network is the new group

Recently, the Defense Threat Reduction Agency (DTRA) issued a request-for-proposal that invited researchers to develop theory to predict when and where social networks would emerge. They want to know how “networks” like Al Qaeda come to be. This caught my eye because I think the idea of networks emerging only makes sense if you are confusing networks with groups.

Networks are clearly popular. Every new club, association, organization and website today wants to be called a network. Network is the new group. What might have been called the “Lexington Preservation Society” in the past would very likely be called the “Lexington Preservation Network” if formed today. The popularity of the term probably does reflect a greater awareness of interpersonal relationships, but I don’t think there is a fundamental difference between these uses of “network” and the now replaced “group”.

Let’s test your network IQ. Suppose I told you I was studying a network with no ties: just a set of individuals. Does this violate your sense of network? If so, I think you are thinking of a group. Groups (certainly in contrast to classes) have a certain degree of internal cohesion. What if I said my network had lots of ties, but they were organized into three fragments or islands, such that all ties were within the fragments and none between. Would you say I have three networks instead of one? If so, I think you’re thinking of a group. Groups have some kind of boundary. They may be fuzzy, contested, or dynamic, but the notion of a boundary is fundamental to the notion of group.*

Networks, in contrast, have arbitrary boundaries and no expectations of cohesion. They are analytical devices. I call a network into existence simply by picking a population of nodes I would like to study and selecting a type of social tie that may connect these nodes. For example, I could choose to study friendships among the set of students living on one floor of a freshman dorm at a university. At the start of the semester, the network may be completely empty of ties: no one is friends with anyone else. A few weeks later, I may find that there are many pairs of students who are friends, and maybe even one or two short chains where A is friends with B who is friends with C. By the end of the semester, I might find that nearly everyone is at least indirectly connected to everyone else by some kind of path, and I may also find that some groups have emerged in which members have more ties with each other than to outsiders.

Note that conceptualizing the network in this abstract way has certain advantages. For one thing, it makes it easy to talk about network evolution. A network doesn’t emerge fully formed out of the Void — it evolves. But what is “it”? If you let me define the network in my arbitrary way, I can watch how “its” structure changes over time from having no ties to the end-of-semester structure (and, over the ensuing decades, perhaps back to having no ties). For another thing, it unconfounds the network from its structure. The number of ties now becomes a variable, so I can do things like test the hypothesis that a team’s performance increases with the number of trust ties. The degree of fragmentation is also a variable. As a result, a counter-intelligence agency can measure the extent to which it has succeeded in fragmenting a terrorist network.

The abstract approach does introduce some limitations. In this way of conceptualizing networks, it no longer makes sense to ask what the best way is to uncover a network. Should I measure interaction or affect, or something else? The answer is: you can measure anything you like – whatever you do measure defines a network. You can study the interaction network, the affective network or even the network of who doesn’t know whom. In the abstract approach, it also doesn’t make sense to ask, as the Dept of Defense has asked, when will a network emerge? The answer is: whenever an analyst conceives it. The network is always there. It is only the structure that changes over time. It is not a thing in the same way that a group is.


*Ok, I know that the fragments bit is actually more about cohesion than boundaries. So sue me.


Written by steve borgatti

December 30, 2010 at 6:23 pm

Posted in networks

9 Responses

Subscribe to comments with RSS.

  1. Great post. I think I’m going to make it required reading for all my future microsociology classes.



    December 30, 2010 at 7:34 pm

  2. This is a very nice observation. What’s more interesting about it is that research into “groups” (which gave us the early social psychological, gestalt inspired stuff that would be core of early network theory and “group processes” research, mostly at the University of Michigan) also emerged out of army funded research after war.

    I wonder if this state of affairs is both a product of the popularization of network research at one (pop science) end and the now full mathematization of network theory at the other (scholarly) end. I think that while the notion of a network with no ties now sounds counter-intuitive because of the semantic connotations that the term network has acquired—which as you point out has become somewhat synonymous with the term “group”—network analysts can unproblematically speak of “graphs,” “edges” and “vertices” which are technical terms which carry very little semantic baggage. So it is easy to talk about an “empty graph” or a “disconnected graph” without violating any semantic intuitions than it is to refer to a group with no ties or three disconnected cliques as a “network.”

    Because of this, I’m afraid that was going to happen is that “network” is going to continue on its way to being conventionalized as a semantic analogue to group and network analysts are just going to continue to write papers using graph-theoretic language to sidestep any confusion. So what I’m trying to say, is that your campaign to reclaim an “abstract” meaning for the term network may be doomed Steve, since the term “graph” is already doing that work at the technical level.



    December 30, 2010 at 11:01 pm

  3. […] the rest here: network is the new group « Tags: are-thinking, sense, test-your, think-you, told-you, violate, your-sense Category: default […]


  4. The term network has (for some time) had specific meanings in other disciplines, most notably computer science. Unconnected nodes, in that usage of the term, are un-networked nodes and a network is an interconnected group.

    Given the explosion of interest in network (and networked) computing since the early 90’s, I shouldn’t think that lay usage of the word “network” in this form is terribly surprising. Even if it is applied to people.



    December 31, 2010 at 3:08 pm

  5. Sure, I can imagine a network with no ties.
    Here is A 5×5 sociomatrix with no ties


    I would call this a maximally sparse network ;)



    December 31, 2010 at 3:41 pm

  6. hello steve — it’s been awhile since last contact, and i’m just an occasional outsider interloper here, but i’ll reiterate an old point anyway, given my persistent interest in distinguishing tribes, hierarchical institutions, markets, and info-age networks (TIMN) as distinct forms of organization — an interest that runs counter to the interest of social network analysts and network scientists who seem to view all organizational forms as networks:

    the abstract view you propose here fits in the latter camp. okay, but i’d like to ask for more explaining. a network with no ties? that violates minimalist definitions of a network (including in your network-theorizing paper). a “maximally disconnected network”? interesting notion. but surely that’s not really a network? if so, then it’s at a terribly abstract level that is more about categorization than real social organization. indeed, you mention that a network may be just an “analytical device” (not to my liking, of course).

    if one is going to be that abstract with the network term, then why not remember that it’s also been done with the group and class terms, as i recall. if one could treat some set of “maximally disconnected” actors as a potential network that may, or even in fact does, evolve into a real network, then i’d suppose that conceptual judo could also be used on the group and class concepts as well. those terms can be treated as analytical devices too. yes? no?

    sometimes it may make sense to talk about a group of networks, or, in a twist, about a network of groups. note the reverse in wording. each phrase means something a little different. and it seems to me that the group term shifts definition more than does the network term in the two phrases. does that lead to clarifying anything?

    anyway, i’m delighted to see you guest-posting here. this is the only blog i’ve found, except for the p2p foundation blog, that often has informative posts explicitly about the four forms of organization that continue to interest me. — onward, david


    david ronfeldt

    December 31, 2010 at 9:46 pm

  7. Hi David, great to hear from you! I was planning to address the organizational forms issue in another post, so stay tuned. I might also post something on whether labeling something an analytical device, as I did, accomplishes anything.


    steve borgatti

    January 1, 2011 at 2:53 pm

  8. […] & hierarchical and informal & non-hierarchical, as well as everything in between. See Steve Borgatti for more on this point. There has been something of a debate as to whether such ‘open source’ actors are indeed any […]


  9. steve (and others): i gather — and call your attention — that milton mueller, in chapters of his new book “networks and states: the politics of internet governance” (2010), contrasts networks as a school of analysis to networks as an organizational form.

    see the review recently posted by clay spinuzzi at his blog:

    i’ve been trying for over ten years to point out differences between these two views and to field similar points, usually without success. so it’s good to have company on this theme, and to learn that mueller provides a detailed analysis. i must read it for myself.

    meanwhile, it seems to me that the contrasts between the two views appear in some discussions about the dynamics of the networking behind the protest movements in north africa and the middle east. for example, i’ve seen an academic social network analyst argue that these networks are likely to evolve (devolve?) into hierarchies of some kind because of the dynamics of preferential attachment (and the “iron law of oligarchy”). in contrast, many activists are intent upon sustaining their networks as a distinct and viable form of organization.

    at times i wonder whether social network analysis may turn out to be as much a hindrance as a help in understanding the rise of networks as a form of organization distinct from tribes, hierarchical institutions, and markets. — onward.


    david ronfeldt

    March 7, 2011 at 2:48 am

Comments are closed.

%d bloggers like this: