I am unsurprised to see that there is a fresh paper out that states that there are strong mathematical commonalities between gang formation and guild formation. (I am also unsurprised to see Nic Ducheneaut and Nick Yee among the authors of the paper).

In particular, we find that the evolution of gang-like groups in the real and virtual world can be explained using the same team-based group-formation mechanism. In contrast to the quantitative success of our team-based model, we find that a homophilic version of the model fails. Our findings thus provide evidence that online guilds and offline gangs are both driven by team-formation considerations rather than like-seeking-like. Interestingly, each server’s Internet protocol (IP) address seems to play an equivalent role to a gang ethnicity.

– “Human group formation in online guilds and offline gangs driven by a common team dynamic”

The source data comes from WoW guilds and from Long Beach CA street gangs — and with lots and lots of data points to boot.

What’s the mathematical model underlying the two? Well, it’s a basically pragmatic approach to team-building:

• teams try to recruit people with varied skills or attributes
• people join based on judging what they can add to the team
• the people joining a team assess the team as a whole, rather than looking at every member
• the team accepts someone based on what they think the person can contribute
• people leave when there’s lots of members who offer the same attributes they do
• people always look for teams where they can contribute more
• when membership n a team is stable, the team grows by merging with other teams

This kind of sounds like how companies work too. In fact, the paper suggests at the end that maybe this is just how human teams work — street gangs and online game players have fairly little in common, demographically speaking.

To me the thing that is interesting about this is that in all three cases, of course, we’re talking about a group formed to accomplish ongoing operational objectives. The above recipe, barring sentimental attachments to teammates, sounds like exactly the most rational way to attempt to construct a team to win at things. I’d be curious to hear about formation models in cases where the objective is different.

One neat tidbit: the relative sizes of guilds in WoW seems to be on a basically invariant curve — similar to how the relative sizes of cities in the US across a century of census data seems to magically keep to the same curve decade after decade, even though the absolute size, and indeed position of a given city on the list, changes over time. However, unlike cities, this doesn’t seem to be on a power-law curve.

via  Boing Boing.