Clay Shirky Social Software "hello everybody. welcome to my cult." been reviewing the literature since the 1970s, pattern: "a group is its own worst enemy" and he's going to outline the main challenge he sees for largescale social software his definition: software that supports group interaction internet supports lots of interaction patterns: one-to-many outbound, many-to-many prior to the internet the way people communicated was the table. conference calls don't work right. we're still learning what works. easy routefinding isn't here yet. "software that supports group interaction" is an unsatisfying definition because there's no archetype. email isn't it: it's point-to-point and two way, mostly. it doesn't *necessarily* support group discussion. and weblogs are mostly broadcast. a cluster of livejournal weblogs are social, on the other hand. -> "groups are a runtime effect" you cannot in advance specify what the group will do pattern: . build software . group comes along and uses software . we're surprised by what the group does this happens over and over again three parts of this talk * one how and why group worst enemy * two why now? what's happening now that makes it worth talking about * three identifying things which are core to largescale social software so, onwards. one: "WR Bion": book from mid 1900s. groups of neurotics conspire to defeat therapy. not deliberate. but whenever he did anything, the group would tend to work against it. do you view groups as aggregations of people, or as entities in their own right? answer -> BOTH. necker-cube-like. Bion's thesis that group cohesion occurs much earlier than we'd expect... even before people are labled as being in a group (like, you say you're in a guild for an online game). then it's clay's example about being at a party. even when you want to leave, you *don't*. group cohesion; social stickiness. so there's a certain effect with groups: they protect the group Bion also says that there are basic patterns implicit in groups: . sex talk (you'll always see this in IRC channels, regardless of topic, overt or otherwise). pairwise bonding [now, i'm sure i've heard of something like this before. damn. tip of my brain] . the identification & vilification of external enemies of the group (eg opensource community in the 1990s. loads to do, but slagging off microsoft). this produces a pleasant feeling of group cohesion . religious icon [audience: "claaaay, claaaay"]. group structure exists to protect itself, to stop it sliding into these basic (earlier) patterns. early BBS: Communitree, 1970s. adults on the board were overrun by school kids (mainly boys) talks rude nonsense. too much open access -- Communitree couldn't defend itself from the action of its own users. -> but this was a requirement. q: was this a technical or a social problem? clay's a: it doesn't matter, they're closely intertwined. they designed this system but were unable to defend it. Communitree was shut down by people logging in and posting, even though that was what it was designed to do. [-> groups as a runtime effect, i guess] other people don't read this documentation. "learning from experience is one up from remembering." learning from reading is way better... but we don't do it: lessons from Habitat (good doc, 1990) reads just like Communitree, from the 1970s. another one "lambdamoo takes a new direction": the wizards say they're only going to do technological things. then 18 months later, they come back as a government otherwise it wouldn't work. people working in social software are more like economists, scientists rather than people writing compilers. cohesive groups tend to *ask* for additional structure. TWO - why now? clay says there's a revolution in social software. the number of people writing things to help groups is astounding. maybe because the www was all about making things really big, so we tried to make huge groups. then the downside is that small groups have a *really* different form of action, but we just went straight past it. now suddenly we're getting the small groups we skipped before: rss, weblogs, wikis. it's cheaper and quicker to build stuff now: ludicorp's Confab took only 2 weeks to build. clay makes a good point: weblogs aren't about technology. we've had that tech since day 1 of mosaic. why didn't we get weblogs then, why did we just got Geocities? because we didn't know what we were doing. why was there an 8 year gap between a forms capable browser and phil's pepysdiary (which will be around for 10 years)? because we needed to get used to it. a weblog is web native. it's not a crazy huge lotus notes thing. and then of course it's the small pieces, loosely joined culture. he's talking about a meeting (joi ito and others) being a conference call, then plus a chatroom, then plus a wiki. [this is the curious thing. why was www successful and not other hypertexts, etc. we talked last year about what made certain apps/platforms successful and not others? will kay's open croquet be successful? and maybe this is why this *type* of technology is successful: because you can use 3 things simultaneously, and the developers don't have to be socially aware enough to choose those channels in advance -- that Kay can obviously do very well. but other people can't, because we're not good enough. this is why we need that social rhetoric. to democratise clay's or kay's social awareness.] other reason for now -> ubiquity. in certain situations *all* people have web access, so you can start taking certain things for granted. oh, there's this kind of ubiquity too. face to face *and* IRC/Wiki at the same time. library of congress meetings now have wikis, like ~permanently~. THREE - what should we do? q "what is it that makes a large, longlived group successful?" a "it depends" . natural grace (get to heaven by doing the right things) . supernatural grace (as above, plus being annointed by god) the normal experience of social software is failure. map out yahoogroups subscriptions: it's a power law. there are about a half dozen thing universally true of online constitutions/ software that supports successful groups. if you were doing to design a piece of social software, you have to accept: . you cannot completely separate technical and social issues (you can't just have two mailing lists to break into technical and philosophical, for example). everyone has the epiphany: the software shapes what people do! but only up to a limit. emergent effects, the group can't be programmed, etc. deciding what value is and defending that value -- put that in the control of the group, not the software. . members are not the same as users. there is some group of users who cares more than average about the integrity of the group. this is the core group. they garden [oh yes] the environment. the software must let the core group express itself, otherwise it will find new ways to express itself. (eg, a usenet group setting up a mail list for meta discussion). "members in good standing" . the core group has rights that trump individual rights in certain situation. this pulls against the one-member-one-vote system. tibetan discussion groups being voted down by chinese users on usenet. the core group needs to be able to defend itself. [sounds like starship troopers and the ethical ladder functions.] -> formal constitution is instantiated in the code. the informal constitution is the "way we do it round here". what would you design for? 1 a handle for the user (not the hot-button "identity" issue). a handle that matters. [this is true. if the technology doesn't guarantee a single unique handle, then why bother building things that assume it.] -> reputations are not linearisable, and not portable reputation systems are poor, because the best one is in your head. eBay starts with a linearisable transaction, which means that can do reputation system -- but it's not social software. so you tax people for changing handles, because they're necessary for the group to work. kaycee nicole teaching us that violating handles is viewed as a major trangression. the community will punish [like the way people punish for lies, even in one-shot games, even though punishing doesn't matter]. 2 design a way to have members in good standing. most reward people [like those b3ta logos]. he mentions a good reputation system: your username *includes* your sponsor's username. if you defect and are booted, your sponsor is also booted. 3 must be a cost to join or participate. segmentation of capabilities. it has to be hard to do some things, otherwise core group can't defend itself. (moderation, for example, you have to be around for a long time to be allowed to do it.) ease-of-use shouldn't always be calculated from the individual's perspective. 4 spare the group from scale. scale kill conversations. metcalfe's law "is a drag". density of conversations drop off very quickly with scale, but good groups rely on density! . livejournal does "soft forking". average size of a group/cluster is about a dozen. it self-clusters. . IRC channels are self-correcting. if there are too many people, people drop off . MetaFilter turns off the new-user page. it raises the bar to entry. makes it better. what happens with scale? the group collapses or turns into broadcast. [<- I like that line very much. waveform collapsing metaphor] more interesting things: conversational artefacts, leaving behind stuff, like wikipedia [the present encodes the past. like cities]. Coates gets a mention on the subject of social software/ communities http://www.plasticbag.org/files/comm/the_excesses_of_social_software.shtml and it's not just online communities, but also online PLUS real world. q: why don't we have VRML yet? a: because VRML has the capacity to make *any* computer run slow. "immersion is not as universal a pattern for human behaviour as we though it was" q. round tables, square tables. what's the social software analogue? a. - it's been architectural so far. maybe that's not a good metaphor - we haven't been very good as goal-oriented software [tellic groups! oh yes. that's what this social rhetoric is all about] "optimising social software for different shaped tables" because it's so easy to build now, it's worth building for a specific purpose. it no longer needs to be all things to all people.