A distinguishing feature of human networks

OK, so the real distinguishing feature is that humans are involved. Duh. But I had a particular feature in mind ...

In the example of the spread of contract bridge, I guessed that a critical event, then as now, would be a well-known authority publicly endorsing a new idea. Well-known authorities are by definition connected to a large portion of the general population, because large numbers of people read/watch/listen to their pronouncements. This is a classic "small world" feature, which keeps the overall diameter of the graph (the maximum number of "degrees of separation" between any two members) small.

In a network like the internet, there are also a relative few "hubs" that are connected to a large number of more peripheral nodes. These tend to have monstrous bandwidth available and will carry huge amounts of traffic, both in and out. When I publish this post, for example, it will almost certainly pass through one or more of a relatively small number of "backbone" servers along with a mass of completely unrelated information.

In the human case, the well-known authority produces human-sized traffic, just like everyone else. It's just that this information gets broadcast, verbatim, to a large number of people. Similarly, anyone can try to send a message to the authority, but only a human-sized portion of it will actually get through. In practice, the authority's input will be heavily biased toward a small number of trusted people (who thus may be influential without being well-known), with maybe occasional input from random people.

I'm not sure what all the consequences of this may be, though I've been grasping at them a recent post or two. One way to look at it is that in a human network everyone's CPU and network are more or less the same size, while in a computer network they can vary by orders of magnitude. This in turn affects scalability.

Nor is it at all clear (to me, working off the top of my head here) that throwing more people at the problem would work, even if they'd sign up for it. Even if I decided to have ten or a hundred or a thousand people act together as a virtual hub, there's a limit to how fast they can talk to each other.

Hmm ... to what extent does something like Wikipedia act as a virtual hub?

The small world of contract bridge

In the spread of a new idea, which is more important: the speed at which information flows or the speed at which we absorb it?

I have no doubt that modern communication, including the net, has increased the speed of information, but let's not underestimate our forebears. From a recent New Yorker book review on contract bridge:

The modern version, contract bridge, was created in 1925 by the railroad heir and master yachtsman Harold Stirling Vanderbilt, who had been annoyed by what he felt were deficiencies in the previous version, auction bridge. Vanderbilt was a passenger on a ship that was travelling from Los Angeles to Havana by way of the Panama Canal, and on the evening of October 31st, while playing with three friends, he introduced several improvements that he’d been mulling over, including a method of scoring that required players to more accurately assess, during the bidding, the number of tricks they would take, a prediction known as a contract. Vanderbilt shared his ideas with a few other friends in Newport and New York, and his game spread across the country and around the world at almost unbelievable speed. “Half a year after Vanderbilt’s voyage,” McPherson writes, “a notice appeared in the Los Angeles Times announcing that a Chicago woman was suing her husband for divorce on the inexcusable grounds that he trumped her ace.”

I doubt that communication speed was the limiting factor here. More likely, it was the time required for a player to try out the new rules, mull them over, deem them good and decide to try to introduce them in their next game. By the small world principle, it doesn't take too many such hops to reach the entire community.

From a graph theory point of view, you have a set of players connected by lines of communication (we play with the so-and-so's at our Tuesday game; I correspond with such-and-such; we all read thus-and-such's column in the Times). Propagation through the graph is a matter of incubation time (playing, mulling it over) and transmission time (the epidemiological terms are deliberate here; I'm sure there is quite a bit of relevant research in that field).

In the present case, there is probably a classic small world graph, with a few well-connected nodes (major celebrities like Vanderbilt and the major columnists) and lots of less-connected nodes (the Tuesday night games of the world). It would be interesting to see when the new rules first appeared in print. It would also be interesting to see how long it took contract bridge to supplant auction. That's clearly not a matter of propagation speed.

These days communication time is much more limited by the reader than by the medium. If a headline appears in a major news source, it may still take some amount of time before everyone gets around to reading it. Blogs and online news sources shorten the time from writing to availability to practically nothing, but they don't necessarily make me read it that much sooner.

Which has had more impact overall: "Old-school" electronic communication (think telegraph, radio, TV) or the internet? One could make a decent case for the old school.

Limits on human bandwidth

Along the lines of the "Rules of Thumb" posts:

I won't claim that the internet has changed nothing. It's at least changed how far and how fast news travels, and this has a number of subtle and unsubtle effects. But no matter how fast the network, storage and processors, as long as people are using the web there will be certain hard limits. Some that come to mind:

How much information can a person absorb?

If we're talking about raw sensory input, which appears to be dominated by sight and (to a lesser extent) sound, then my guess is that HD video comes pretty close to the limit. That's on the order of 20Mbit/s, or 10GB/hour, 250GB/day or 100TB/year. I'm taking the MPEG compressed rate as opposed to the raw frame rate as that more closely represents what the visual system is really processing (because successful lossy video compression is finely tuned to the way the visual system works)

Given that disk capacity increases about 100-fold every decade, in ten years one could reasonably afford to buy enough storage to store a fairly immersive audio/video stream that would take a year, 24/7, to watch. Taking time out for things like, um, sleeping and eating, it would probably be more like two or three years.

Conversely, if you wanted to record everything you saw and heard, you could do it for a reasonable -- and decreasing -- annual cost in the not-too-distant future. Anyone could do it, unobtrusively. "Be careful, his bowtie is really a camera".

If you want to boil that raw content to the more abstract images stored in the rest of the brain, there's a pretty well-established medium that covers that reasonably well, though not perfectly: words and pictures. It's trivial now to store all the words a person could reasonably process or produce in a lifetime, or even every mouse click or keypress, timestamped to, say, the nearest millisecond.

A picture on disk is probably worth more like hundreds of thousands of words, but storing tens of thousands of pictures is no big deal these days, either. That's a lot of pictures, if you want to take the time to look a them.

In short, when it comes to words and pictures, the limitation is not what the computer can handle but what the people using it can handle. Audio and video are rapidly approaching the same state.

How many people can a person keep in touch with?

With modern technology, I can now keep in touch with people all over the world, but I can't keep in touch with any more people than I ever could. Somehow the advent of the internet didn't add any new hours to the day. I don't have objective numbers handy on how many people people interact with, though I'm sure there are studies on the subject. At a guess, I'd expect the usual log-normal distribution, with a handful of people accounting for most of a typical person's interactions and maybe a few dozen accounting for almost all of it.

Balancing that is the small world property of social networks and many other structures, including the web itself. In such cases the degree of separation between any two individuals grows very slowly, if at all, as the network expands. In the movie version, there are no more than six degrees of separation between any two people. The actual number (neglecting any groups that really are totally isolated) is probably larger but not much larger. [See these later posts for a bit more on the topic]

How quickly can a group reach consensus?

Whether it's everyone deciding that magenta is the new chartreuse or a deliberative body deciding that the bylaws should be amended to allow for amendments to amendments, the game has probably not changed appreciably in recorded history.

In the mass-consensus case of global pop culture, the scope is bigger, but one of the "small world network" results is that both the average person's view of the social network and the overall structure of the network itself change little as the network grows. In other words, fashions in the malls of the world are driven by the same basic forces as those in, say, Louis XV's France or Julius Caesar's Rome, just on a bigger scale.

In the small-scale case of a deliberative group, the limiting factor is how quickly the members can get the others to understand and (ideally) accept their view of the world. Again, it seems to matter little whether the members are sitting around a campfire or exchanging messages electronically.

I do find it interesting that most of the distributed groups I've been involved with develop a mix of email, live conferencing and face-to-face meetings. Most of the routine stuff can be worked out via email, sometimes you have to pick up the phone and talk to a particular person, and every so often you should all meet. Most of those meetings can be by phone/IRC, but a few of them should have everyone sitting in the same room. It'll be a while yet before technology can completely replace this.