专家访谈 ‖ 哥伦比亚大学社会网络分析专家Peter Bearman 教授

snachina

说在前面的话

Peter Bearman教授是哥伦比亚大学Jonathan Cole 社会科学教席教授，美国科学院院士，也是社会网络分析领域的世界知名学者。他师从于社会网络分析方法奠基人之一的Harrison White教授，自上世纪70年代末起，开始使用并研究社会网络分析方法。如今，Peternal已有博士毕业生50余位，其中不乏社会网络分析方法领域的世界级专家。

Bearman教授的研究领域：社会网络、比较历史社会学、社会学理论、质化研究设计。其质化研究专著《Doormen》即将译为中文出版。

我很有幸参加了Peter主持的社会网络分析工作坊，并于课后就社会网络分析方法访问了他，请他为我们简单介绍社会网络分析，畅谈该领域的前沿应用，并推荐些学习此方法的工具书。Bearman教授认为：“社会网络分析是一种认识世界的思维方式。”在他看来，一切皆网络。他和他的团队用社会网络分析历史数据、神话故事、个体行为，甚至是神经系统。在他的研究中，没有什么是不可以用网络的思维囊括的，关键是要学会网络的思维方式。那么，网络的思维方式是什么呢？看看Peter怎么说。（王荣）

1

Q：Peter教授，您好！受我导师王锡苓教授的委托，想就社会科学研究的研究方法问题请教于您，对您做个访谈。我知道您是研究社会网络的专家，请问，您认为什么是社会网络分析？

Peter：社会网络分析是一种认识世界的思维方式。你可以想象这个世界是由各种关系联结在一起的，社会网络分析是一种从概念上思考世界相互联结的方法。社会网络分析由一系列技术和方法组成，以此来描述这些联结，以及将个体与个体、群体与群体、个体与群体连接在一起的结构。这套方法是用来评价这些联系是如何发生的。所以，社会网络分析是由一套概念和一套方法组成的。

Q:这种结构式的思维方式与传统的社会科学研究有何不同？

Peter：我认为，在典型的社会科学研究方法中，人们将世界作为许多独立的个体来进行观察。所以他们觉得可以对抽离出社会联系、家庭、社区、工作、环境等社会关系的个人进行研究。但是社会网络分析可以将个体嵌入他们的环境去了解，去理解他们所处的关系，以及由于与他人的关系而得到的机会。所以说，这两种思维方式还是有比较大的区别，是试图从独立个体的角度去理解世界，还是说从与他人关联的视角去理解我们是谁，我们希望什么，以及可能发生什么。

Q:那这套思维方式是如何运行的呢，能为我们举个例子来说明如何用社会网络的进行思考？

Peter：这是一种结构式的思维方式。比如，如果你对某件事感兴趣，但你却不去研究这件事，而是研究它是如何嵌套在一个更大的世界中的。例如，我们在研究人们描述税收时所使用的词语。但一个词的含义，通常取决于上下文。因此，如果我说“沙发”，既可以理解成是家具的一种，也可以解释为互联网络中第一个发帖的楼主。如果我知道“沙发”同“客厅”同时出现，我就知道是家具的意思。

在很多演讲文本中，具体情况要复杂的多，例如平等、自由等等词语的意思，都要取决于出现在它周围的上下文。因此，具有社会网络分析的思维方式，就是我不会去研究平等是什么意思，而是将其作为一个概念，去看它周围环绕着哪些词。我也不会去研究这个人是谁，这就是在试图将个体从影响他的环境中抽象出来进行研究。我会去看是什么环绕着他们，进而挖掘与这个人相联系的主要的环境层面。因此，我在努力理解环境。但是这个环境不是说任意的环境，而是一个关系的环境。

2

Q：您是从什么时候开始接触这一方法的呢？

Peter：我1981年在哈佛大学读博士，此前大概在1977年，还在读大学，我那时候感兴趣一个社会网络分析的经典问题——控制美国经济的500强企业中，有很多公司的董事会共享相同的成员，那些人左右着不同的公司，我对这种从所有公司董事成员中抽象出来的互易网络（dual network）很感兴趣，这些公司是如何通过这些重叠董事联系在一起的。总的来说，我当时对美国公司的结构很感兴趣。

到哈佛之后，我的导师是Harrison White（哈里森·怀特），他还健在，他确实算得上社会网络分析的奠基人之一了。他发明了早期的区块模型（block modeling），以及一大堆理念和方法，他就是现在我们所谈的社会网络很多理念的创始人。还有很多社会网络分析的名人，比如写作“弱连接的力量”那篇论文的Mark Granovetter（马克·格兰诺维特），也是Harrison的学生之一。Harrison的很多学生都活跃在这一领域。我的博士学生有五十多名，其中很多已经是社会网络分析领域非常重要的学者，有些已经超过了我。

Q：根据您的经历，请为我们简单梳理下社会网络分析方法的发展历程

Peter：社会网络分析从社会计量法的领域开始，这是基于观察一群人，也许是住在村庄里的一群人，或者去参加葬礼的一群人，用点线的方式画出他们之间的关系图谱。这就是从人类学家需要用图表的方式来描述一个群体中的小团体开始的。

这一领域开始没有任何测量方法，从1970年代到1980年代，主要发展的是测量。因此，我们现在可以去思考类似于“势（power）”这样的概念，并去描绘在一个网络中，“势”可能是什么样的。现在，我们甚至可以用不同的测量方式来表述“势”。“势”可以是你在两个人之间的独特性，我们继而称其为中间中心性（betweeness centrality）；“势”也可以就是尊重你的人数，这就是度数中心性（degree centrality）。

另外，人们形成或终止联系，交朋友或结束朋友关系，签订协约到不再签订，这一宏观进程是如何运作的，就要研究这种规则下的网络如何变化。现在，人们尝试从历史的视角看待在特定环境下展开的网络。

3

Q:我们知道您在尝试用分析《圣经》，但论文尚未发表，能先简单介绍下用社会网络分析文本的思路吗？

Peter：最基本的想法是，语义结构是有意义的，就是说文本组织语言的方式是有意义的。而且我们可以去评估这种结构。

Q:随着六度分割理论的流行，小世界理论开始得到人们的注意，但我们知道，小世界理论并不简单，能帮我们讲一讲小世界理论吗？

Peter：每个人都知道我们生活在一个小世界中。其实，小世界就是我们生活的世界是一个巨大的世界，小世界只有在有相当多人的时候才存在。小世界有两个特点，一是平均的连接步数，即随机筛选的两个人之间通过多少人能联系在一起的这个数量，要非常小。而人们又生活在许多高度聚集的社区中。这看起来有点矛盾，其中有一半的连接步数会比较短，而人们又都属于不同的聚类群体。所以，小世界问题就是，这两件事如何同时发生。Duncan Watts（邓肯·沃茨）多年前的工作，为我们展现了这一过程如何同时发生。也使得这一理论成为有趣的事情。

4

Q:您在研究过程中会将其与其他方法结合吗？

Peter：当然。每个人都希望能呈现这个，你希望将网络特点视为独立的变量。而且确实有一些在个体层面的变化。这对结果而言也很重要。因此，人们拿到网络数据，将其转化为个体层面的变量，用一些标准的统计方法来做预测。网络可以适应这些分析。我的大部分工作，都是试图用网络展示些东西。用一些其他方法，我们用空间分析的方法，也用普通的回归分析，比如线性模型。所以，我不认为有哪些方法是我不用的。

Q:那这是属于量化的方法吗？

Peter：这一方法大部分是量化的，当然，也不需要一定这样。我们也做很多混合方法，你收集网络数据，生成数据集，将其可视化。再你不需要用统计的方式去描述那些网络，你也可以用质化的方式去描述，用质化的理解去解释。用社会网络分析，你也可以非常质化。网络数据非常有理论建设性。所以，如果你有网络数据，也有助于你生成更好的理论。我在哥伦比亚大学教授的就是社会理论课程。

Q:如同您的《Doormen》那本书吗？

Peter：我的《Doormen》是一本人类地理分布学的书，我用了一些社会网络分析的理念在里面。关于人们如何找到工作，这是一个网络过程。但是这本书只有一小部分的社会网络。还有很多其他的观察。只有一章是关于人们如何找工作的网络。不同的方法推动了不同的章节。

5

Q:请您为我们推荐些学习社会网络分析方法的参考书

Peter：

1.从理论学习的角度，推荐John Scott的《Social Network Analysis》一书，这也是对于入门者而言较好的一本书。

2.统计的角度，推荐Matthew and Jackson的《Social and Economic Networks》。

3.从方法的角度推荐Wasserman and Faust的《Social Network Analysis》。

Q：最后，非常感谢您接受我的访问！祝您愉快！

访谈者：王荣，系中国传媒大学新闻学院博士研究生；

                          英国伦敦政治经济学院访问学生。

采访原文

PART ONE: BASIC INTRODUCTION

Q: Very nice to meet you Prof. Bearman. My supervisor Prof. Xiling WANG is now organizing a set of interviews on social science research methods aim at helping to introduce more about methods to Chinese college students for free. We all know that you are the world’s expert on social network analysis. It is our honor to have you here tell us something about this field. Thank you very much. First of all, what do you think is social network analysis?

A: well, social network analysis is one level a way of thinking about the world. So it is a way of imagining the world as connected by relations, and that’s a conceptual way of thinking that the world is connected. And then social network analysis is a set of technics and methods to describe those connections and to review the ways in which the structures that bands people together, band people to groups, band groups to each other. The set of methods to review how that happens. So it’s both a set of concepts and a set of methods.

Q: So what’s the difference between this way of thinking with other ways?

A: I think in typical social science approaches, people approach the world as the units of observations who are independent of each other. So they imagine they can learn something about people by abstracting them out of their social (connections), out of their families, out of their communities, out of their work places, out of what setting they are, and asking them questions. Social Network Analysis can learn something about people we have to actually embed them in those settings, to understand the relational precious that are on them, the opportunities that they are given by virtue of other relations with other people. So it’s pretty profound difference between thinking that the way we are to try to understand the world is as if they were independent as versus saying the way we want to understand who we are, what we hope for, what’s possible is my focusing on the relationships that we have with other people.

Q: Can you tell us an example to learn to think like a networker.

A: I think structurally. So it is the same kind of principle if you are interested in something. You don’t study it, you study it in how it is embedded in a larger world. So, for example, let’s take taxes, we have been studying what’s the meaning of one word people use to say taxes. The meaning of a word, what a word means, is depended on words around it. So, if I say, the word “batter”, batter is both someone who hits a ball, and it is also the material that you made pancakes or cakes. Or I say “strike” it is both an accrual throw and an activity that labor dose, they don’t go to work, they strike. So these are obvious examples that the meaning of the word is conditional on the words that around it. If I know that batter is associate with pancake or cakes, I know oh that’s what that means. If I know batter is associate with a ball, I know what that means. Much more complexity in any kind of the speech that meaning of a concept like equality, freedom, and so on is can change depending on what are the other concepts that are around it. So, thinking in terms of network is, I am not going to study what equality means, I am going to see what is around equality, as a concept. I am not going to study what this person is, I am going to see what is around them, that’s always when you are trying to abstract the setting in which is the person is impacted. It is to pry the major level context in which a person is touched. So I am interested in students I want to put them in the universities. I want to understand the universities not understand the students. So I am trying to get the context. But it is not any context, it is a relational context.

PART TWO: HISTRY AND FUTURE

Q: When did you start to use this method?

A: I was a PhD student in 1981, at Harvard University, but even before that in the 1977, I worked as an undergraduate student, I work on a problem that was a very classic social network problem which was, in America we have major companies that control the economy, the top 500 companies of the united states, and each one of those companies has a board of directors, individuals who direct them, and those individual direct multiple companies, and I was interested in the dual network that I rose from the fact that all individual sat on company boards, and you could ask how our companies connected by the overlap of their directors. So I was interested in the structure of the US company.

Q: That is when I first know about the Social Network Analysis, the typical company network example.

A: the structure of the corporation net, yeah, that is quite what I did in 1970s, my work have not been published since I was just an undergraduate, and then, when I went to Harvard I learned more about Social Network Analysis, with the guy named Harrison White who was my tutor, he is still alive, he is really one of the founders of Social Network Analysis. So he invented what we talked about earlier block modeling, and he invented a whole bunch of some other things, so when people think about social networks, he is the father of many of the ideas that we work with. Also many of the people you might have heard of who do Social Network Analysis, like Mark Granovetter, who wrote the paper we called the strength of weak ties, which shows the weak ties, he was a student of Harrison white’s. So many of Harrison white’s student have gone on to field in Social Network Analysis, I think I have cheered 50 doctoral students, some of those people are very important in Social Networks, more than me.

Q: Tell us more about the development of this method.

A: Social Network Analysis starts in a field called sociometry, which is a way just based on just looking at a collection of people, maybe people who lived in a village, or people who attend to funeral, just drawing a picture of their relationships, representing the circles in the relationship there is the edge. So it starts from the idea that anthropologists had to try to graph the clique to describe the community. And it was a field that didn’t have any measurement, so what happened from 1970 to 1980 to the present is, increasing sophistication in measurement, so we are now able to think about a concept like power, and say what would power look like in a network. And we have now measures for different ways in which power could be expressed. Power could be how uniquely between two people you are, and we then call them betweenness centrality; power could just be a share number of people who look up to you, that’s degree centrality. But, one of the things that we have been talking about is how the micro-processes that lead people to form a tie or to dissolve the tie, to make a friendship or end a friendship, to make a sign to treaty or not to sign to a treaty. So those process are understandable with respect to very simple rescission rules, like avoid relationships that are designate, that are contradictory.

PART THREE: HOT TOPICS

Q: You mentioned before that you are currently using this method to analysis the <Bible>. Will you please introduce us some basic thoughts about this research?

A: The basic thought, first is that the semantic structure is meaningful you know that the way in which words are organized in texts is meaningful. And we can review that. That’s the new idea.

Q: Then tell us about the small world theory briefly.

A: Everybody thinks they live in a small world. So, the small world is the world we live in is super large, it’s a world that only exist when there is lots of people. And it has two characters one is that the average path lines, the average number steps between any two randomly select people is very small. And that people live or people are in very clustered communities. These things seem contradictory that half lines will be short and yet people are in clustered communities. So the small world problem is how do those two things happen at the same time. And the work that shows how the happen at the same time was what Duncan Watts did years ago. And that’s what makes it interesting. Like I bet you are 5 steps away from

the present of your country.

Q: They said it is now from 6 steps to 4 steps already.

A: Yeah, so it is getting smaller.  I’m sure that’s probably true if you took two randomly people this would basicly be true. Actually it is interesting now because we have met, you are now 4 steps away from Barack Obama, my wife, my wife’s friend, Obama. And now it’s you, me, my wife, my wife’s friend, Obama.

It is only meaningful if you have hundreds of thousands of nodes. It’s perfect for China.

Q: How about relationship change over time, how can we measure that?

A: Relation changes over time. It looks different patterns on the network. So, the meaning of the relationship is the pattern of the network. One of the ways you can tell if the meaningful of a relationship that change over time is network patterns change. Of course you could also go the other way and say we could look at a relationship like trust and say that the structure of trust is much different than it used be, then we learn something about trust. So networks give you a way of actually capture important changes that we don’t often see.

PART FOUR: PRACTICE QUESTIONS

Q: Do you use other methods in your research?

A: Oh, sure. Everybody wants to be able to show that, you want to be able to treat a network characteristic as an independent variable. And some did vary in the individual level. And that’s important for outcomes. So, people take network data, turn it into individual level or people level variable for an inner regression, use the standard statistical methods to predict something. So, there is no reason that network cannot fit that strategy. Almost all of my work, uses networks try to show something, by some other methods. So, we use spatial methods, spatial the geography, we use ordinary regression, like linear modeling which is a subset of regression. So, I don’t think there is a method that I don’t use.

Q: So it belongs to quantitative method?

A: They are generally quantitative. While, you know, they don’t have to be.

Q: Can this be a mixed method?

A: We do a lot of mixed methods, you collect network data, you generate a plot of that data, you visualize it, you get to see it. You don’t need the statistical way to be able to describe those networks. You can just say, you know you can describe them qualitatively. You can draw your qualitatively understanding to make sense of them. Statistics are just that kind of thing that people use to, make it appears in the argument are better.

Q: Then can this be qualitative?

A: But I think you can be very qualitative with network data. Network data is very theory construction. So, if you have some network data, it helps you to get better theories. The courses I teach in Columbia, I teach social theory. That’s what I teach.

Q: Like your book <Doormen>?

A: That’s anthropography. I use some of the ideas of social network analysis in that book. About how people get jobs, which is a network process. But the book is only little bit of networks. Some other kinds of observations. There are other kinds of, there is a chapter on networks about how people get jobs. A chapter on queueing theory which, queueing like standing in line, there is a chapter on statistic process, so, different methods motivated different chapters.

PART FIVE:  RECOMMENDATIONS

Q. Give us some recommendationS on choosing text books.

A：I think there is a book in 1994 by Wasserman and Faust. Called <social network analysis>. That’s excellent on methods. Until 1994. So all the methods that are developed worked out for 1994, W&F is the best. Then I think after 1994, to 2012, maybe this book is written by Matthew and Jackson, <social network analysis>. Then I think a very simple book for beginners, is John Scott’s book, for beginners. Looks like a good book. I will start with John Scott, and then if I got more other interests I go to other books.

1. Statistical: Matthew and Jackson

2. Theoretical: John Scott

3. Methods: Wasserman and Faust

Q. Thank you for accepting my interview! Thanks a lot!