Data Science

Cyber Anthropology: A Unified View of Communities and Social Networks

Dr Michael WuMichael Wu, Ph.D. is 927iC9C1FD6224627807Lithium's Principal Scientist of Analytics, digging into the complex dynamics of social interaction and group behavior in online communities and social networks.


Michael was voted a 2010 Influential Leader by CRM Magazine for his work on predictive social analytics and its application to Social CRM.He's a regular blogger on the Lithosphere's Building Community blog and previously wrote in the Analytic Science blog. You can follow him on Twitter at mich8elwu.



Welcome back, and thank you for being patient with me during my relocation to Lithium's new office in San Francisco. It has been a very frantic week for me. After several weeks of digression, we can finally revisit the topic of cyber anthropology – the social anthropology of online communities and social networks. I’ve previously written about the difference between communities and social networks and their distinct roles in building and maintaining relationships. In those posts we examined the structure of communities and social networks in isolation. In reality, however, they coexist and overlap each other – at any given time we are part of many communities, and are connected to people in these communities as well as those beyond our communities.


Today I will give you a unified view of communities and social networks in an attempt to understand how these two social structures fit together. This post is a little more abstract and theoretical in nature – in comparison to previous posts – but I believe it will give you a novel perspective on the inner workings of social media.


The Social Networks inside Communities

1045iB3D4AEE83074DE6BAs we’ve learned, most of the strong relationships in our personal social network (a.k.a. personal network) were developed in communities that were once part of our lives. Figure 1 illustrates the dynamics of a person (hypothetically Bob, the red dot in Figure 1) joining a community (yellow).


From personal experience, you probably recognize the fact that when someone joins a community, the rest of his personal network does not necessarily join with him. That is, your friends generally do not follow you everywhere, at least not immediately, and not all of them. Why is this? The main reason is because people have different interests and they have communities of their own. Upon joining the community and interacting with other community members, Bob would create weak ties (dotted lines). If the interactions were mutually desirable and have the proper environment to develop over time, then some of these emerging weak ties will grow into strong relationships, which will ultimately become part of Bob’s personal social network.


As you can see, it is definitely possible to have social networks within a community. In fact, social networks will develop naturally in communities that provide a rich medium for their members engage and interact. Members who have invested much time and effort in the community will tend to have a more extensive network (e.g. the orange dots), whereas new members (e.g. the red dot) will tend to have fewer connections. But this is just one part of Bob’s social network (i.e. the localized social network inside this particular community).


1047i837BE1F5ECECC83AIn reality, the picture is more complicated. Because Bob, like most people, doesn’t just belong to one community. Most people are part of several communities at any given time. Figure 2 illustrates this situation, where Bob is actually part of three other communities (blue, green, and orange). Note that within each of these communities, there are actually complex network structures just like the yellow community, but we’ve kept these redundant details hidden.


The Communities inside Social Networks

Although the entire social network covers the globe, your personal network (which includes only your direct connections) is more limited in scope. This means we can examine Bob’s network in greater detail. Since this section focuses on Bob’s personal network, we will hide all the relationships that are not directly connected to Bob. As we discussed in the previous section, Bob has just joined the yellow community, he didn’t know anyone in this community before and hasn’t established any long term relationship yet. So if we hide the details and only look at Bob’s personal network, he actually has no connections in the yellow community.


1049iCB5BE38B722913D1However, Bob is already a member of three communities, so we can examine the relationships that Bob has established in these communities. Figure 3 shows all of Bob’s friends and the communities they belong to (the black lines indicate direct connections to Bob). It is not surprising to see that Bob has friends in the blue, green and orange community. Moreover, one of his friends (the blue dot) shares two common communities (blue and orange) with Bob. Notice that Bob also has friends that do not share any communities with him. These friends are in the pink, violet, and cyan community respectively. Where do these friends come from?


Remember, as we grow and our lives change, it is natural for people to move away from a community and move into another. Since we maintain the relationships we established in communities with our social network, it is therefore logical to find relationships across community boundaries. This means sometime in the past, either Bob has left the pink, violet, and cyan communities, or his friends have left their common communities and then subsequently joined the pink, violet, and cyan communities.


Now that we understand the community structure around Bob, we can begin to fill in the network structures. Figure 3 is incomplete because it doesn’t show how Bob’s friends are interconnected among themselves. If none of Bob’s friend knew each other, then this picture would be correct. However, this is unrealistic, because the social dynamics (such as the triadic closure rule, homophily, etc) that govern tie formation tends to create cliques among friends.


1051i5F0FA43D63017D18This brings us to Figure 4, which shows the structure of Bob’s personal social network, and how it fits with the community structure around him. The black lines are Bob's direct relationships, and the gray lines are connections among Bob’s friends. Together they make up Bob's personal network. What do you see inside this network? Do you see communities? Not only will you find communities within personal networks, communities exist inside the global social network too. This will be more obvious if you recognize the fact that the global social network includes everyone in the world. Consequently, any community consisting of a subset of the world’s population must lie inside this global network.



So what have we learned? Although the topological structure of communities and social networks are very different, and each social structure has its unique functions in human history, they are tightly interwoven. In fact they complement each other.

  1. Social networks develops naturally inside communities (e.g. Friends on Yelp)
  2. Communities can (and often do) exist inside social networks (e.g. fan pages and groups on Facebook)

We have examined the social structure around a hypothetical person, Bob, and it looks something like Figure 4. But this is only for one person. What is not shown in this picture is that everyone of Bob’s friend has a similar social structure around him/her. This also applies to all the people inside the communities that weren’t shown, because they weren’t directly connected to Bob. In fact, these social structures exist around every single human being on earth. The complexity is truly perplexing, yet beautiful.


Alright, now that we (hopefully) have a more holistic understanding of community and social network, as well as how they fit inside each other, we can begin applying it. Next time, we will use this knowledge to analyze an interesting social media platform – Twitter. But, before that, maybe we can have some discussions around this topic. Comments of any kind are always welcome, so keep them coming.




good discussion on building social network within a community or a community with a social network. actually, both of cases have deep meanings: social network within a community means that we can have certain-type of 'closeness/friendship' build on top of the professional/interests network and community within a network means that we can turn friendship into professional/interests relations.

The diagrams say it all. A great way to explain this concept.


I was particularly interested in knowing the influence of trend-setters on this. From what i have noticed, individuals tend to follow their friends not just within a particular network, but also from one network platform to the other (for example, when i started with the social media, Orkut was pretty big in India. Today i am more active on Facebook. Reason: My friends whom i looked upon as tech savvy, migrated to fb). 


From the perspective of the diagrams shown, if we say that the trend-setters usually are the focal point where most of these communities overlap, would that be right?



Do you have anything that charts the nature of online friendships that charts influence in a community
Data Science

Hello Dr Chen,


Thank you for stopping by Lithosphere and commenting on my blog.


This is a great comment to this somewhat abstract and theortical post. I just like to rephrase the 2nd point you mentioned.

1. People with similar interests and professional groups can turn into long term friendships.

2. You can find groups among your friends with similar interests and professions, which form cliques in your personal social network.


Great observation!

Thanks again for the excellent comment.


Data Science

Hello Nirmal,


Thank you for the question.


To answer your question, that really depends on what you mean by trend-setter, because trend setter can have many meanings. From the context in your question, I think trend-setters means adopters of technologies who like to explore new networks, right? Certainly this one kind of trend-setter, but I would not generalize my comment here to other types. To avoid future confusion, I am going to call them network explorers for the time being, and I think from my answer below you will understand why.


What I've shown is the overlap of communities within a social network, but I think what you were asking is the overlap between different social network platforms. Keep in mind that the different social network platforms is an artificial separation of our social network. Whether your friends are on Orkut or FB, they are still your friends, and they are part of your personal social network. With that in mind, I think that the network explorers are probably at the overlap between different social netowrk platforms (if we view different network users as a macroscopic communities). However, they do not necessarily have to be trend-setters. It is certainly possible that these people move to another network created some ties in the new network, but found it not as good as the old one, then move back. That being said, I believe that explorers, early adopters, who tend to explore other networks and communities will tend to be at the overlap between communities (as shown in my diagram). That is a direct consequence of the fact that these explorers tend to have much movement between communities/networks.


Alright, I hope I've address your question. Thanks for asking it and I hope to see you around on Lithosphere again.


Data Science

Hello Peeyoosh,


I've written quite a bit on influence in communities, and you can find them through the links provided in this summary blog article.

     My Chapter on Influencers


Thank you for the comment.


I like the term "network explorers" . . . could this idea be linked to a computational "curiosity" metric?  Any ideas?

Data Science

Hello Colleenp


Thank you for asking the question.


I think we can use it as a proxy for curiosity, but I wouldn't just take that literally. Many people often play with terms and they can be very misleading.


For example, if the metric is a measure of network overlap, it means network overlap, and that is it. Although network overlap can be a result of people who are curious and like to explore different networks/communities, it doesn't means that network overlap = curiosity. For example another user can have high network overlap due to the fact that he's a traveling salesman, and his work requires him to move from place to place, therefore switch between many communities/networks. But he is totally not curious at all. In fact he hates exploring. In this case, he will have a high network overlap metric, but he may be totally not curious at all. His high network overlap is simply a result of the fact that his work require much traveling.


So one danger about interpreting data is don't try to over-generalize the data. A metrics only mean what it is designed to measure, and that is it. We should not over-interpret it, or else it could lead you to a totally wrong conclusion.


However, we do compute a interest focus metric (bease on the participation entropy) in our communities. It measure how focus a person's participation is. I think that is a much better proxy for curiosity, although it is still not equal to curiosity. It tells how focus a user's interest is. To measure curiosity, we would have to design a metric that specifically measures that. It may use a combination of network overlap, interest focus, and more, because there are many behavioral traits that are a result of curious behaviors.


I hope this addresses your question. Thanks again for this great question. See you around.



Soy Boliviano y uno de los pocos antropologos especializados en Redes Sociales Virtuales, tengo un blog donde escribo diatribas sobre el tema. (

El análisis me pareció fundamental, en especial por que se basa fuertemente en las estructuras de redes que no son extrañas para la antropologías.

Me quedo con una pregunta ¿Los antropologos están listos para analizar las modificaciones de las conductas pautales en este nuevo ecosistema?


Data Science

Hello Jean,


Hmm, unfortunately Spanish is not my native language. So I've used Google Translate to help me. For those who can't read Spanish (like me) here is what Jean said according to Google.



I am Bolivian and one of the few anthropologists who specialize in Virtual Social Networks, I have a blog where I write diatribes on the subject. (Http://

The analysis seemed essential, in particular that relies heavily on network structures that are alien to anthropology.

I'll take one question: Anthropologists are ready to analyze the behavior pautal changes in this new ecosystem?



Now my response.


I think that anthropologist are ready. To understand social media, we need a collaboration between many disciplines. I've used everything from statistics, machine learning, data mining, social network analysis, graph analysis, econometrics, sociology, social psychology, and social anthropology to understand the "social" and user behavioral aspects of social media. It really spans everything from the really rigorous mathematics and statistics, to the softer side of science. Social is not just a challenge of one discipline, it is really every aspects of our life. So I think it is important to look at the problem from multiple lens. If you are interested in the basics of social network analysis. I've written a short blog article on that: Social Network Analysis 101.


Although it is in Spanish, I still appreciate your comment Smiley Happy

I presume you are using the term social network and community for both online and offline relationships.

Data Science

Hello Larry,


Yes, social network and community are both very robust social structures that existed thousands of years before there are even computers. The idea here is to see if we can get a better understand of the "social" aspect of social media using principles from social anthroplogy, social psychology, social economics etc. Because it is these social aspects of social media that makes social media work. Technology is only an enabler of what the social things that people normally do for ages.


If you have not read the earlier articles in this series, I recommend checking them out:

  1. Community vs. Social Network
  2. How Do People Become Connected?
  3. From Weak Ties to Strong Ties
  4. Maintaining the Strong Ties

I'm sure that if you read the prequel to this article, things will make a lot more sense. But thanks for asking the clarification question, and hope to see you around Lithosphere next time.


Hi Michael,


As every marketer would like to know who are the influencers to best spreed out the message and to attract more users/customers, the next task would be: Find and Engage the influencers (they are from the intersection of friend network and the professional network), and let them to market your brand/products/services. That reads, there is a need to find out the answers for the questions: the quality of the influencers friend network & professional networks (against your business target). from business point of view, you wand influencers to' develop' your users/customers from BOTH of their friend network and professional networks. Then there is an question a marketer has to find out: why the influencers 'works' for him...and this is key to social CRM from business point of view.

one more thing, from my facebook experiences, facebook provides more automatic notification tools for friend networks, NOT PAGES. And this is also a reason we need to blend friends network and professional network.

Data Science

Hello Dr Chen,


Great comment.


Finding influencers is definitely the first step and I've created algorithms and blogged about this topic earlier (See My Chapter on Influencers). Engaging the influencers to co-create value is a much longer term problem that involves strategies, business processes and a change in corporate culture. I've also speak about this topic at webinars and panels, even though I have not blog extensively about it.


Base on my simple 6-factor influence model, we must take into account of the targets (people we want to influence) when looking for influencers. These targets are precisely the personal network of the influencers. Remember, the friendship network and professional network are both part of someone's personal social network. The division between these two network is quite artificial, and is only a result of the inability of social networking platforms to assign and keep track of the different types of relationships we have (See Social Network Analysis 101).


One way to design an algorithms that identifies influencers would be to filter out the irrelevant targets first, provided that we know something about the target. Then we can look for the influencers among the relevant targets. This will ensure that the targets are matched to the business and marketers would know why the particular identified influencers works.


I hope this address your commnet concerning why a particular influencer works for a particular target audience. Thanks for commenting and hope to see you next time.


Hi Michael,


As someone with an anthropology education I'm pleased that the world has stumbled upon a value to our discpline.  A relief that my student loans weren't in vain.


My add to this: I've come to think of social media as a real-time annotated Rolodex.  Many say that they use social media to keep in touch with people they know IRL  Sometimes with people they met on other platforms.  Hardly ever do I hear that people meet and become close cross-network without meeting in person.  Thus we become voyeurs of other people's network interactions.


Communities allow us to engage with strangers without the implication that we must be introduced by a common network connection, such as on Linked-in.  And those relationships are unlikely to expand outside the Community unless they are bolstered with an IRL engagement.


Until online engagement approximates IRL engagement of people in related networks, with real emotional bonding of some kind, online remains a model of IRL engagement, merely supporting IRL.  Maybe it's better that way - don't know yet.


thanks, Michael for addressing my question on the influencers. your suggestionwould be a feasible approach (i would call those people the 'Conditional Influencer'): 

One way to design an algorithms that identifies influencers would be to filter out the irrelevant targets first, provided that we know something about the target. Then we can look for the influencers among the relevant targets. This will ensure that the targets are matched to the business and marketers would know why the particular identified influencers works.

Data Science

Hello Maggid,


Glad to know a real anthropologist by training. I believe that the softer science (sociology, social anthropology, social psychology, social economics) will definitely be valuable in terms of understanding the social aspect of social media. As I mentioned in my reply to Jean social is inherently a interdisciplinary subject. We need both the hard core rigorous science (mathematics, machine learning, statistics, social network analysis, etc.) as well as the softer science to understand human behavior.


I believe that (in-real-life) IRL interaction and talking to people face to face still has its value in building strong relationships. Online communication platforms gave us a much wider reach, but also limited us in the depth of interaction. However, I would argue that it is possible to develop strong relationships if people spent enough time together and there is enough reciprocity between them. I've definitely get to know quite a few members on the lithosphere community quite well. Eventually, we linked up on LinkedIn, and I've even met some of them IRL. So, even though I agree that it is much harder to build close and strong relationships online than IRL, I think it is definitely possible.


As we learn more about human behavior, psychology and needs, we may be able to develop a platform that enables deeper level of interaction, which could then make online relationship building easier. That would require a fusion between the softer sciences with the rigorous science.


Thank you for the comment, and hope to see you next time.


Data Science

Hello Dr Chen,


Welcome back. Great term. I can see your probabilistic modeling expertise leaking out from this term.


However, the set of influencers may be outside the set of potential targets. When filtering out the irrelevant targets, we only get the set of potential targets. The set of conditional influencers, as you call it, is the set of all 1st degree connection to set of all potential targets. The real influencers are within this larger group. That's what I ment when I say finding influencers among them (sorry for not being clear).


Thanks for the comment and see you around.


Hello Dr Chen,


is, sociologicaly spoken, 3 types of social group (small, medium and big) more pointed to analyse web sociation and influencing?

Michael -


Thanks for this useful description of the interplay between Communities we participate in and the Social Networks we have as individuals.  


I'm intrigued by the conversation - and your related posts - around influence.  Recently, I've been wondering about the various roles played by individuals within communities - and how roles - and their respective "behaviors", interact to impact the stability and integrity of a community.   In that sense, I think "influencer" would be less of a role and more of a descriptor of community-building efficacy ... meaning that you can become an influencer for a given target audience by fulfilling any number of more specific roles (e.g. moderator, aggregator, networker, etc.).   Does that sound right to you?


How do others feel about the importance of understanding roles within a community?   Is such an understanding useful and actionable?   For example, if we could identify pivotal roles in certain community "types" that must be filled effectively, then this might lead to some sort of community health or maturity metric.    Would love to hear more about the role of roles in community at some point. 


- Skip

Data Science

Hello Skip,


Haven't seen you for a while. Welcome back.


Roles in community is also a feature that arise quite naturally. Because members have a life cycle in the community, more senior members tend to take on some functional roles because they are passionate about certain common goal in the community.But role and influence are not really the same. Althugh I've found that influencers tend to correlate highly with superusers (product experts in the community). This is not surprising because domain credibility is a necessary (althought not sufficient) factor for being influential.


What I've found from analyzing our community data is that depending on the types of communities (e.g. support, enthusiast, innovation community, etc.) different roles are important. For example, in support communities, the product expert role is extremely important, whereas in a enthusiast community, the discussion leader roles is more critical, and finally, in an innovation community, the critics roles is essential. However, in a healthy community, we usually need a mix of different roles.


The understanding of roles is definitely actionable. For example, if you are a support community, you want to have more product experts. But if all you've got is critics and discussion leaders, you may want to focus on cultivating more product experts. We are currently working on metrics that enables our application to automatically identify users having these functional role characteristics.


I will definitely make note of your interest and maybe write a series on the different roles in different types of communities.


Thanks for the comment, and I hope to see you next time. Actually, this week's post will be about influencers and whether they really matter or not.


Thanks, Michael.  Your response regarding roles is quite helpful.   I like the characterization of community types (support, enthusiast, innovation) you surfaced as well.   Do you have any posts or references you'd suggest we review to gain a deeper coverage of community types?  I think this may have been covered to some extent in Bernoff and Li's Groundswell, but its been a while since I read that.

Data Science

Hello Skip,


You are welcome.


Personally, I don't know, off the top of my head, a post on these community type. But communities can really be used for anything. And there are can be many different classification of communities based on size, age, purpose, audience, industry, etc.


The 3 types that I mentioned is a segmentation of communities by purpose, and it arises from an analyis of our client base. But for each of these three types, they can also be divded by audience into B2C, B2B, or internal, and industry, as well as other attributes such as size and age.


We analyze that because our benchmark clustering algorithm is based on these 5 different dimensions. There is a tribal knowledge base article on this. I've also analyze some participation data base on these segmentation in a post I wrote on the Economics of 90-9-1.


In my limited view, early on, business uses of community has been primarily to cut support cost. So support communities has been prevalent for a long time, (10+ years). As social media gain popularity, more and more companies are using it to drive marketing and sales. So later on, maybe 3 years ago, more and more companies are adopting the use of enthusiast communities for marketing and sales. Then shortly after, people are realizing the importance of value co-creation, that is why the innovation use case was created. But this is only my limited view from our client base.


Sorry I couldn't provide any more detail reference. I will seriously consider writing a post about this topic. Thank you for asking the question.


The Strength of Weak Ties 


cafe press needs to have neckties with these social network  patterns.


Michael, Through Facebook API can we find our fans (fans of our FB page) likes and their further friends likes/dislikes?



Data Science

Hello Gunjan,


Thank you for the questions.


I don't believe you can get any non-public data on fan's friends and/or their likes/dislikes through the Facebook API. I believe they have to opt-in specifically for you to do that. And most people are not going to let you do that. That is why so few people have Facebook data for any analytics.


Alright, thanks and see you later on Lithosphere.


Dear Michael!


Thank you so much for your brilliant articles about interactions between social networks and communities! I was so inspired by them that decided to share my impression to others. I have translated all of your 5 articles of your mini-series into Russian and published them on - the one of the influensest IT community in the russian-speaking world. We have awful huge lack of the discourse in Russian on such topics. I hope you don't mind.


Here are my translations:


Community vs. Social Network

How Do People Become Connected? Community vs. Social Networks 2

From Weak Ties to Strong Ties: Community vs. Social Networks 3

Maintaining the Strong Ties: Community vs. Social Networks 4

A Unified View of Communities and Social Networks

Data Science

Hello Maxim,


Thank you for the comment and for your interest. I'm very glad to hear that you are inspired by my work.


Thank you for letting me know that you've translated this mini-series into Russian. I would have completely miss that if you didn't tell me. I think that is great, and I really appreciate your effort in helping to diseminating these knowledge.


Personally, I certianly don't mind you translating and re-publishing my work, however, our marketing does request that you follow good social web etiquette.

      1. Mention the author: Dr. Michael Wu or Michael Wu PhD.
      2. Mention the fact that I’m the principal scientist at Lithium.
      3. Link back to the original source for the blog article.


You can see examples of how others does this when re-publishing my work at this URL. I see that you already did #1 and #2, so you just need to link back to the source article like what did at the end of the re-post.


Thanks again and hope to see you around Lithosphere in the future.


Michael, in fact my translations do have the back link to your original articles. The form of publication on suggests you to put a back link in a special field. And then this link is depicted at the end of the publication under the name of the author of the original text. Look at the image beneeth. Is it not enough? Should I dublicate the back link in the text of publication directly?


Data Science

Hello Maxim,


Thank you for the clarification. In this case, you are all set then. 

I just didn't know this convention at    ;-)


Thanks again for all the translation and re-publishing.

And see you next time.


Even if I am late at it, Welcome back Michael.

I agree that a shared interest is a key driver for a community, just like neighborhoods, people who live their comeback because they live there and have a shared interest within that community. “A rough mind map”, but what really drives retention? I figure if I purchase a house in that neighborhood that automatically displays a vested interest. But if I rent, what's my standpoint?


 I would like to gain a deeper understanding of community types, can you suggest anymorearticles on this matter?


Lithy Awards 2017

The winners in digital CX have been crowned!

See the winners!!