“Connectors” and “hubs” may be genetically inclined
I still pay a lot of attention to social networking theory (not the stuff about the sites, but the research around how humans form networks of influence), ever since doing all the research that led up to my 2003 “Small Worlds” presentation. So this Reuters report that scientists have found a genetic component to having tight friend clusters was interesting to me.
To dig into what’s going on here a little bit: social networks are very discontinuous. They “clump.” We know from datamining that some people have many friends and some have few. The ones who have many are often referred to as hubs or connectors. These folks are also often the ones that “bridge clumps.” And when we say they have friends, we mean, like, they have a crazy amount more than ordinary people do. (The distribution of “number of friends” follows a power law, so the folks at the high end are very very very rich with friends, to a radically disproportionate level).
I suppose it isn’t surprising to think that there is likely some genetic component to this aspect of it. Most people are not like those guys.
However, within any given social circle, there’s also the question of how tightly knit it is. Does everyone in the clump know each other, or are they mostly held together by the person in the middle? You could think of this as a “clumpiness” metric (it is technically called the “clustering coefficient” and it was developed by Watts & Strogatz). And that is what these folks found to have a genetic component, which I find fascinating. 🙂
“We find that how interconnected your friends are depends on your genes. Some people have four friends who know each other and some people have four friends who don’t know each other. Whether Dick and Harry know each other depends on Tom’s genes,” Christakis said in a telephone interview.
…”We found there appears to be a genetic tendency to introduce your friends to each other,” Christakis said.
Relevance for anything useful? Well, singly-connected people tend to be at far greater risk of exiting a given social circle. And if you like, you can read that as “product or service” instead of social circle. If your highly connected friend is on Facebook, but you don’t know anyone else there, you are way more likely to stop using Facebook than if you are also connected to other people. So it would be to Facebook’s advantage if your friend were the sort of person who introduced you around; it makes it more likely that you’ll stick.
The flip side, of course, is that if that friend is the central node that matters the most in keeping those ties alive, she is more likely to take the entire group with her if she exits. This would be what we see in the “guild migration” phenomenon.
Alas (or hurrah), we don’t have genetic tests for users, so you can’t identify these users with a cotton swab just yet. (And I do say “yet.”) That said, you can identify these people through datamining; just search for clusters, then identify the most connected hub in a highly connected cluster and treat them preferentially to highly connected hubs in weakly connected clusters.
13 Responses to ““Connectors” and “hubs” may be genetically inclined”
Sorry, the comment form is closed at this time.
While accepting friend requests recently, I noticed that Facebook has added that very feature. I’m surprised you didn’t mention it. Or do you mean that the connector needs to make those introductions?
Tangent onto talking about myself.
I swear I must have a recessive trait for this. I’m not a connector, not whatsoever. (Though I might qualify for one of the others, but I haven’t read Tipping Point, so I won’t make that claim.) But I really enjoy connecting friends together. Very few things make my day better than pulling in two friends and seeing them get along well. I developed my taste for it in 7th grade, and I’ve continued to work at doing it ever since. The problem is that (1) I don’t have oodles of friends to play mix-n-match with and (2) I really suck at actually pulling it off.
I wonder whether or not they’d let me donate blood for a DNA check. 😛
Social networking, DNA and … guild migration phenomenon. I was not expecting the latter but it makes sense in the context.
Anyway, I am always wary of such founding. First, the results are expected. If you look at any software, there obviously is a dependency between its use and its development, isn’t it? Second, you are very likely being able to find explanatory genes. Can you remember me how many genes humans have?
I’m wary too. This gets into dangerous territory that can lead to racism. There’s a gene being called the “warrior gene”. It can be set in low activity or high activity and affects chemicals in the brain such as dopamine, and others. There’s been a correlation between the low activity set of this gene and ethnic groups that have a history of warlike activity. And somehow this has been connected, maybe just speculatively, with “stupid gene”. And then a doctor in New Zealand did some testing and found that a particular ethnic minority group has an unusually high rate of this “warrior gene”. See where this goes? Now, there does seem to be some correlation in this gene in low activity to aggressiveness in studies.
But is that all there is to it? Lets look at something else. Genes can turn on and off, both during development, and even on the spur of the moment.
This study shows that interest in another sex actually does this. It causes genes to temporarily turn off, and turning back on later. If this is caused by an outside factor, can this be controlled? Can training, for example, change a person’s biological response to whatever outside stimuli are doing to them?
Is a “warrior gene” really as important as intellectual exercises? Can social environment make a difference? How about a “stupid gene”? Does cranium size matter?
And back to the topic, which isn’t about genes other than that it reminded Raph of a particular topic of interest. These hub people seem to me to have something very friendly in their disposition. Something that removes the natural tendency of the rest of us to judge others, thus lowing our acceptance of them. They seem to remove negative feelings. Maybe it’s their energy and doing interesting things, in sort of a “busy hands” sort of way.
UO was classic. The worldliness of the game allowed for players to do so much. It’s the biggest reason for my love of that style of MMORPG. And “hub people” could really shine. The Mage Towers, Fishing Councils, Kazolas Tavern! Several player cities. Communities. This stuff is lost in the grind of current games. And there’s a terrible price to pay in the social loss for young gamers. The “numb brain gene” is turned on to max.
Or just figure out who throws the best parties. They’ll have the most friends or the best friends (not necessarily both).
I wonder what the graph of a server’s communications would look like and if any of the game’s hav ebuilt one. Take the say and tell channels. Everytime character A sends character B a tell or is in the vicinity of a say then draw a line between A and B. Increment the weight of the line for each say/tell that happens outside of some threshold to try and note the number of “conversation threads” between the two characters. The resulting graph of the entire character population could be transposed into accounts behind the scenes. I would bet that you could identify your hubs and connectors this way from the logs.
Of course you would need to filter out gold spammers and identify the half dozen most popular zones to eliminate says from weighting the conversations improperly but beyond that it just might show useful information. Add in some of the automatic relationships such as a line drawn between two characters everytime they grouped over the course of some threshold and those that are in a guild together and there’s more data to crunch. I suspect you’ll be able to quickly visualize 20 – 40 accounts on your server that act as hubs.
From there, you need to figure out what their roles are in the community and then implement things to make those roles easier/more functional in order to increase retention.
Amaranthar, I’m not sure that’s a fair reading of what was going on in UO. “Hub people” will really shine regardless of their environment or the activity they’re clustering around. They’re the sorts of people that’ll create vibrant guilds in diku-derivatives, and keep regular contact with people they meet in pick up groups. Worldy games don’t really inherently improve their ability to do what they do.
What worldy games do is force all of the non-hub people into interacting with other people as much as possible if they want to really play the game. Interaction becomes a requirement for success. But that sort of interaction isn’t the same as what connectors would be doing naturally.
And you’re also selling gamey games completely short as well. WoW has communities that are just as tight as any of the UO ones, and bigger besides; the biggest difference is that they aren’t as open to outside intrusion. You stumbled into UO communities; you have to seek out WoW ones. This has a side effect, and that is that it’s harder to cut across multiple clusters at the same time, but that’s a design flaw in how guilds are structured, not in how the game actually plays (Raid level play notwithstanding anyway. Raids create other issues).
To my direct knowledge, the following games have datamined this sort of information:
Everquest 2
Uru
Shadowbane
The data for EQ2 is part of what Dmitri Williams is writing his material on; the data from the Ubisoft games was presented at AGC several years back by Mike Steele and Patricia Pizer.
@Raph,
If that’s the list, then it is awfully short. I know those are the only one’s you know of but you know of a tun of projects so that’s short given the list of projects you’re likely aware of. Then again, I come from the web world and we do this all the time so it seems much more natural to me to mine for information like that. I’ve always thought that this topic is one that could be used to directly influence retention rates assuming you have the right data and the right model built around that data to draw conclusion from.
Derek, doing social graph analysis in games is still, I think, something of an untapped vein. When I did that talk in 2003, it was sort of a revelation for a lot of folks (it wasn’t entirely current in the web world either, back then).
Eolirin, you might be right, but I don’t know how you can tell. But then, I was never deeply involved in any of the current models. I don’t know how you can tell when guilds are based on the player’s need to group to run quests and end game content. Is there really any glue there, socially speaking?
In UO you had things like Kazola’s where hundreds of player participated not because of any need, but purely out of a social desire. The Mage Towers had, probably, thousands between the shards. And again, there was no real need. It was a desire to be a part of it, and this seemed to me to stem from a desire to join with key leaders in that effort. Even then, the Mage Towers had conflicts within due to PKers, which ultimately led to their falling apart, but after quite a bit of time. The Fishing Council of Britain had events that lasted years. Once again, no need to participate, but lots of players did for the social aspects of fun.
Still, maybe you are right and the gamey games did have hub players who people drew people together by their own nature, I might simply not be aware of it.
Worldly games forcing players into interaction? I don’t see it on a big scale. You can play with 4 friends that don’t know each other and do just as well, as far as playing the game. Maybe UO did due to PKers, but that doesn’t mean all worldly games have to have PKing on that scale. It doesn’t mean a worldly game has to have PKing at all, really. If either game style forces interaction, it’s the level grinds for group quests. And worse yet, they force it by levels, dividing people according to their current level which may be different that the players they grouped with a month ago.
And size doesn’t matter! Not when comparing a game that had coding problems with large sized guilds and a much smaller base to boot.
In regards to the gamey/non-gamey discussion, I think we may be talking about apples and oranges. I can think of at least two types of “hubs” in a gaming context; I’ll call them Generals and Coordinators.
Generals are action-oriented “hubs”. Their social interactions revolve around assembling and leading effective teams in PVE or PVP raids. They may know the combat effectiveness of every member of their social network, without having the vaguest clue about RP backstory or real life interests. They tend to hit maximum level as quickly as possible, and they tend to have little interest in people who aren’t at least close to the level cap. Obviously, these thrive best in an environment with a wide variety of group-oriented challenges that test their mettle.
Coordinators tend to ignore or at least backburner advancement and adventuring altogether, focusing instead on organizing communities and events. They tend not to have the equipment or training to be effective combatants, but they have the personal charm and charisma to convinced grizzled mercenaries to sit still for weekly poetry readings. Coordinators will remember your RP backstory and write new story arcs to weave it into the fabric of the community. Coordinators will ask after the health of your ailing cat — by name. Coordinators tend to flourish in environments where the level grind, mass raids, and PvP are deemphasized, at least to the point where there’s enough breathing space to organize an event and have some decent turnout.
In my experience, the individual who can successfully pull off both roles is exceedingly rare. In UO, both types of individuals coexisted, and the biggest and best guilds usually had both Generals and Coordinators. But when MMOs and other virtual worlds really started to take off, Generals gravitated towards PvP titles and big raid games, Coordinators towards worlds with support for events and communities.
So while I would agree that there are hubs in all MMOs, I’d argue that the game system has a strong impact on what flavor of hubs are most attracted, successful, and likely to stick around. If you’re trying to draw in both, be prepared to be hammered from both sides for having too much x and not enough y.
As long as no relationship types are automatically imputed, you can mine connector data. IOW, you can’t impute they are friends or mates. Two people who hate each other may talk a lot too. My intuition is the frequency/amplitude curves will look different when viewed at a different scale, but unless you are mining the text itself, you can’t assume anything topical, just that a relationship of some kind exists. Then a time domain analysis can tell you if it is casual or strongly coupled.
If I were mining Facebook, I’d want to know who has been rejected for a “Friends-of” request, but I’d also want to know how many and who are sitting on the list of invitees to which no response was given. IOW, the person to whom the invite is sent is sitting on that request and for how long. The same for ‘friend recommendations’.
As far as Burger King goes, I would have removed my ten best friends and sent them a backchannel message to tell them to do the same. Then we could have gotten as many others to do that as we could and then show up at BK with a Whopper of a crowd looking for a free meal. In the world of guerilla advertising, turn about is fair play.
[…] partially be genetically determined. I got the news from Raph Koster’s blog entry titled “Connectors” and “hubs” may be genetically inclined. Fascinating […]