All posts made in Apr. 2004:

Pinkish knuckles on the backs of my pale hands, knobbled. Knobbled veins too, standing out over the bones and tendons. Skinny fingers, boney finger knuckles. Weathered somehow, older and more tired hands than when I last studied them. Chewed fingers, pink thin skin rounds the nails. Flecks of ground in white paint caught on the edges of the bitten skin round my thumbs.

I never knew the finger biting habit was shared until I was told, I'd never noticed when I was young. But looking down at my hands as I washed them, I recognised them from over a decade before. The tiny pattern of the white paint on the bitten skin edge skin (a thin line on the top left of the thumb, a fleck at the tip), hands that look like thinking's being done with them. My father's hands.

On social software consultancy | With Jack Schulze, I've been doing a little consultancy at TimeBank recently on how thinking in social software, adaptive design and associated areas can improve a couple of their projects, and tie into their strategic thinking.

I'm not going to say anything about the projects or the outcome of the seminar day at the moment, just say what social software looks like, deploying in anger, a year in. This is what consulting and teaching non-expects in this area looks like.


Defining sosowa was the source of many posts this time last year. Having to compress it to be useful was a useful process. We're using something like this:

Social software's purpose is dealing with with groups, or interactions between people. This is as opposed to conventional software like Microsoft Word, which although it may have collaborative features ("track changes") isn't primarily social. (Those features could learn a lot from social software however.) The primary constraint of social software is in the design process: Human factors and group dynamics introduce design difficulties that aren't obvious without considering psychology and human nature.

This ties nicely with adaptive design, in that social software encourages you to fulfil latent needs first, then embark on not a development cycle but a dialogue with user concerns in which you listen to their emerging needs and implement them in code -- but you have to give users the ability to stretch the system otherwise you'll never even notice those new needs.

Areas of interest

To design effective social software, you should have some awareness in a number of areas.

So one of the things Jack and I are doing is producing a Primer in these areas to provide an overview but also a bibliography so the TimeBank team can dig deeper themselves. We've spent quite a lot of time running through their projects and these ideas to learn what's appropriate, where they need extra knowledge and so on.

Clay Shirky's essays (these in particular) over 2003 figure pretty big when the areas of concern to social software are summarised. That's not a surprise, they're great essays. But also, looking back, they're the only standalone, well-written essays there are. Outside the context of the early 2003 discussion, most of the weblogs posts just don't make any sense.


So already we're got a way to put the sosowa ideas into practice. I use this example:

We might consider the way, in groups of three or more, how there's always the possibility of two people being disloyal to the gathering, and how to moderate that behaviour. In the physical world, disloyalty is visible because all interactions within the same context are visible to all local group members, and the disloyalty is moderated through politeness and social pressure. There is a different mechanism to exert social pressure in a peer-to-peer small group (of 6 people, say) than a broadcast larger group (a classroom type situation). Contrast this with online small groups and we see misunderstandings by email, people being left off cc lists and so on.

Now that's become a solveable scenario. We need mechanisms in the online software to bring in a similar incentive structure to the offline world.

The single most useful piece of thinking I've been using is Stewart Butterfield's March 2003 post on the devices in social software, mechanisms successful pieces of social software tend to have.

  • Identity
  • Presence
  • Relationships
  • Conversations
  • Groups
  • Reputation
  • Sharing

I'll describe each of these, as I see them, critiquing AOL Instant Messanger (just as an exmaple), and then describe how we put them into use.

  • Identity | Your identity is shown by a screenname, which remains persistent through time. There are incentives not to change this, like having your list of friends stored on the server and only accessible through your screenname. This acts as a pressure to not change identity. Having a persistent identity is more important than having one brought in from the physical world.
  • Presence | Presence is awareness of sharing the same space, and this is implemented as seeing when your friends are online, or busy. AIM isn't particularly good at group presence and visibility of communication, although other chat systems (such as IRC and early Talkers) use the concept of "rooms" and whispers.
  • Relationships | AIM lets you add people as buddies. From that moment, their presence is visible on your screen. This is a relationship, you're allowed them to have an effect on your environment. Not terribly nuanced however.
  • Conversations | Conversations are implemented as synchronous messaging. There's a difference between messaging and conversations. Messaging is just an exchange of text with no obligation, but conversations have their own presence and want to be continued. AIM does this by having a window for a conversation. It's difficult to drift out of it, it hangs there, requesting you continue. Contrast this with email which often is just messaging, and conversations die easily.
  • Groups | AIM isn't great at groups. Although you can have group chats, the group is transient. People have more loyalty to a group when there's some kind of joining step, when they've made some investment in it. Entering a window just doesn't do that, and there's no property of the group that exists outside the individual user's accounts.
  • Reputation | Reputation is used more in systems which allow meeting new individuals. AIM's simple version of this is "warning". Any user may "warn" any other user. A users total "warn" level (a figure up to 100) is shown to everyone they communicate with. Unfortunately, it's not a trustworthy reputation system, and reputation is notoriously difficult -- but humans are great at dealing with it themselves, given certain affordances: persistence identities, and being able to discuss those identities with other people. AIM's simplistic relationship system makes reputation not so important though.
  • Sharing | People like to share. With AIM, sharing is often as simple as giving a friend a link to follow. Other systems, such as Flikr, are about sharing photographs. These act as small transactions that build genuine group feeling.

Clay's essay A Group Is Its Own Worst Enemy provides a great overview of many important concepts here, especially on having a cost to join (which provides the feeling of membership). Incidentally, knowing this is a fantastic rule of thumb: Forums which require a joining step are better that forums that feel like window-shopping. It's a more social design decision on the same level as not putting the reply box at the top (so the user has to read the whole conversation).

Putting this into practice

These seven items act as a tremendously useful framework to critique and comment on social software. We combined them with a further process:

  • Define the goals of your software (the strategic goals, and the user goals)
  • Consider incentives: For goal-oriented systems, why do people take part? Why do they come back?
  • Consider moderation: If you have a disloyal user, how do you stop them starting fights and poisoning the community? If you block them, you might enter an arms race.

We took a dynamic approach to this. It's very simple-minded design which only gets a person to use the site once, because it's easy to operate or whatever. A more dynamic approach is to consider the consequences, and use mechanisms such as people like sharing to design in the appropriate incentive and moderation feedback loops.

And dynamic's the key. We've attempted to understand user flows, and apply the ideas above to different types of user within the system, using one or two of Matt Jones' UCD tricks (they're handy to use and easy to teach). We've been teaching the technical and other project team members to use these tools and ideas to develop a common vocabulary, and to use social software ideas to make simple social tools more likely to succeed.

(So, for example, we can see that some users aren't going to be comfortable approaching people online - they're educated, highly intelligent folk who aren't good with computers - then we can use mechanisms such as slowly building relationships (making use of presence) and reputation to build an approach with a gradual gradient.)

And then of course it all gets into adaptive design and development processes, and also onto the specific small touches we recommended for the individual projects, but I'll get onto those another time.

I've put many of the social software links we'll be using in the Primer in my links directory.

Of course half-way through gathering these, the Many-to-Many Social Software Reader (and Timeline) launched. Handy, but unfortunately not completely appropriate -- I think they're aimed at providing a history and resource for practitioners, and not a quick way in to pick the most appropriate ideas for people in other fields. The latter is what Jack and I are providing.

Comments always welcome.

The cream blinds against the windows against the white walls of my new bedroom. The bright morning sun projects an image of the east-facing window sharp onto the blind: condensation patterns, each droplet a halo, rough edges of the frames where I wasn't neat with the gloss, an xray edge-filtered. The south-facing window is more diffuse. A subtle shadow grid intersecting blurred rectangles. A milky Rothko.

Leif Inge's 9 Beet Stretch is being performed in San Francisco, in its entirety. If you're a local, and free for 24 hours or so following 10pm next Friday (23 April), then go. Beethoven's 9th symphony timestretched to 24 hours. Each movement lasts about 5 hours. 2 minute crescendos now build over 20 minutes. It's slow. The performance is a sleepover.

I listened to the first four movements on my train ride from New York to San Diego (for Etcon, this year). My power ran out, but I'd listened to the 5th before. Over that length of time, you can't pay attention. The music affects your consciousness without you noticing, shapes your thoughts. It's impossible to pinpoint anything about the sound itself, only your own cognitive state. I made notes immediately afterwards. For a nearly a day's worth of listening, this is all I could muster, summarising each 5 hours chunk in a sentence:

1st movement is excellent, near the beginning, very exciting. End of the 2nd movement is scratchy. Not so keen on the 3rd. It seems quiet. The 4th is dark to begin with, then sad. I've heard the 5th before, but not this trip. I remember it being vocal and dramatic, especially at the beginning and in the middle.

(But I remember, within it, there were long portions where I was rapt, carried by the music, and even five minute transitions that were especially touching (that I'm not actually sure I could hear, now, outside the context of the movement).) And then:

9 Beet Stretch, like this long train ride, isn't a spectacle to be taken in and anticipated, but rather a smell: a field of potential to be used as extelligence, to be lived, the journey being the destination.

You can listen to 9 Beet Stretch yourself, in streamed Real, or download it as mp3s.

(More background and quotes from Inge at the New York Times: Beethoven's Ninth Around the Clock, What you hear in normal time as a happy Viennese melody lasting 5 or 10 seconds becomes minutes of slowly cascading overtones; a drumroll becomes a nightmarish avalanche, and the trance state induced by listening.)

(Packing yesterday. Moving house. Objects going into cardboard boxes, boxes stacking on top of boxes. Packing isn't random. Objects have shapes, and some things aren't really objects: knowing that I need to pack my jumpers together means they comprise a population. They'll stick together if they can, but will be split up if necessary (it doesn't matter that they're together that much). They afford padding delicate things, or wrapping multiplicities together. As do small cardboard boxes, which are used to keep the aromatic oils from rolling around, but which can also be put in other boxes. Lamps ask to be wrapped in tshirts, and ask so hard that they're as inseparable as a pair of shoes (you could separate them, you'd be a fool to). Ornaments ask to be separated by clothes too but because they're smaller, all separated yet in the same box (you wouldn't want to find one by surprise and break it). A box's shape asks for a book in the certain size range, the actual book resonating and asking for more books of the same size. Size matters, subject matters. The shape of the free space changes, the ideal book to fill the hole changes, and the aggregate weight acts as a stop. Bulk properties lead to heavy boxes, fragile boxes, those without lids, those that can be turned, and filling the van we play the same packing game again: each packed object influencing the space and the shape and what can be put in top.

These are local relations, and then (at the van level) more, emergent, local relations (abstracted by the cardboard). As solid as the surface of each thing is, its objecthood bleeds out into proximate neighbours, a three dimensional jigsaw of affordance niches, all filled. I can no longer refer to the table lamp, it's the table lamp shirt box-with-wooden-game more shirts not-heavy assemblage. The reality of packing has forced me to stop thinking of units, not even to regard the mass of my belongings as an association network (what does it matter that my clothes are in this box and that other box? It's obvious when I look at them), what matters is the immediacy of packing/unpacking, the local actuality of the so-far packed population: it's an extremely dense local-link network, and as Steve Grand (talking) referred to the brain: not a network of modules and circuits (objects) but more like a gel, the kind of medium in which broad phenomena can occur, fields, waves.

And indeed that's what happened (what's happening), unpacking. Before moving, the order of unpacking would have been the reverse but otherwise identical to the order of the packing. But the new house has provided a new ecology of niches, in new positions. In their new configuration, in the new environment, the boxes are held in tension between these two surfaces: the packing order encouraged by the shelves, layout of the old place; the packing order encouraged by the new.

Set like this, it becomes inevitable. The objecthood smeared out across the packed gel produces interference patterns, waves. Each box acts as a cavity, bouncing the ripples back and forth, amplifying, and out of the grey primordial packedness forming (objectifying) things, objects, belongings once again. The constellation of boxes likewise, the properties (on top, behind, can't reach, teetering stack) form circuits, switching gates: they act as channels, opportunity-to-release-objects rushes, converges, highlights single items: unpacking is easy, it's a matter of listening to the packed. Like the carpenter who sands with the grain (there's no rule that says you have to, you can sand against the grain, it'll just look terrible), we know what we have to do.

The packed becoming unpacked. It speaks to us. The affordance-gel serialising into a steam of newly minted objects, freshly abstracted and given independent surfaces, emitted one by one like puffs of steam from this opportunity machine. As it unfolds, this ribbon encodes the intelligence of the assemblage, like a computer chip that does nothing static but collapses dynamically into work.

It's just a matter of reading it. By reading the collapsed string, we can see, like a hologram, the layout of the old house and the belongings arranged in it: we can reconstruct the mindstuff of the packedness, and the thought-processes that occupy it, the same thought-processes that are/were isomorphic with the configuration changes inherant in the unpacking, our hands as the mechanism of its thinking. And we can read it. Mind reading.

The first object to be reformed/regurgitated was my camera. As there are four unpopulated rooms, take the fourth letter "e". Books I need to consult come next, and those went on existing bookshelves in the house, so fourth letter again. "K."

The dresser and bedside table arrived at this point, from outside, which changed the shape and occupied the bedroom, and so altered the field defining the [static, it's only really thinking-conscious while dynamic, being disassembled] conscious state of the packedness. So for the next item (statements, in boxfiles), take the third letter, "a".

Third letter also from hammer ("m") which we used to repair the moved bookshelves. A remarkably introspective thought, given the outward-facing nature of previous steps of unpacking.


Which appears to be Sanskrit, part of a mantra, meaning one, without a second. Where a second cannot even be contemplated, not even a virtual second, like "the number 4" I guess, or the isness. (And unlike the Earth, because a twin Earth can at least virtually exist.)


I must say, I didn't expect the process of unpacking several hours worth of belongings to be speaking to me in Vedic Sanskrit, much less to be reminding me to regard the matter in my hands and in my house (and the happenings, and becomings) as part of a single - and all-encompasing - isness, just as simultaneously (and in contradiction) the packedness is dissolving its haecceity [deferred] into a linear stream of freshly manufactured objects, condensed out of the gel, objects that display the very opposite of ekam. Well, I guess that's why it needs to be restated, and anyway it's good to be told.

Ekam, backwards, is "make." Which is presumed to be significant.)

It's not just primates that react to babylike features: some songbirds can be duped by fish who rise up above the water's edge and gape like baby birds. A gaping mouth is the feature that elicits parental behavior in birds, the avian equivalent of big heads and large eyes, and so the poor birds dip down and stuff food into the mouths of the fish instead of their own young. -- The Other End of the Leash by Patricia McConnell, page 123.

The final curious fact about a dog's vision is that their visual actuity is sufficient to resolve the European transmission frequency of 625 dots per second into a visual image. American television transmits at 525 dots per second, not quite fast enough for the canine eye to see anything other than a screen full of fast moving dots. -- The Dog's Mind by Bruce Fogle, page 35.

Take the Martian topology (datafiles captured by the Mars Orbiter Laser Alimeter, currently making a height map of the whole planet to within 5 metres), put them into the Terragen scenery generator, then add in a bit of speculation: get these beautiful renderings of a fictional Mars with surface water [via More Like This]. Olympus Mons with an ocean. The coastline of the Deuteronilus Mensae.

(The pattern reminds me of the Atlantic coastline of Long Beach (maps, aerial photos), which I saw flying into New York earlier this year, a wonderful mix of curvy inlets, long spits and fractal islands, so unlike the coasts of anywhere I've seen in the UK.)

Loosely joined griffins everywhere.

How dogs perceive the universe.

There's a different between using smell as the primary sense and vision. Vision is all about surfaces, about being outside. And it's at a distance too. We take input, then cognition occurs, and then act. (That's not to say that sensing and locamotion are separate, they're not.) There's a self there.

But smell. Vision first. If you see a lion, you can do something about it before you have to deal with it. You can hide. Or run away, if it hasn't seen you. But there's thinking time. Vision decouples sensing from action. And it's very binary: it's only an edge case where you see a hint of something. Sense exceeds reach; vision affords anticipation, apprehension.

But smell! Smell is all about hints. You don't smell a lion, you smell 70% of the likelihood of a lion -- is it nearby in space, or in time? How close is it? How much does it smell? There's no chance to think about it, you can't hold the sense-of-lion at a distance: even 10% of lion is 10% chance of getting eaten. Wherever you are in the field, there's the chance that something will happen.

There is no cognition step between sense and act with smell.

Smell is all about moving through the insides, through a field of intensities, of potential. There's no hiding.

For a dog, the sensory input of smell leads to actions (or potential of action) so thoroughly that it comprises a large portion of the mind of the dog. The smellspace is not just input, it's the beginning of output too. So when the dog walks along, it's like they're thinking. It's not undirected thought by any means. A dog can decide where to walk, so moving equals thinking.

Navigation is cognition. But there's no concept of moving. Just being (continuing).

For us, this is almost like reading (especially reading someone like Markson). You can't read and articulate thoughts at the same time. The input is how your minds moves; the book comprises part of your thought processes. And because you decide what to read, it's like a dog exploring the landscape.

Of course something there's an unexpected scent. Or an event unaccompanied by a scent. To a dog, this is like a burst of intuition, or a joke. If you creep up on your dog, and surprise them by tickling, that's a surprisingly witty remark (or just rude, I can't tell).

We live in a world dominated by vision. Vision is directed, so we look at something. And because of that decoupling, we can decide what to do, how to interact with the thing we see. That's what gives room for hands, for manipulation. Rather than locomotion (which just moves us along a sensory gradient), we have reach too, which lets us perceive with better detail without moving the body.

We live in a spectacular world, where what can be sensed on the surface dominates (because it happens first), and what can be reached/affected is often neglected. A spectacular world neglects the capacity to be affected.

We think we live in a spectacular world. And unfortunately our artificial systems (cities, cyberspace) take as their point of construction something to look at but not adapt. (We bring cities to life when we inhabit them. Ditto becoming embedded in cyberspace. But it's a stuggle.)

I'm not saying it's a bad thing. Spectacle is what lets us say that plants can be dug up and put in a place together (that the land and the process of growing are separate from the growth of the plant). Spectacle lets us abstract, and claim that a dog's nose is like a man's nose. Spectacle gives us science, and speech, and maps.

Dogs understand that the world is part of our mind. The world as extelligence. Dogs are able to inhabit the same world as one another, to follow one another's thoughts. They don't just explore the environment, they are literally of it, encoded in it. The dog says that the map is not the territory, however much its spectacular surface convinces us. The dog says that there are no surfaces, just occassional strata of unusual density, and that there can't be abstaction because deep isomorphism is always required.

Third order cybernetics.

Some books, and links.

I think of Dennett saying that for early bacteria, sensing a chemical gradient is isomorphic with doing something about it. (Do you sense something if it doesn't change you at all?) Ben in those dinner conversation notes: their cognition is smeared out across the ocean itself. the brain is not a locus of the intelligence. Clarke and Chalmer's The Extended Mind [my notes]. And Gibson (p226): This is why to perceive something is also to perceive how to approach it and what to do about it. Information in a medium is not propagated as signals are propagated but is contained. Wherever one goes, one can see, hear, and smell. Hence, perception in the medium accompanies location in the medium.

Oh, go watch a dog.

I'm not very good at generating random words on demand. Count 1, 2, 3 (at 1hz), then say one: artichoke. 1, 2, 3, pterodactyl. The problem is, you get used to listening to your brain queueing up the next word, and you feel the association vectors turning into place, or the sound at the beginning just starting to form. The rule is, you're not allowed to say those words. You mustn't even have a sniff of the word until you say it. 1 2 3 cavernous. That's why the pauses are so difficult, because they give you time to think. Excitation and lateral inhibition, the increasing dominance of a single idea, and the growing difficulty in finding new ones. 1 2 3 slat. Possibly you can begin to approach a state of immanence, although the tempation is always to look back to find hidden patterns. There's a technique, certainly. 1 2 3 furniture. 1 2 3 blackbird. 1 2 3 ampere.