digitaldust

Over the last couple of days I’ve had the strange experience of people (who are not Heather) talking to me about my blog posts. Mostly about the cats, even someone on the other side of the world coming up and asking me if we had a name for Burton. Yep, we do, he’s an explorer, named for Richard Francis Burton. And it’s quite apparent that no one comments on or talks to me about the turgid attempts at high brow posts I put up… those just seems to scare people. Just goes to show it’s the human angle every time, reality TV, not docs. Also, I’ve put some a couple of posts up in a bit of an un-reflective and unprofessional way, without thinking how they might be read from other points of view and in my new context as a lecturer. I took them down. So I’ve become painfully aware of the public nature of my blog. Only a very small public, I’m not claiming fame, just unfortunate recognition. I suppose that’s what one wants when blogging, but I had always sort of assumed that the only people who read my blog were Heather, Ed and my Mum… Hi Mum.

On top of that has been some discussion with a couple of people about what I would use Facebook for. Facebook (as well as many other social networking or social blogging tools) allows selective privacy/publicity. So I could limit my audience to people who might get obtuse jokes or forgive me for obsessive cat postings. However I consciously chose to join Facebook as a tool I could use to connect with students, not as something to keep up networks with friends.

But all of this raised again for me the whole social construction of knowledge thing. How meaning is constructed in a social context, and with very little to do with the content of the message. I’ve been thinking about both Facebook and the blogosphere in these terms over the last day.

Profile/Social Networking sites, like Facebook, run on this concept of social constructivism. The so called content is light, but they are not about that, they are about links to other people or saying what movies one likes. From the constructivist angle these profiles, when filled out, become very rich nodes in the framework due to their relationships, but not because of any unique or creative content. And from that point of view you cannot treat the individual nodes separately, you must always be addressing the whole. Or addressing big chunks, which is inherently difficult as there are no logical boundaries in these networks.

The same goes for the blogosphere. Most of which is not filled up with unique, new, creative content. Much of it is links other posts, which again are links to further posts, which might end up in some unique content.

So to me both these experiments in closed social networks and the open knowledge networks out there on the web seem to prove and should be read as socially constructed and intrinsically hermeneutic knowledge.

Networks, communication and problem solving:

Sometimes problem solving in groups of people can be harder than individually and sometimes it is easier. What is the difference in the problems, what characterises one from the other?

Everyone knows (well at least I do and I assume everyone else does) that as a group gets larger it’s average IQ drops. It’s capable of doing more, but in less smart ways… Many hands make light work, but too many cooks spoil the broth. Small groups of people often seem very able to take on large organisations, solving their problems by being very creative and agile, not brute forcing things through. I’m thinking here of comparing, say the BBC, with companies like Six Apart, Ludicorp (flickr) or even Google who took over the world with a couple of dozen people.

There are a bunch of other reasons why these people can create global phenomena, not the least because of the Internet, egalitarian tools for creation and green fields development. And I’ve not listed the hundreds if not thousands of failures either.

Much of this is down to the team effect. Organisational studies, management science, knowledge management, and all that malarkey know about teams. But why do teams work compared to organisations? Also why do teams fail to make a difference within an organisation? I’ve seen small companies and a monster one and even in the small ones, each team truly feels like the others are working against it, not with it. Is that a mislead perception or does it hint at the truth?

So what is the difference between a small group of people and a large one creatively? I think it must have something to do with the network effects; transmission and loss of information. Can this be solved by understanding the nature of networks of communication and information theory?

Is this anything new? I don’t think so but has anyone researched this before?

I’m writing up a spec for a “Trust” system to power the Beeb’s message board system. It occurs to me that Reputation is now very 2003, which is why it seems to now be all the rage here among BBC management. The BBC is never with the state of play in any of it’s media, it’s always a few years off… sometimes way ahead of their time, sometimes woefully behind.

Anyway, Reputation and Faith. Or more accurately I’ve just had a bit of a epiphany about this whole rating/recommendation/reputation thing, that it’s not about creating reputation systems it’s about creating epistemological systems, simple ones mind you, but at their heart they are epistemological. All the debate feels very similar to old school philosophy of knowledge.

I was writing my usual stuff about trust living on top of reputation and suddenly thought about faith in the philosophical sense. I got to thinking about the relationships here. Knowledge being a justified true belief and faith being a belief that is not justified, and it’s truth is unknowable.

This lead me back to thinking about what reputation is in the sense that we are trying to use it in social software at the BBC. And it is a simple epistemological system. I don�t think in that the above definition of knowledge is right in a practical world. It occurs to me that knowledge is a probability function around repeated experience, and belief is where the probability approaches 1.

I’m going to have to think/discuss this one more I think. Back to work for me.

Well, I may be being too harsh in the title, but Reed’s Law and Metcalfe’s Law are both slightly wrong. They both treat the subject at hand, Social Systems too simplistically.

One of the phenomena of social systems is that they undergo connection reduction when they reach a certain size. Luhmann identifies two levels of scale in social systems. Simple systems where every actor in that network is connected to ever other (a simple network graph, which does match Reed’s Law). But when they reach a certain size they have to become complex networks; the linkages are not evenly distributed and not every actor is linked to every other. This in my opinion happens when a social group gets above about 150 in the real world (maybe a few more when computer supported). This is your tribe, or wider circle of friends. Beyond this point clustering happens. So in a network of up to 150 then Reed’s law applies, beyond that it is 2^N less some level of conection simplification.

There needs to be some kind of expected value or statistically reduced value of the links that could possibly be made above a certain level. Social systems do not split evenly down the middle when they reacha certain size, they end up as a complex but not fully linked system.

Even Metcalfe’s law seems to me to be mistaken. The network of communication is only as valuable as the number of people you want to talk to. So althought the global communication network is constantly adding people in developing countries the value to the network to me isn’t increasing. If the people I want to talk to are hooked up, and all the people I could ever really want to talk to are hooked up then the value to me is fixed. I think that Metcalfe’s law is stuck in first order cybernetics. The value of the system needs to be worked out in a slightly different way, and taking into account the observer; the person making the value judgements.

Robin Hamman sent on this paper this morning. Interesting in depth statistical review of Slashdot’s moderation system. Some previous knowledge of it is useful before reading.

Slashdot’s moderation system is not what I would think of as pure moderation it is more of a rating system. Slashdot don’t delete.

A few pull quotes…

Part of Slashdot’s ethos is that posts are never deleted from the site.

Of users who commented 41% moderated.
Of moderators 68% commented while 32% lurked.

Of users who commented 31% also meta moderated.
Of users who meta-moderated 66% also commented.

Only 28% of comments received one or more moderation action.

The median half-life of a conversation on Slashdot was 174 minutes (the time taken to receive half the total comments on a story). The median to get 90% of comments was about 18 hours.

90% of anonymous posts either stayed at the same rating (zero) or went down.

Most importantly the design recommendations that come at the end are worth reading.

• Worth thinking about how quick you want the community to self-moderate or self select.
• It seems that, in the case of Slashdot, the time taken for moderation/rating to be effective is longer than the half-life of the conversation.
• Nearly 3 / 4 of the postings don�t get moderated or rated at all

Slash(dot) and Burn: Distributed Moderation in a Large Online Conversation Space

I was just thinking about messageboard moderation and reputation. I’m off sick so I actually get the chance to think things through. Actually I was thinking about moderation, reputation, the Soham murder case - Ian Huntley’s history is now coming out, privacy, big brother states, anonymising democracies and the possibility of open reputation based, whuffie style societies.

There are two schools of thought as far as reputation and message moderation goes. One goes “You should moderate every message based on the content of the message alone” and there is the other that says that “You should moderate each message within the context of it’s poster’s history.�
(more…)

[cond-mat/0205383] Identity and Search in Social Networks on arxiv.org.

Mathematical analysis of an old sociology experiment; How can individuals in a social network pass messages to each other when they are unaware of the links between them. Most relevant for P2P networking and P2P search.

It uses some network theory and probability to solve this. Depends primarily on two things. That individuals in the real world cluster on more than one thing. Each node (person) has a number of dimensions of similarity/dissimilarity, for individuals in a social world these could be things such as profession or geographical closeness, plus many more besides. All of these together give a vector of similarity/dissimilarity for each node relative to each other. Two people can cluster round either geographic similarity (live on same street) or professional (work at same place), so either of these dimensions can be used for similarity. We don’t have to perform any triangulation to determine distance. Thus each individual node can cluster with more other nodes, and share multiple networks. (This sharing of networks is one of the most interesting aspects of the paper, one not examined).

Its also relies heavily on a low probability of chain termination. In automated systems this can be set arbitrarily low. So little problem here.

Just read a few reviews of the Sims online. It’s not getting much good press. Seems like the Avatars are not especially individual. Not enough ability to differentiate. After the experience of the offline version people expected more.

PC Game Review
Amazon

Will try and read some more and post a bit more to this.

I just remembered about this article on Kuro5hin about reputation. “Pondering Digital Reputations”.

Worth a read. Drifts on and off the topic of completely digital reputations as I’ve found everything does. When it comes down to quantifying reputation it becomes quite dry. Which is fortunate of unfortunate depending on ones outlook. Mostly this dryness is due to the dimensionality reduction that goes on when one converts from a fuzzy real world brain based system to a limited exact digital system. The price you pay for switching from a system based on millions of years of evolution to one based on less than 100 years of rational research.

I wonder if there is some kind of proportionality between what evolution can achieve and human research can discover.

I’ve given this talk a couple of times now and have messed around with the presentation based on how it was received. So here’s a link to the new one.