📚 Who's Who Of Epstein's Black Book | 爱泼斯坦黑名册中的人物网络分析
Epstein's infamous 'Black Book' is more than a scandalous contact list; it is a complex social network that can be systematically examined using graph theory. In this article, we apply IB Mathematics concepts — particularly from the topic of graph theory and network analysis — to uncover the most central figures, reveal hidden structures, and illustrate how mathematical modelling brings clarity to a web of connections.
爱泼斯坦臭名昭著的“黑名册”不仅仅是一份充满争议的联络名单，更是一个可以用图论系统分析的复杂社交网络。本文将运用IB数学中的图论与网络分析概念，揭示最核心的人物、发现隐藏结构，并展示数学建模如何让一张关系网变得清晰可辨。
1. Networks as Graphs: Modelling the Black Book | 将黑名册建模为图
aph forms the basis for our centrality analysis. We denote the graph as G = (V, E), where V is the set of vertices and E is the set of edges.
社交网络可以表示为一个图：顶点（节点）代表个人，边代表已记录的连接。在黑名册中，如果两个人出现在同一条记录中或通过共同的关联人相连，则他们之间存在一条边。这个无向、无权重的图构成了我们中心性分析的基础。我们记该图为 G = (V, E)，其中 V 是顶点集，E 是边集。
For IB Mathematics, understanding the adjacency matrix A allows us to perform further calculations. The element aij is 1 if individuals i and j are connected, and 0 otherwise. The degree of a vertex v, denoted deg(v), is the number of edges incident to it, which equals the sum of the corresponding row in A.
在IB数学中，理解邻接矩阵 A 有助于我们进行后续计算。若个体 i 与 j 之间有连接，则元素 aij 为 1，否则为 0。顶点 v 的度，记作 deg(v)，是与它相连的边的数目，等于 A 中相应行之和。
2. Degree Centrality: Who Has the Most Contacts? | 度中心性：谁拥有最多的联系人？
Degree centrality is the simplest measure of importance: it counts how many direct connections a person has. The degree centrality for vertex v is defined as CD(v) = deg(v) / (n – ee centrality indicates a person who appears alongside many others, suggesting a broadly networked individual.
度中心性是衡量重要程度最简单的指标：它统计一个人有多少直接联系。顶点 v 的度中心性定义为 CD(v) = deg(v) / (n – 1)，其中 n 是顶点总数。这种归一化使得不同规模的网络之间可以进行比较。在黑名册中，高度中心性意味着此人多次与他人共同出现，表明其拥有广泛的社交网络。
For example, if a person X has 15 direct links in a network of 100 people, CD rality is easy to compute, it only captures local influence and misses those who connect different subgroups. Thus we need more sophisticated measures to answer the question of 'who's who' in the hidden structure.
例如，某人 X 在一个100人的网络中有15条直接联系，那么 CD(X) = 15/99 ≈ 0.152。虽然度中心性易于计算，但它只能捕捉局部影响力，而无法识别连接不同子群的人物。因此，我们需要更复杂的衡量方式来回答黑名册中“谁是谁”的结构性问题。
3. Closeness Centrality: Reaching Everyone Quickly | 紧密中心性：快速触达每一个人
Closeness centrality measures how close a vertex is to all other vertices in the network, based on the shortest paths. The closeness centrality of vertex v is given by CC(v) = (n – 1) / Σu∈V d(v, u), where d(v, u) is the shortest distance (number of edges) between v and u. A person with high closeness centrality can spread information efficiently through the entire graph.
紧密中心性基于最短路径衡量一个顶点与网络中所有其他顶点的接近程度。顶点 v 的紧密中心性公式为 CC(v) = (n – 1) / Σu∈V d(v, u)，其中 d(v, u) 是 v 与 u 之间的最短距离（边的数量）。紧密中心性高的人能够高效地在整个图中传播信息。
Think of a graph where some people are clustered tightly together while others sit on the periphery. Closeness centrality highlights those who are not necessarily the most connected, but are strategically positioned so that the total distance to everyone else is minimised. In Epstein's network, this could identify intermediaries who provide rapid access to many individuals.
试想一个图中有些人紧密聚集，而另一些人处于外围。紧密中心性突出那些不一定是连接最多，但处于策略位置、到所有人总距离最小的人。在爱泼斯坦的网络中，这可以识别出能够快速触达众多人物的中间人。
4. Betweenness Centrality: The Power of the Bridge | 中介中心性：桥梁的力量
Betweenness centrality captures the extent to which a vertex lies on the shortest paths between other vertices. It is defined as CB(v) = Σs≠v≠t σst(v) / σst, where σst is the total number of shortest paths from s to t, and σst(v) is the number of those paths that pass through v. This measure reveals gatekeepers who control the flow of connections between different communities.
中介中心性反映了一个顶点位于其他顶点之间最短路径上的程度。其定义为 CB(v) = Σs≠v≠t σst(v) / σst，其中 σst 是从 s 到 t 的最短路径总数，而 σst(v) 是其中经过 v 的路径数。该指标揭示了控制不同社群间连接流的把关人。
In the context of the Black Book, a person with high betweenness centrality would be essential for connecting otherwise isolated groups. Removing such a vertex could fragment the network into disconnected components. Mathematically, computing betweenness requires Floyd–Warshall or Brandes' algorithm, which are within the scope of IB HL graph theory investigations.
在黑名册的语境中，中介中心性高的人对于连接原本隔绝的群体至关重要。如果移除这样的顶点，网络可能分裂为互不连通的组件。从数学上讲，计算中介中心性需要用到 Floyd–Warshall 或 Brandes 算法，这些都属于 IB HL 图论探究的范畴。
5. Eigenvector Centrality and PageRank: Influence of Influential Neighbours | 特征向量中心性与PageRank：被重要人物关联的重要性
λx, where the centrality of vertex i is proportional to the sum of centralities of its neighbours. A variant known as PageRank introduces a damping factor to model random walks.
特征向量中心性基于连接高得分顶点的贡献更大这一原则来分配相对分数。它由邻接矩阵的主特征向量计算得出：Ax = λx，其中顶点 i 的中心性与其邻居中心性的总和成正比。一个称为 PageRank 的变体引入了阻尼因子来模拟随机游走。
PR(v) = (1 – d) / n + d × Σu∈M(v) PR(u) / L(u)
Here d is the damping factor (typically 0.85), M(v) is the set of pages linking to v, and L(u) is the number of outbound links from u. Applying PageRank to the Black Book network helps reveal figures who, even if they have few direct contacts, are linked by very influential people — a subtle but powerful indicator of hidden importance.
这里 d 是阻尼因子（通常取 0.85），M(v) 是链接到 v 的节点集，L(u) 是 u 的出链数量。将 PageRank 应用于黑名册网络有助于发现那些即使直接联系人不多，却被极具影响力的人所关联的人物——这是隐藏重要性的微妙而有力的指标。
6. Worked Example: A Subgraph from the Black Book | 实例演示：黑名册的局部子图分析
Let us analyse a simplified subgraph of 10 individuals labelled A to J. Their connections are abstracted from common appearances in the book. The adjacency relationships are: A connects to B, C, D; B connects to A, E; C connects to A, D, F; D connects to A, C, G; E connects to B, F, H; F connects to C, E, G, I; G connects to D, F, J; H connects to E; I connects to F; J connects to G. This small network allows manual calculation of all centralities.
让我们分析一个简化的10人子图，标签为 A 到 J，其连接关系从书中共同出现的情况抽象而来。邻接关系如下：A 连接 B、C、D；B 连接 A、E；C 连接 A、D、F；D 连接 A、C、G；E 连接 B、F、H；F 连接 C、E、G、I；G 连接 D、F、J；H 连接 E；I 连接 F；J 连接 G。这个小网络便于我们手工计算所有中心性指标。
By constructing the adjacency matrix and computing shortest paths, we obtain the following centrality values (normalised where applicable):
通过构建邻接矩阵并计算最短路径，我们得到以下中心性数值（适用时已归一化）：
| Vertex | Degree (CD) | Closeness (CC) | Betweenness (CB) | Eigenvector |
|---|---|---|---|---|
| A | 3/9 ≈ 0.333 | 9/14 ≈ 0.643 | 0.254 | 0.437 |
| F | 4/9 ≈ 0.444 | 9/13 ≈ 0.692 | 0.318 | 0.512 |
| C | 3/9 ≈ 0.333 | 9/15 ≈ 0.600 | 0.197 | 0.401 |
| G | 3/9 ≈ 0.333 | 9/17 ≈ 0.529 | 0.164 | 0.298 |
| E | 3/9 ≈ 0.333 | 9/16 ≈ 0.563 | 0.225 | 0.366 |
From the table, vertex F emerges as the most central across all measures: it has the highest degree, closeness, betweenness, and eigenvector centrality. This means F is not only widely connected but also lies on critical shortest paths and is connected to other central actors. In a real network, F would be identified as a 'key player'.
从表中可以看出，顶点 F 在所有指标上都最为突出：它拥有最高的度数、紧密值、中介值和特征向量中心性。这意味着 F 不仅联系广泛，还处于关键的最短路径上，并与其他中心行动者相连。在真实网络中，F 将被识别为“关键角色”。
7. Community Detection: Uncovering Hidden Groups | 社团发现：挖掘隐藏群体
on. A popular method is the Louvain algorithm, which maximises the modularity Q = (1/2m) Σij [aij – kikj/(2m)] δ(ci, cj), where m is the number of edges, ki is the degree of vertex i, and δ is 1 if i and j are in the same community. Groups with high internal density reveal clusters of closely associated individuals.
除了个体中心性之外，我们还可以利用模块度最大化将黑名册图划分为社团。流行的方法之一是 Louvain 算法，它最大化模块度 Q = (1/2m) Σij [aij – kikj/(2m)] δ(ci, cj)，其中 m 是边数，ki 是顶点 i 的度数，若 i 与 j 在同一社团则 δ 为 1。内部连接紧密的群体会揭示出彼此高度关联的个体组成的圈子。
In the subgraph above, we can visually identify two main communities: {A, C, D, F, G} and {B, E, H, I, J} with F and E acting as bridges. Mathematically, community detection validates that certain individuals, despite not having the highest overall centrality, are vital for inter-community links — a role often overlooked by simpler measures.
在上面的子图中，我们可以直观地识别出两个主要社团：{A, C, D, F, G} 和 {B, E, H, I, J}，其中 F 和 E 充当桥梁。从数学上讲，社团检测验证了某些人尽管总体中心性不是最高，但却是社团间连接不可或缺的一环——这一角色常被简单的度量所忽略。
8. Visualising the Network: Graph Theoretic Drawings | 网络可视化：图论绘图
Using force-directed layouts such as Fruchterman–Reingold, we can position vertices so that connected nodes are pulled together while disconnected ones repel. This produces a visual map where highly central vertices naturally gravitate to the core. Students can reproduce such visualisations with software like GeoGebra or Python's NetworkX, reinforcing their understanding of graph algorithms.
利用力导向布局（例如 Fruchterman–Reingold 算法），我们可以将顶点放置在合适的位置，使得相连的节点相互吸引，而不相连的节点相互排斥。这样生成的可视化地图会让中心性高的顶点自然地聚集到核心位置。学生可以使用 GeoGebra 或 Python 的 NetworkX 重现此类可视化，从而加深对图算法的理解。
In the Black Book context, a central dense cluster with several peripheral nodes would indicate a core group of tightly connected individuals surrounded by less integrated associates. Such patterns are common in real-world social networks and can be quantitatively expressed through graph density and average path length.
在黑名册中，一个中心密集的聚类带着若干个外围节点的结构，表明存在一个由紧密关联个体组成的核心群体，周围环绕着很少融合的关联者。这种模式在现实社交网络中十分常见，并可通过图的密度和平均路径长度进行量化表达。
9. Comparing Centrality Rankings: Who's Really Who? | 中心性排名比较：究竟谁才是关键人物？
No single centrality measure gives a complete picture. A person who tops the degree ranking may not be the best intermediary, while a high betweenness vertex may have low degree. We therefore compute a composite score using z-score standardisation: for each vertex, we sum the normalised values across all four centralities. This multi-criteria approach reduces bias and provides a robust 'who's who' ranking.
没有哪一个中心性度量能够给出完整的图景。在度数排名中位列前茅的人不一定是最佳中间人，而高中介中心性的顶点可能度数很低。因此，我们使用 Z-score 标准化计算综合得分：对每个顶点，将四种中心性的标准化值相加。这种多准则方法可以减少偏差，给出可靠的“谁是谁”排名。
Suppose we apply this to our example. The composite scores (sum of standardised values) are: F 3.82, A 3.15, E 2.74, C 2.53, G 2.10, D 1.97, etc. Thus F is confirmed as the most central figure, followed by A. This quantitative assessment mirrors how investigators might use network science to prioritise individuals of interest.
假设我们将此方法应用于上述案例。综合得分（标准化值之和）为：F 3.82、A 3.15、E 2.74、C 2.53、G 2.10、D 1.97 等。因此 F 被确认为最核心的人物，A 次之。这种量化评估反映了调查人员如何运用网络科学来优先关注某些关键个体。
10. Robustness and Sensitivity: What If Edges Are Missing? | 鲁棒性与敏感性分析：如果缺失了某些连接会怎样？
Real-world data is often incomplete. We can test the robustness of our centrality rankings by randomly removing 10% of edges and recalculating. If the rankings remain stable, confidence in the identified key players increases. The Jaccard similarity index on the top-5 sets before and after perturbation measures stability. A value above 0.8 indicates high robustness.
现实世界的数据往往不完整。我们可以通过随机删除 10% 的边并重新计算，来检验中心性排名的鲁棒性。如果排名保持稳定，则对已识别关键人物的信心就会增强。扰动前后前5名集合的 Jaccard 相似系数可以衡量稳定性，高于 0.8 的值表示鲁棒性高。
In the Black Book case, due to the dense core, removing a few edges seldom alters the top-tier actors. This suggests that while the book may not list every interaction, the central figures are reliably identified. Such sensitivity analysis is a valuable component of an IB Mathematical Exploration (IA) project.
在黑名册的例子中，由于核心层密度高，删除少量边几乎不会改变顶层角色的排名。这表明即使名册未能列出每一次互动，核心人物的识别仍然是可靠的。此类敏感性分析是 IB 数学探索（IA）项目的重要组成部分。
11. Ethical Considerations and Data Limitations | 伦理考量与数据局限
When analysing sensitive real-world networks, ethical use of data is paramount. In this educational context, we emphasise that our graph is a hypothetical model inspired by the concept of a 'contact book', and no personal identities are referenced. All calculations are intended solely to illustrate mathematical techniques. In any actual investigation, privacy, consent, and data accuracy must be rigorously addressed.
在分析敏感的现实世界网络时，数据的伦理使用至关重要。在此教育情境中，我们强调本图是从“联络簿”概念获得启发的假设模型，未涉及任何个人身份信息。所有计算仅旨在说明数学方法。在任何实际调查中，都必须严格处理隐私、同意和数据准确性问题。
Additionally, social network analysis presumes that recorded connections reflect genuine relationships, which may not always be true. Missing connections (false negatives) or spurious entries (false positives) can distort results. A good IB IA would discuss these limitations transparently and suggest improvements like weighted edges or directed graphs.
此外，社交网络分析假设记录的连接反映了真实的关系，但这不一定总是成立。缺失的连接（假阴性）或虚假的记录（假阳性）可能扭曲结果。一份优秀的 IB IA 应当透明地讨论这些局限，并提出改进方法，例如加权边或有向图。
12. Conclusion and Further IB Applications | 总结与进一步的IB应用
Graph theory and network analysis provide a powerful toolkit for making sense of complex relational data like Epstein's Black Book. By combining degree, closeness, betweenness, and eigenvector centralities, we can objectively rank individuals and uncover the hidden architecture of a social network. The mathematical processes — from constructing adjacency matrices to implementing community detection algorithms — align directly with IB syllabus topics and offer rich opportunities for Mathematical Exploration.
图论和网络分析为理解像爱泼斯坦黑名册这样的复杂关系数据提供了强有力的工具。通过结合度数、紧密程度、中介性和特征向量中心性，我们可以客观地对个体进行排序，并揭示社交网络的隐藏结构。从构建邻接矩阵到实现社团检测算法的数学过程，与 IB 课程大纲主题直接吻合，并为数学探索提供了丰富的机会。
Students are encouraged to replicate this analysis with anonymised datasets, extend the model to directed or weighted graphs, or compare centrality profiles across different network types. Such investigations not only deepen mathematical understanding but also cultivate critical thinking about the role of mathematics in real-world investigations.
鼓励学生使用匿名数据集复现本分析，将模型扩展到有向图或加权图，或比较不同网络类型的中心性特征。这样的探索不仅能加深对数学的理解，还能培养关于数学在现实世界调查中作用的批判性思维。
Published by TutorHao | IB Mathematics Revision Series | aleveler.com
更多咨询请联系16621398022（同微信）
或截图后用「扫一扫 → 相册」选择本图