information flow between 3 teams collapses. Need to create direct cross-functional ties ✓

### Incorrect Application
- "The CEO has the most connections, so he's the most influential" → CEO has high degree centrality (many ties) but may have low betweenness (everyone also connects to each other without needing the CEO). Violates Iron Law: structure determines influence, not just connection count.

## Gotchas

- **Granovetter's strength of weak ties**: Weak ties are MORE valuable for accessing new information and opportunities because they bridge different social circles. Strong ties share redundant information.
- **Network data is sensitive**: Mapping who talks to whom can feel like surveillance. Be transparent about purpose and anonymize where possible.
- **Networks change**: Relationships evolve. A network map is a snapshot. Remeasure periodically.
- **Centrality is context-dependent**: High centrality in the advice network ≠ high centrality in the friendship network. Define the tie type carefully.
- **Don't confuse correlation with causation**: Central people may perform better because of their position, OR they may be central because they perform well. Disentangling is hard.

## References

- For network visualization tools and methods, see `references/network-tools.md`