The Unspoken of Groups

David Weinberger
home page weblog

This is a recapitulation (by no means a transcription) of my comments
at the O'Reilly Emerging Technology Conference (April 26, '03 in Santa Clara)
under the title "What Groups Will Be."

"Learning from experience is the worst way to learn." That's one of the many right things that Clay Shirky said in his keynote yesterday morning [and later published here]. Learning by reading is far preferable, he said. Absolutely. But, he said this in reference to the fact that when it comes to the behavior of groups, we keep making the same mistakes. Well, why? Even if we're not reading accounts of group behavior the way that we should, we all have had the same  frustrating experiences with groups. So why don't we learn by experience? Why do we keep making the same mistakes?

I want to suggest — eventually — that there's a reason why we keep making the mistake that Clay specifically points to: the failure to adopt a group "constitution." It's not because we don't learn. It's because of the importance — in even the most vocal groups — of the unspoken.


But first some stipulated definitions. For the purpose of this talk, I mean by "group" a set of people who know one another and know they're in the group. There are lots of borderline groups by this definition, but in the paradigmatic one, people know one another and are aware of the group.

This excludes "groupings," people who have something in common but who don't know one another. For example, a demographic slice is a grouping but not a group.

This also doesn't talk about "communities," a word that is important enough to preserve for considered use. I understand a community to be a group in which people care about one another more than they have to. And I'm not going to talk about communities today.


I have two premises today. The first is that groups are really, really important. I believe they're what's driven the public passion for the Net from the beginning. But I suspect I don't have to talk you into seeing the value of groups.

Second, the Net is really bad at supporting groups. It's great for letting groups form, but there are no services built in for helping groups succeed. There's no agreed-upon structure for representing groups. And if groups are so important, why can't I even see what groups I'm in? I have no idea what they all are, much less can I manage my participation in them. Each of the groups I'm in is treated as separate from every other.

Now, the fact that the Internet is bad at supporting groups comes straight out of the End-to-End principle that recommends in designing a network that you put in as few services as possible so that these services can be invented by people on the "edge" of the network. (By the way, I find it interesting that David P. Reed was one of the co-authors of the original End-to-End paper and is the author of Reed's Law that squarely locates the value of the Internet in its group-forming ability.)

So, if groups are important but are under-served by Internet and if the Internet lets us innovate on the edges, then there's a market opportunity for group services.

Room for innovation: Friendster

Lots and lots of services have arisen. I want to pick on one — Friendster — because it's new and appealing. Friendster attempts to provide a social network with special tools loose federations of friends might want. And they've done a good job of thinking what such circles might need and providing it in a clean, easy to use interface.

How do you join Friendster? Very likely it's because a friend invited you to be her Friendster friend. You get an email saying, for example, that Halley Suitt wants to be your friend at Friendster. There's a button to press to accept or reject Halley. This is already indicative of the problem. Of course it's good that Friendster requires my permission to be listed as one of Halley's friends. And with Halley, I have no problem: I know her virtually and I also know her in the real world, and I have no problem saying, yes, I am Halley's friend. But there are lots and lots of people who might ask me to be their friend for which the situation is much dicier. There are people who are acquaintances, or relatives, or former college housemates I've been trying to avoid for years. There are people for whom I'll press the Accept button not because they're friends exactly but because they're not enough not-friends that I want to reject them, or because I want to impress them, or because I want to kiss their butt in public, etc. Friendster asks me to be binary about one of the least binary relationships around.

I'm not suggesting that Friendster made a poor design decision. I'm suggesting that there is no good design decision to be made here.

Then you join Friendster and are faced with a one-page profile. It's relatively inoffensive as far as they things go. And I like that it allows free-form entry of text rather than making me choose among explicitly listed alternatives. But it wants to know my favorite books. I don't have a list of favorite books. When I finish a book, I don't mentally rank it and say "Ah, it's #2057, pushing The Thornbirds down a notch."

Worse, it asks me for my interests. People don't know what they're interested in. Oh, I can list four or five things, but I'm interested in many many more things than I could ever list precisely because I don't have a list implicit in me waiting to be output. For example, I listed "weblogs" on the form, but it didn't occur to me to list that I find it interesting that John Wayne evaded service in World War II. I didn't know I was interested in the history of the telegraph until I read The Victorian Internet.

But we know that these fields aren't really asking me to list my interests. Rather, they're asking me to market myself, to pick out the interests and books that I think will present me well and are likely to pull me into circles of others who I might want to meet virtually. But even there, common interests often are precisely not how relationships are struck up. For example, when you meet someone new, not infrequently the conversation begins with a statement about how you've never even considered a particular area: "You're studying erotic taxidermy? That's fascinating! I never even heard of such a thing." Areas of disjunction over interests often are the most fruitful way to begin a conversation.

Further, as Zephoria suggested when I posted my initial ideas for this talk, the Friendster sign-up sheet assumes that there's only one me I want to put forward. I should probably have a profile sheet for at least several different me's: the blogger who wants to find other bloggers, the consultant trolling for clients, etc.

The real issue is, I believe, that any profile asks me to make myself explicit. And that can't be done without doing damage to the truth about myself.

Wrong model

But making explicit doesn't just do damage to selves. In general, making explicit does violence to what is being made explicit. (In the modern age, Heidegger gets credit for this idea.) Making things explicit isn't like unearthing an archaeological find that's just been sitting there, waiting to be dug up. Making explicit often — usually — means disambiguating and reducing complexity.

The reason is simple. The things of the world exist as they are only within deep, messy, inarticulate, shifting, continuous, fuzzy contexts. This is certainly true of human relationships, although I believe it's also true of all that we find on the earth, waiting in it, or promised above it. The analog world — the real world — is ambiguous. That's a source of its richness. In making a piece of it explicit, we make it less ambiguous and thus lose some of its value and truth.

Back to Clay and constitutions

Clay said in his keynote that we can't keep the technical and social conversations apart. That's because the technical, without the social context, is just a pile of silicon. But when he talked about groups creating "constitutions" (explicit rules of behavior, membership and operation), I want to say the same thing: you can't keep the constitutional and the social apart. You can't disentangle them. And this suggests a reason why groups keep making the same mistake, putting off drawing up a code that acts as a constitution.

Perhaps groups can't write a constitution until they've already entangled themselves in thick, messy, ambiguous, open-ended relationships. First, without that thicket of tangles, the group doesn't know itself well enough to write a constitution. A constitution is descriptive as well as prescriptive. For example, if a group with the disposition of Slashdot were to come up with a constitution that said "No sarcasm will be tolerated," it would fail. The constitution has to match the group's nature, but that nature emerges from the thicket of ambiguous relationships.

Second, writing a constitution is an act of violence because it's making explicit rules and mores that were left unstated until some problem arose that pushed the group into the constitution-writing process it had been avoiding. We all know how ugly constitutional discussions can get. In order to have such a charged conversation, the group needs a web of good will. That web takes time to develop. So groups generally dare not attempt a constitution early on.

Example: Ben Hammersley's talk

I've heard the question of the explicit running through many sessions at this conference (perhaps only because I'm listening for it). For example, it showed up in Ben Hammersley's talk. After explaining threadsML, a standard I've been involved with and encourage you all to support, he talked about the ENT standard from Matt Mower and Paolo Valdemarin. ENT provides a way to embed a blog entry's topic into an RSS feed. This is a good idea, and Ben suggested a way out of the old problem of people using different terms to mean the same thing: if A links to B, then the two words they use for their topics are likely to be similar.

But here's the problem: things aren't about what they're about. "Aboutness" is also contextual and ambiguous. For example, if my blog entry on the JFK assassination links to the 1962 Sears catalog from which Oswald bought his rifle, the author of that catalog will not have labeled it as being about the JFK shooting. And if a scientist publishes a paper about a new polymer, she may in passing reject some closely related compound because it's too sticky...but that may be exactly what you're looking for. So, for you the article is about what the author tosses away in a footnote. Not to mention that in much of the best writing, about-ness is an emergent property. So, while the author's intentions are an important clue, aboutness is ambiguous. Systems that too easily categorize and classify based upon a univocal idea of aboutness do violence to their topic.

Example: Knowledge management

This is a digression, but since Geoff Cohen yesterday said to me that Knowledge Management is the largest example of failed social software anywhere, I thought I'd mention it quickly.

There were two initial impulses behind KM.

The first was: "We have more information than ever but don't understand our business any better. We need to identify the information that counts. That's a special class of information and we'll call it ‘knowledge.'"

But knowledge isn't a special class of information. It's not like gems among the stones. You can test a diamond by seeing if it cuts glass. Knowledge is only knowledge because of its context. In this, knowledge is a particular bit of twisted scrap metal you find in a junkyard. There's nothing inherent about the piece that distinguishes it from the other scraps of metal around, but it catches your attention because you happen to have a hole in your tin roof exactly that shape: it's valuable because of the context. And this is a problem for the type of KM system that thinks of knowledge as a special class of information, for computers are particularly bad at evaluating context.

As an example of this, consider a company whose engineers have a mailing list about computer games. From the company's point of view, it looks like a digression until the  day the company discovers it needs to do some high resolution graphics work. Now what was just plain old scrap metal turns out to be real knowledge.

Thus, KM systems that look for explicit marks of knowledge in the vast sea of information miss the point: what makes something knowledge instead of mere information isn't inherent in the information itself. It is dependent on a messy, shifting and implicit context.

The second impulse behind KM was the fear that Marge would quit. Marge is the company's best tech support or maintenance person or whatever. So, the company thinks it needs a "data dump" from Marge: she needs to externalize her internal knowledge so that others can internalize it.

But because Marge is not a database, she can't do a data dump. In fact, her implicit knowledge isn't sitting there waiting to be externalized. There are people who can do this externalization well, finding what's implicit and making it explicit in ways that others can understand. They're called teachers and writers. Marge may have those skills but she may not. Making explicit is a genuine craft, a skill, even an art. Done well, the violence is mitigated by the ability of the artisan to expose the connections to the context as well.  

Software for groups

If in language and in groups, the unspoken is the source of the greatest value; if it is where richness lies; if the tangly, messy, ambiguous, latent, unstated, continuous, context is the reality of who we are as speakers and as people who share a world, then what software might work for a group?

Social software.

Could a term be more vague? It could be taken to include everything from email to instructions on how to hold hands. But in fact it's coming to mean something more specific: emergent social software. It's emergent in two ways.

First, it enables social groups to emerge. It goes not from implicit to explicit, but from potential to actual: social networks are only actual in the first degree. From the second degree on out, they're only potential. That is, if I know you and you know her, my social network reaches to you but not to her. And that's for a very simple reason: you haven't yet introduced me to her. Until you do, she simply isn't part of my social network, no matter what the maps of the One Unified Social Network may make it look like.

Social software enables social networks to become actual.

Second, social software enables the social network's shape to emerge. Rather than, for example, dividing the company into groups, structuring access permissions, and provisioning them with the toolset it's anticipated they'll need, emergent social software is low-tech and relatively non-intrusive. It may include such familiar items as chats, mailing lists, instant messaging, weblogs and wikis. The access controls are generally turned off at first. The taxonomy is blank. The webspace is unfurnished and undivided. The group builds what it needs as it needs it. The structure of the group's tools follows upon the group's growth into itself. For example, the group may use a wiki — a jointly editable web site — that starts off blank. As the group develops interests, individuals will add in pages, structuring the workspace. Because the site is editable by every individual, what emerges is a workspace that reflects not the group's expected interests or its pretended interests or its satisfy-the-boss interests but its real interests...interests it may not even know it had.

Why social software now?

If social software has been around for as long as software has been, why is it becoming a buzzword only now? Is it because consultants see a new wave to ride? Sure, that's a part of it. But since most social software is relatively simple and inexpensive, this doesn't promise the big consulting bucks that, say, knowledge management did.

Even if consultants get rich off of social software, that's not why the rest of us are excited about it. That's not why it was the undercurrent of conversation at the recent O'Reilly Emerging Technology Conference. It's not why some of the people I respect the most have been drawn towards this "new" software as engineers and as customers. Social software is sending some of the old energy through us again.

Good. Frankly, we're a dispirited clan that could use a lift.

But, perhaps that isn't a sufficient explanation either. Companies that have been burned by groupware and stymied by knowledge management systems are beginning to explore emergent social software. This tells us not only that they're looking for something simpler, but that perhaps they're willing to make the most basic trade as corporations, the trade we made when we first got on the Net as individuals: trust for hope.

Five years ago, the idea that idea of putting up a page that anyone can edit would have been laughed at. What's to stop vandals wiping out a week's work with a Control-A Delete? But now wikis look like they make sense.

Five years ago, it was obvious beyond question that groups need to be pre-structured if the team is to "hit the ground running." Now, we have learned — perhaps — that many groups organize themselves best by letting the right structure emerge over time.

Such beliefs deliver trust and get hope in return.

But I want to go further. If such a change is occurring — I say if — then it, too, is emerging from a greater, implicit whole. And here's where I place my own hope. Could it be that the this turning of the greatest of the beasts of structure, corporations, could betoken an even more significant change? Could we at last be turning from the great lie of the Age of Computers, that the world is binary? Could we be ready to embrace the most obvious of facts: The earth is continuous, with every edge imposed? The world is ambiguous, with every thought, perception and feeling just a surface of an unspoken depth?

We can hope. Can't we? Please?


1. I'm an uncompensated member of the Board of Advisors of a company that makes social software.

2. Here's a place to discuss this article.