[spectre] Social Software -- and groups that blow up

[Bruce Sterling]

Jon Lebkowsky Blogs a Speech by Clay Shirkey
jonl at weblogsky.com

http://www.weblogsky.com/

Clay's been studying social software, has seen a pattern in social software 
for large scale, long term group interactions. What he now thinks is one of 
the core challenges in designing social software.

Social software: software that supports group interaction. Emphasizes how 
radical that is. Prior to the Internet, we had lots of patterns that 
support one-way outbound. Prior to the Internet, the last invention that 
affected group conversation was the table. Ridiculously easy group forming 
is really new. We're just finding out what works, still learning how to 
make these things. This definition doesn't specify class of technology, 
which is unsatisfying. Email is not necessarily a supporter of group 
patterns, but it can. Weblogs can be broadcast if used for one to many, but 
can also support social patterns.

Definition is right, though, because it acknowledges the nature of the 
problem. Groups are a runtime effect. We keep seeing patterns, as on the 
WELL, where the social software was built, and had unexpected results.

Part 1: Best explanation for 'a group is its own worst enemy' in a book by 
W.R. Byum, who worked in group therapy with neurotics. He found that the 
groups were trying to defeat therapy. "Are these individuals taking action 
on their own, or is this a coordinated group?" Couldn't resolve this, but 
lack of resolution is the answer. Humans are fundamentally both individual 
and social. Every one of us has a rational, decision-making mind, but are 
also able to enter into visceral bonds with others that would override 
rational, individual thinking.

Groups of people who have formal membership in groups clearly have a group 
cohesion, but Byum says that this cohesion is deeper, and kicks in stronger,
  than supposed. Illustration: You decide you don't want to be there, but 
you don't leave. Then one person gets up and gets his coat, and everyone 
gets up at once and begins to leave. This effect is so steady, it's called 
the paradox of groups. There are no groups without members, but there are 
also no members without groups. People decide at some subconscious level to 
do something that illustrates a group effect.

Byum found that the group has specific patterns in the way they tried to 
defeat his attempts at therapy. First was sex talk - flirtatious or 
salacious talk among members. Sex talk is always in scope in human 
conversation - one basic pattern they can always devolve into. Second: 
identification and vilification of external enemies. E.g. Open Source 
community against Microsoft (instead of doing stuff). Nothing causes a 
group to galvanize like an external enemy... even if someone is not an 
enemy, identifying one causes group cohesion. Third: religious cohesion, as 
with the Cult of Clay (funny worshipful moans from the audienece). E.g. go 
to a Tolkien newsgroup and say The Two Towers is a little dull, and see 
what happens. In most places you'll be flamed, because you're interfering 
with a religious text.

Groups sandbagging goals in favor of these patterns. The way around this: 
group structure (as in Robert's rules) is necessary to defend a group from 
itself. Defends the group from the action of its own members.

Communitree in the early 70s. Dialup. Throw off structure, and new, 
beautiful patterns will arise. This does work... at first. As time sets in,
  new patterns emerge. In this case, a high school got hold of a bunch of 
modems, and they weren't interested in sophisticated conversation at 
Communitree. They were overrun by these students. The place that was 
founded on Open Access had too much open access... free speech but too much 
freedom. They had no defense. The site shut down.

Was their inability to defend themselves a technical or a social problem? 
In a way it doesn't matter, because technical and social issues are deeply 
intertwined, and can't be disconnected.

Attack from within is the pattern that matters here. Communitree was shut 
down from within, and at the machine level, you couldn't distinguish the 
good from the bad.

Why people haven't learned within this arena: they're not reading stuff 
like Rose Stone's description of Communitree.

Lambda Moo: wizards decided to be involved only in technological stuff, but 
had to come back because they couldn't separate the technical from the 
social. They had to structure and run the community.

When you're dealing with people as one of your runtime phenomena, that's a 
very different practice.

In political realm, this is called a constitutional crisis. Constitutions 
are necessary elements of long term groups. As a group grows, the chance 
increases that they will look for moderation (a way to defend themselves 
from themselves).

Why important now? Observes a revolution in social software. The web was 
into a size thing for several years... loosely coupled and growing. Scale 
only worked if you werent' responding to users. The dense interconnected 
pattern that drives groups and coversations wasn't possible... we blew past 
the interesting midsize scale of groups wehre people could actually have 
conversational forms that can't happen with tens of thousands or millions 
of users. Now, all of a sudden, these things are emerging: weblogs, wikis,
  platforms (rss). Platform allows interesting developments at lower cost, 
like Stewart's Confab.

Why do we have weblogs now? Why did we have geocities instead? We didn't 
know what we were doing - it took a while to realize that conversation was 
better than pictures of cats.

Pepys weblog... Phil was asserting that weblogs would be around for at 
least ten years.

We've internalized, now people are building, and what they're building is 
web native. A weblog and a wiki is web all the way in... lightweight, 
loosely coupled, easy to break down and extend. RSS is a web native way of 
doing syndication.

We can now start to have a small pieces loosely joined pattern. Joi Ito's 
Emergent Democracy happening: conference call was lower bandwidth, so added 
chat, and then wiki... two realtime, one annotated. It's difficult to 
coordinate a conference call without interrupt logic - here that moved to 
the chat room, and the call flowed well. Could type urls in the chat or 
refer to the wiki. This is a broadband conference call, but it's not a 
giant thing. But this is an incredibly powerful (but light) pattern.

Final thing is ubiquity. The web has been growing for a long time... and in 
ways 'all' have access to the network. For some groups of people, everyone 
they work with and know is online. This pattern of ubiquity lets you start 
taking this for granted. We're starting to see software that assumes that 
all offline groups will have an offline component - that is lightweight and 
easy to manage.

As here - you assume people who assemble are both online and offline at the 
same time. Library of Congress meeting: SocialText set up a wiki to capture 
large complex bits of stuff. For future memories they take for granted that 
this will be the case. Shared ubiquitous computing can lead to new patterns.

What should we do re social software, knowing that a group can be its own 
worst enemy? Clay has been looking at this problem for ten years, and 
pretty hard for a year in a half. What makes community work online... he 
can say with confidence, it depends. (laughter)

Natural and supernatural grace... you never knew if you have supernatural 
grace or not. Social software is like that... something supernatural about 
a group working... more groups fail than work. It's like a cake recipe: 
nothing you can do to make it come out right every time. But here's a list 
of a half dozen things to do:

Accept 3 things: 1) You cannot separate technology from social issues. 
Separate tracks for technical and social, for instance, don't work. 
Mentions ymmv list: the conversation could not be forked. You cannot 
completely program social issues... you can't specify all issues in 
software. The group is real, it will exhibit emergent effects, and it can'
t be programmed in the software up front. 2) Members are different from 
users. Core group that cares more about the success of the group. They 
garden the environment to keep it growing and healthy. The software doesn'
t always allow the core group to express itself, so it'll invent new ways 
to do this. E.g. old hats mailing list outside a newsgroup email list... 
three tier system ultimately, also a young hats group. 3) The core group 
has rights that trump individual views in some situations. Voitng is a bad 
idea where citizenship is associated only with an ability to log in. 
Example: Tibetan culture newsgroup voted down beca! use of participation by 
Chinese students... sufficiently contentious groups would be voted away. 
Tyranny of the majority. Core group needs a way to defend itself as it gets 
started. Wikipedia has similar system - 'volunteer fire dept'... their 
leverage has kept it up despite repeated attacks.

These are things you have to accept.

Constitution: formal (code) and informal (custom).

What would you design for in social software? 1) Handles the user can 
invest in (not identity, which is a hot button issue). Anonymity doesn't 
work in group settings, but weak pseudonmity doesn't work either... we need 
to be able to see sustained identities. 2) Reputations are not linear and 
not portable from one situation to another and is not sufficiently 
expressed. Ebay's rep system starts with a linear transaction, so it works 
even tho linear. But you can't extend it. Reputation is stored in your head,
  but it must be associated with a handle. There should be a penalty for 
changing handles. E.g.: Casey Nicole story. Changing your identity is 
really weird, and when the community finds you've been doing it and faking 
it, they will expend astonishing effort to find you and punish you. Some 
systems have additional accretion so you can read involvement as an aspect 
of identity. Describes a sponsorship deal for member in good standing. 3) 
You need barriers to par! ticipation. It has to be hard for some users to 
do some things, so that core group will have a defense capability. Ease of 
use is the wrong way to look at this situation. The user is the group, not 
the indvidual. You have to find some way to protect the group. Soft forking 
(e.g. LiveJournal) - can be connected to multiple clusters. IRC chats and 
mailing lists are self-moderating... people drop off as the volume 
increases. Or Metafilter which occasionally turns off the new user page. 
Human interaction doesn't blow up like a balloon, it collapses.

There are lots of other effects, but these are primary.

The users are there for one another, so control is limited.

Q&A

Online community is more than message boards and chats.

Clay feels that online community is not a great term, because there is an 
offline aspect as well. It's online plus real world overlap.

(Find Tom Coates on online community ... reference in the discussion.)

Discussion of the 3D problem - it looks like the real world if you're drunk 
and in a cave.

Non-immersive software has tended to be more sustained than the immersive 
because it allows for the small pieces loosely joined effect. Immersion is 
not as universal a pattern for social behavior as people have thought.

How tune the software for the needs of the groups and users? Past model has 
tended to be architectural: make a place for people to be together. Now 
more of a shipbuilding pattern - build something and go there together. 
Wikis: general bag of capability that can be used flexibly. Optimizing 
software for particular shapes of tables... because of platform effect it'
s easy to build software customized for specific use.

Discuss Clay Shirky - A Group is its Own Worst Enemy: Social Structure in 
Social Softwar

--
Posted by jon lebkowsky to weblogsky at 4/24/2003 2:16:52 PM