Large scale social networks panel

Anil Dash:

when typepad started most people had ratio of blog to readers – 40:1 now it is closer to 1000:1 at Typepad.

LiveJournal – huge community – primarily or 2/3rd woman 18–24, very geographically and technologically disparate. Communities like this allow interest groups (That guy who loves cars can talk to car lovers) without boring people. Live Journal ratio of writer to reader is about 1:8 – much smaller ratio.

Almost no one on Live joural has more than 150 connections in an organic fashion.

Many people check their live journal “Friends” page many times a day– to see how friends are doing, what they’re writing about, if people left them comments, ect.

I realize I’m more interested in listening to Anil than taking notes now.

 Christopher Allen: Life with Alacrity

The Dunbar number – here’s Chris’ blog about this. Smaller the neo cortex in a primate, the smaller the groups people collect in.

People can handle about 147 to 150 people in a group on a regular basis. -this tends to apply to nomadic tribes, sustinance villages, terror organizations, ancient armies (Roman Centurion – 100 soldiers).

To maintain a group of this side, you need to spend 40% of time social grooming – glancing, listening, talking, responding. If you don’t do this with a group, you can’t maintain the connections.

Groups don’t scale – what are the limits to group size in modern culture – based on Christopher’s research –

from Online games – ultima online – guild numbers – most guilds are 37–85 in size, 61 is the major high point in a histogram. World of Warcraft – Xerox Parc – 250k users, all data for month – how did groups interact – Subgraphs – max is 6 for a particular guild – for all guilds – it is more difficult to get people together after around 77, and very large groups it is very difficult to maintain cohesion. Easy to form group of 10, but it gets harder after that.

Groups that behave in particular way – between 5 and 11 (7 is the peak) and another one at between 50–70, but around 12–15 there is a problem with 12–15 – the ‘judas number’ small group dynamics don’t work and large don’t work – it is a dangerous time in a group.

Too small– insufficient critical mass, unable to sustain conversations, group think or echo chamber, feels like you’re alone.

Too Many People –

Too noisy, not enough signal, lack of trust or unequal trust, clicques and bad gossip,  inappropriate politics

Other social contract failures – flames, trolls, tragedy of the commons

 As software gets to understand this number – they’ll be able to design for it. One example: YackPack – BJ Fogg – software for intimate social networks.

Electric Sheep:
A cyborg mind – 30k computers, plus people and a genetic algorythm. Runs on evolution and mathematics.

It would have taken 100 years to create this video using single computer, but with collective power, you can get it in a few months by itself.

It is a distributed screensaver using same architecture as SETI at Home. This project doubles every 9 months. The animations are each called sheep – each has 250 floating points as its’ “genetic code” – they reproduce with other animations and there is family resemblence. If people like what they see they vote (up arrow key)– the animations that recieve the most votes “live” longest and reproduce.

In 2003 people created editors,and people could manipulate and see their creations, and this introduced an element of “intelligent design”  – people upload what they like and if others vote for them, they live.

Half audience votes for humans vs half for the genetic algorythm.

There is also mutation and crossover.

To deliver all sheep to all users – 20Terrabytes/day would be needed  – so they need to build in bit torrent in some way.

The main network is the lineage of each sheep – there are hundreds of thousands of them.

He did go from a linear model to logrythmic model for voting -because reproduction was following a power law – so he changed it to log, and that works better.

Each video gets rendered to 1000 clips, each clip is related to 5 others, so you get almost seemless transitions between the clips – true nonlinear video.