This ideas in this post are based on conversations with Bernie Hogan and should be interpreted as the production of our co-thinking.
All too frequently, someone makes a comment about how a large number of Facebook Friends must mean a high degree of social capital. Or how we can determine who is closest to who by measuring their email messages. Or that the Dunbar number can explain the average number of Facebook friends. These are just three examples of how people mistakenly assume that 1) any social network that can be boiled down to a graph can be compared and 2) any theory of social networks is transitive to any graph representing connections between people. Such mistaken views result in broad misinterpretations of social networks and social network sites. Yet, time and time again, I hear problematic assumptions so let me start with some claims:
- Not all social networks are the same.
- You cannot assume network transitivity.
- You cannot assume that properties that hold for one network apply to other networks.
To address this, I want to begin by mapping out three distinct ways of modeling a social network. These are not the only ways of modeling a social network, but they are three common ways that are often collapsed in public discourse.
Sociological “personal” networks. Sociologists have been working hard to measure people’s personal networks and much of the theory of social networks stems from analysis done on these networks. Different scholars have taken different approaches to measuring personal networks, but, most stereotypically, this takes the form of a clipboard and pencil as a young grad student queries an individual to recall who they talked to yesterday and indicate who they would lend money to or call when they are having an emotional breakdown. On classic measurement survey is an appendix in the back of Claude Fischer’s “To Dwell Among Friends.”
Most sociological theory stems from analyses of these personal networks. Social capital, weak ties, homophily, … all of those theories you’ve heard about are based on personal networks. Given that these are typically measured by eliciting people’s understandings of certain categories (e.g., “friend”), there’s a strong overlap between everyday language around social networks and the categories being measured.
If you’re a sociologist talking to anyone other than sociologists, you would probably speak of personal networks as the golden standard, the baseline truth. Of course, if you were being honest with yourself or your colleagues, you will note that these measurements have their methodological flaws and biases which is why the scales for measuring personal networks haven’t stabilized and why scholars still struggle with the best ways to elicit meaningful information from people being surveyed.
Behavioral social networks. Behavioral social networks are the networks derived from encounters between individuals. In their efforts to measure personal networks, sociologists have often tried to get people to manually document encounters with others through diary studies. With new technologies in place, folks have gone on to generate behavioral social networks through the traces people leave behind. For example, a record of someone’s email exchanges provides a handy accounting of that individual’s behavioral network. New technologies introduces new opportunities for measuring behavioral networks. Many genres of social media let us see who communicates with who. GPS technologies let us see who shares physical space.
Behavioral social networks provide valuable insight into people’s practices and interactions, but they do not confer meaning. This is not to say that they don’t have value. I would love to find the strangers that I regularly share space with as I traverse Boston. But we cannot assume that these are my friends or acquaintances. Yet, there seems to be a tendency (especially among geeks of all stripes) to overlay meaning-laden terms on top of these networks, to assume that high connectivity means friendship. This is where trouble often arises. Just because I spend a lot of time with my physical therapist does not mean that she is more important than other people in my network who I see less frequently.
The other difficulty in measuring behavioral social networks is that, at least to date, we measure distinct channels of connection. This complicates our ability to do meaningful comparison across people. If I use AIM as my primary way of keeping in touch with Person A and email as my primary way of keeping in touch with Person B and you only look at one medium, you get a distorted picture of who I communicate with. As communication channels proliferate, this only gets messier. So even when we talk about behavioral social networks, we have to talk about them in across a particular channel.
Publicly articulated social networks. Articulated social networks are the social networks that you intentionally list. In some senses, this is what sociologists are eliciting, but people also articulate their social networks for other purposes. Address books and buddy lists are articulated social networks. So too are invitation lists. Most recently, this practice took a twist with the rise of social network sites that invite you to PUBLICLY articulate your social network.
At this point, I would hope that most of us would realize that Friends != friends. In other words, who you connect to on Facebook or MySpace or Twitter is not the same list of people that you would say constitute your closest and dearest. The practice of publicly articulating one’s social network can be quite fraught because there are social costs to the process of public articulation. Issues of reciprocity emerge and people find themselves doing a lot of face-work to navigate the sticky nature of having to account for their social relations in a publicly accountable way. Thus, the list of who you might list as a Friend is often a mix of friends, acquaintances, family members, people from your past, fans, professional colleagues, familiar strangers, and people you don’t particularly like but don’t want to offend. Oh and the occasional celebrity you think is interesting.
Relating Different Social Networks
These networks are NOT the same. Your mother may play a significant role in your personal network but, behaviorally, your strongest tie might be the person who works in the cube next to you. And neither of these folks might be links on your Facebook for any number of reasons.
Our instinct then is to ask: which is the “real” social network? Frankly, it depends on who you ask. Your mother may be cranky that you don’t talk to her as often as your colleague and she may resent your refusal to Friend her on Facebook, but this doesn’t mean you love her any less. Of course, this doesn’t stop her from thinking you don’t love her. If we’re trying to understand emotional affinity, the behavioral and publicly articulated social networks aren’t particularly helpful. But if you’re mother thinks that time is not only a proxy for emotional depth but a proof of it, your behavioral social network might really upset her. (Note: behavioral social networks have gotten people into trouble in the past. See Cobot.)
The truth of the matter is that there is no “real” social network. It all depends on what you’re trying to measure, what you’re trying to do with those measurements.
We do ourselves an intellectual disservice when we assume that these different types of networks are interchangeable or that studying one automatically tells us about another. Most scholars get this, even when they’re quoted out of context by journalists to suggest otherwise (see Cameron Marlow). But I get the sense that a lot of journalists, marketers, advertisers, politicians, and everyday folks don’t. This is a problem.
Those who treat different social networks interchangeably project properties onto the network they’re analyzing that don’t hold. People aren’t inherently cool or connectors because they have a lot of Friends on a social network site. Bus drivers and waitresses are much more likely to encounter more new people on a daily basis than executives, but this doesn’t mean that they have more social capital. People who email regularly do not necessarily have strong tie strength.
This is not to say that structural information in behavioral social networks or publicly articulated social networks may not work as a proxy for personal networks. Perhaps the networks derived from a particular social media tool or through a particular channel of communication do actually provide insight into a person’s personal network. There are great ways to empirically test this hypothesis involving the combination of structural analysis and interviewing. But you cannot simply assume that they are meaningful proxies just because they are both social networks.
There are also many opportunities for new research when we tease out different types of social networks. What if we overlay the different types of social networks? Can we get a better sense of how someone manages their social network? Can we see new structural properties that give us new insights into how people connect, share information, gather support, etc.? So many possibilities!
I’m super excited that so many people from so many fields are getting interested in social networks, but I’m also scared that there are a lot of assumptions flying around that make it difficult to make sense of people’s contributions to this emergent field. Increasingly, I see sociologists and computer scientists and mathematician and economists outright dismiss work outside of their field as “wrong.” I think that part of the problem is that we’re each failing to account for what we can and cannot say based on the types of analysis we’re doing. And I think that we often talk past one another because we’re all talking about social networks but we’re talking about different social networks. In accounting for three types of social networks here, I’m not trying to be all-inclusive, but I am trying to point out that there are differences and that we cannot assume transitivity either in terms of structure or theory. If we can find a way to better identify what kinds of social networks we’re talking about and when and where what theories apply, I think that we’ll go a long way in bridging different intellectual discourses.
woot! awesome break-down danah. thank you!
An example of assumption overload, perhaps, was neuroscientist Baronness Susan Greenfield’s comment about social networking before the House of Lords last winter (http://www.netfamilynews.org/2009/02/social-networking-infantilizing-users.html). There definitely needs to be more cross-disciplinary discussion, because one of the biggest assumptions made is that social networks are technology, an add-on to “real life.”
Anne, part of this thinking was pushed forward by Baronness Greenfield’s comments. She helps us understand the naivet� that comes with misunderstanding the difference between media, social media and social interaction.
– –
To follow up (and similarly consider this my thinking co-evolving through conversations with danah), I tend to lean on the sociological term “reify”, as in to make real. We false reify networks when we forget that it is really the network of, for example, at least 3 reciprocated email between ties, or the network of ‘very strong’ ties. Robin Dunbar is the worst victim of this. While he considers networks as moving outward in a periphery of activity and intensity his ideas have been distilled into a Dunbar number of 150, which is then reified as the “size of the ideal network”.
Networks are all to easy to reify when we see the powerful images that give us ‘visions of omniscience’, as in, look how much data we can throw onto the page. But as we suggest, it is important to remember the processes that go into making these ties and these networks, for it will also help us understand the processes, tensions and constraints that emerge from the networks as well.
Finally, I can’t overestimate the importance that danah highlighted in referring to “a” social network rather than “the” social network. I suspect that simple article shift will help our thinking immeasurably. @blurky
danah,
thank you thank you. wonderfully concise and precise. a quick question and a comment.
question: communication over current “conversational media” would be considered behavioral, i’m guessing, insofar as communication is the social action and interaction?
comment: to put a media theoretical spin on this, i think conversational media transform talk (tweets, comments, status updates and all other feed content, including system messages like activity updates) communication into a symbolic form of sorts. that the act of communication is separated from the form it takes online (writing). communication lives on, and becomes available for commercial use, for search, you name it.
once separated from its function as interaction or communication, it’s no longer a reliable index of intent or behavior. social context is not longer directly coupled to interaction and communication. tweets become capable of more than talk (among people), while they are at the same time less than capable of talk.
how would one research this space? if the communication (eg tweets) can’t be taken simply as talk (social behavior)?
i’m interested in your take on this. conversational media lack so much of the structure and so many of the cues available on page-based profile-centric social networks.
personally i look at conversational media as failed socialities in which ambiguity in intent and arbitrariness are in part responsible for their appeal. (authenticity would be a related concept, because appearance and identity are so easily manipulated in conversational media.)
cheers,
adrian
> network transitivity
I’m not sure what you mean by that, but for network-as-a-complex-graph physicists and CSist out there (who would tend to agree whole-heatedly with your message, or at least should; and who should be the main readers of that draft) transitivity means “A knows B, B knows C therefore A (should) know(s) C.” If I understand you properly, you actually mean “A calls B therefore A emails B” I would rather call that “network overlap”.
> Increasingly, I see sociologists and computer scientists and mathematician and economists outright dismiss work outside of their field as “wrong.”
That sentence is right, but what follows misses to account for your own point of view: those are increasingly open-minded, but you spend an increasing amount of time with them, so you come across is more reserved then you’d want, but it’s not a bad sign.
To do interdisciplinary science, we need more categories then “right” and “wrong” – actually, economics (and medecine) is the closest thing to such a bastard science (being an economist, I say that with all the love in my heart) and we have “observations”, “stylised facts”, “conjectures”, “demonstrated effect”, “measured effect”, “something journalists know and is actually true”, etc. You’ll need to accept that a theoretic physicist will frown until you reach a very high point of certainty: from my short experience, I’d say that it’s the mismatch of expectation in cross-field gratification that is the real issue.
> We false reify networks […]
Not so sure it’s false: it is the discourse people are reading, getting used to, etc. Very nice people not on Facebook *are* missing on invitations because of their Luddism, and loosing their very real friends. Many consider that anyone with more then 150 Facebook Friends is a superficial bitch and should not be considered a real friend – well, not that strongly, but they do tune their behaviour. A friend (and Friend) of mine as 1400 Friends, most of which are actual friends (she is amazingly pretty, intelligent, talented, worldly, travelled a lot and speaks more languages then I can count, just like her twin hence the impressive figure): she does suffer from that and from contextual collapse a lot, as well as a continuum of behaviours from spam to distant relations assuming she has low Friending standards. What was an amazing asset couple of years ago now turns into a modern (i.e. ambiguous) scarlet letter.
Thank you danah (and Bernie) for this very interesting piece. You surely get a point here. That’s why, now more than ever, when someone wants to study social networks have to be very clear about the adopted methodology: to be able to define clearly what he is observing and what kind of social network is the research dealing with. This is particularly true when you came to describe social practices that occurs within the social networks, I’m currently observing gaming in SNSs and “hanging-out gaming” is not the same in SNSs and in personal networks… even if you’re always playing with your “friends” :-).
An interesting side note about bus drivers etc – there’s a new book out which sounds like it revalues their role and connectivity. I’ll be reading it…
Blau, M., & Fingerman, K. L. (2009). Consequential strangers : the power of people who don’t seem to matter– but really do (1st ed.). New York: W.W. Norton & Co.
And now a word from a journalist who’s trying to get it right: Great piece, danah and Bernie. I struggled with these issues while researching “Consequential Strangers,” my upcoming book about that vast territory of relationships outside the inner circles.(www.consequentialstrangers.com) It’s a book about relationships, not networks per se, but of course in order to describe CS, I had to bring in a discussion of networks. But because of the complexity you describe, I adapted Antonucci’s “social convoy” instead,leaving it to the reader to analyze their own constellation of contacts, past and present. (I do my best to explain network types and measurements only in the endnotss, as mine is a lay audience. In reprints I’d like to add this link!)
One of the reasons the term “consequential strangers” (coined by my academic collaborator Karen Fingerman as a way to describe peripheral relationships) piqued my interest is that it gave a name to relationships that had been given little coverage in the press. People can’t value something they can’t name. While sociologists have long known the value of weak ties, lay people have not. This project began in 2006. Three and a half years later–light years in Internet time–it’s clear that no matter which type of network you look at, it is filled mostly with consequential strangers! I have no idea whether reearchers or lay people will adopt the term in everyday parlance. But the important point of both my book and your essay is that we need to think about the broader social landscape in more nuanced terms.
And now a word from a journalist who’s trying to get it right: Great piece, danah and Bernie. I struggled with these issues while researching “Consequential Strangers,” my upcoming book about that vast territory of relationships outside the inner circles.(www.consequentialstrangers.com) It’s a book about relationships, not networks per se, but of course in order to describe CS, I had to bring in a discussion of networks. But because of the complexity you describe, I adapted Antonucci’s “social convoy” instead,leaving it to the reader to analyze their own constellation of contacts, past and present. (I do my best to explain network types and measurements only in the endnotss, as mine is a lay audience. In reprints I’d like to add this link!)
One of the reasons the term “consequential strangers” (coined by my academic collaborator Karen Fingerman as a way to describe peripheral relationships) piqued my interest is that it gave a name to relationships that had been given little coverage in the press. People can’t value something they can’t name. While sociologists have long known the value of weak ties, lay people have not. This project began in 2006. Three and a half years later–light years in Internet time–it’s clear that no matter which type of network you look at, it is filled mostly with consequential strangers! I have no idea whether reearchers or lay people will adopt the term in everyday parlance. But the important point of both my book and your essay is that we need to think about the broader social landscape in more nuanced terms.
If one is old enough, it is possible to remember the days when a cute reversal probe was “now computers have networks; soon it will be that networks will have computers.” I think an analogous shift is coming in the language that seems to be somewhat problematic when thinking about how to infer the nature of both individual and multi-lateral relationships from the explicit and implicit graphs of social networks. Focusing on the networks in one’s social milieu and interactions, and using language like “Friends,” confuses the issue, it seems to me, and creates complication when we would be better served by complexity.
Why not turn the problem around? Why not approach the analysis from a ground of the nature of relationships that create emergent, autopoietic social structures, rather than the other way around? I’m finding that this way allows one to create a whole lot of clarity in what previously were pretty complicated issues.
(Hi, Bernie! [waves])
Looking at this as a psychologist, I would say that the greatest pitfall will be to look at structure first – that tempts us into conceputalizing a network as unchanging.
We are better off looking at underlying dynamics that might lead a network to ‘flourish’.
And we need to distinguish social capital/power from flourishing. A heavy weight executive can be like a bomb- inert until it goes off – that is with power but not flourishing. A youngester riding the Tube for the first time is a light weight but flourishing in that they are going some where.
The work on happiness and phase states is where I am going with this – Frederickson and Losada 2005, Am Psychologist.
There are so many of these social networking sites these days. An interesting note is how dangerous they can be for a person, especially for one’s career. I just read a pretty interesting blog posting about it here: http://lawblog.legalmatch.com/2009/06/12/be-careful-what-you-write-social-media-marketing/
Another great thinking piece danah! ^.^
I agree that there are so many types of social networks at play, and just because you’re facebooks ‘friends’ with someone doesnt mean you’re even more than acquaintance with them in ‘real life’. I would say that a great percentage of the people i am ‘friends’ with on facebook (mostly old school peers) I would not even consider talking to in real life and from what I’ve seen on their facebook profiles they would not want to talk let alone be ‘friends’ with me. I am a collected item for them, one in a set. It is like the trophy cabinet, who can have the most high school ‘friends’ (read peers) as their facebook ‘friends’ on their site. These people dont even say anything, they just add you because they remember your name. I dont know if its actually they have some friendly outreach, maybe we couldnt have been friends in highschool but they are trying now… One example, my partner (we went to the same school) is ‘friends’ with a girl who i never really had anything to do with (we were in the same class) because she was ‘cool’ and i was a ‘nerd’. she starts chatting (the IM feature) to him asking about us and saying that she always really liked me. He recalls to me that she always asked about me in school. Now, she is my ‘friend’ on facebook and has never posted anyhting on my wall and i havent on hers. We have literally never talked on facebook nor since highschool (does ‘hi’ count?). Boggles the mind yes? Not really.
Thanks for this analysis – incredibly useful.
It occurs to me that these definitions of social networks are going to be useful to most people personally as our social networks become augmented by online interactions. It can help when we ask “what *is* a friend” (as Euan did the other day, http://www.euansemple.com/theobvious/2009/7/26/friendship.html) to be able to think which kinds of networks different relationships fit into.
Hi, great post. I picked it up from George Siemens blog, and responded on http://roys-discourse-typologies.blogspot.com/search?q=digital+ecologies. Might be of interest.
Hello danah,
I know this is old post but I’ve been thinking about it a lot recently and I’d like to make sure I understand your Cobot reference.
Did you highlight Cobot to demonstrate that the researchers mistook users interactions/behavior with respect to Cobot as being more meaningful than it was? Was your point that although Cobot was one of the most talked to/talked about “personalities” on LambdaMoo it in no way indicated that Cobot was influential, popular or important to the systems users? Or was your point simply that Cobot somehow got the researchers in literal trouble?
I ask to ensure I am not misunderstanding the context/point of the reference.
Siddiq
Some Outrageous Propositions -by Miller McPherson in 1990
Propositions 1-8 at link below
9. Many of the smartest sociologists are moving in the direction of other disciplines, because these other disciplines allow them to think more clearly (among other reasons). Our discipline makes us stupid, rather than smart. This fact is caused by our not having a distinctive and well connected set of ideas. A consequence of the movement away from sociology by our smart people is a further decrease in our ability to talk to one another, which makes us more stupid. As our most important ideas become more and more integrated into other disciplines, we lose the centrality of our ideas, and thus our disciplinary identity.
http://www.u.arizona.edu/~mcphersn/outrage.html
I’ve come across this post thanks to your recent keynote at SXSW. During this keynote you talked about Google Buzz non-technical mistakes due to the confusion they made between personal, articulated and behavioral networks. I am wondering how much this confusion was led by the fact that Buzz is embedded within Gmail. Indeed IMHO Webmails at their core nature aggregate or collapse the three different kind of networks (only exception would be that not all of my encounters would be listed in my email contacts but they could potentially be) since Buzz comes right into gmail, it was meant to collapse the three different kind of networks.