Sunday, March 11, 2007

Sunday Rumination: Compartmentality

It's been a cold month or so. We forget that the early part of the winter was like Florida. Remember? Just February, it got cold. And today, it's warm, nice, sunny. The last snow-ploughed heaps of snow were washed away by yesterday's rain. This morning, this is the good time. Sun shining, snow's gone, WPFW on the radio: a classical guitarist, the dog has been out, coffee is cooking, kids are asleep -- what can you say? This is good.

I was trying to explain to my daughter last night, but it really is kind of hard. Can you imagine? Me, skinny, with a beard, standing up in a honky-tonk along I-5 somewhere, or in North Carolina, or Alaska, Tacoma, Tucson, Colorado, Nebraska, singing Willie and Waylon and Billy Joe and Hank (senior more than junior) and struttin' my stuff and scaring those other guitar players ... Can you imagine me reading poetry at Linnea's, leading poetry workshops in the schools ... How about me, in a suit, giving a keynote address in China, or Copenhagen, or Spain? Or speaking to an audience of government survey methodologists about changes in technology and how they affect the survey process?

To you I'm known for the "sex ed thing." But more people know me for something that gets called "swarm intelligence." (Google "Kennedy" and "swarm" to see what that's about.) Well, because that's the title of my book, I guess. Want to hear about that? That's your cue to choose whether to keep reading or hit the back-arrow. I got a full pot of coffee, I can go on for a long, long time.

I'm a social psychologist, really, and I am impressed that people, starting in ignorance, can figure out what's going on. Talking to each other, people can determine what are the important aspects of a situation, they can figure out which of those aspects deserve their attention -- the fact is, people working together can accomplish amazing things. Science is the big one, the way we've made sense of the world, but other things are incredible, too: movies, art, literature, architecture -- music -- government, culture, Cheetohs. It seems we are a Big Step beyond the next monkey down.

Sometime ten years or so ago I put some of these thoughts into a computer program. I showed that if you wrote a program that had a population of individuals start from random guesses and just talk to each other, learning from each other and teaching each other at the same time, all blind to any details of the problem solution, all equal and all absolutely unintelligent on their own, you could solve the hardest problems known to mathematics and engineering.

We call it a swarm, because if you look at two dimensions of it on a computer monitor the guesses start out dispersed and end up clustered in one part of the screen, usually the part where the answer is. On a dynamical Cartesian graph, they look like a swarm of mosquitoes or gnats. The little guys start out all over the place, and after a few iterations of the program they have converged in groups around the optimal solutions. When a problem has two or more solutions, they will cluster around the various ones, but usually the best solutions will attract individuals until the lesser solutions are left abandoned.

At the heart of it is the idea of cooperation, or collaboration. Yes, every one wants to be the best, but the thing is, the population members -- we call them "particles" because recursively updated vectors can be conceptualized as moving points -- are completely unintelligent by themselves. All their power comes from the ability to learn from each other, and to share their insights.

There is a kind of theory of the evolution of ideas called memetics. Memes are like genes, they are like an idea that passes through a society, from one person to another, mutating slightly as each person interprets it a little differently, and as each person blends the idea with the ideas he already held. Richard Dawkins, who invented the term, is an evolutionary theorist, rather controversial right now because of his criticisms of religion, but whatever, it has been obvious to anthropologists and other observers of the human lifestyle that culture evolves, ideas evolve -- Dawkins did us a service by showing that the well-developed theory of Darwinian adaptation was useful for describing this other kind of evolution.

My swarm theory turns memetics on its side. You can't really talk about the evolution of ideas, because ideas don't exist by themselves. You have to have a mind to have an idea. And minds change over time.

The evolution of ideas is really -- c'mon, you know this is right -- the evolution of a pattern of changes in people who have ideas. Ideas wouldn't evolve if there weren't people thinking about them and talking about them. So swarm intelligence models people, abstracted as simplified computational entities, and has them interact in a simple way to solve hard problems.

There's a lot to consider there, and I have spent the past ten-years-plus writing computer programs and publishing books, chapters, and papers trying to understand what goes into it. For instance, the communication structure -- the social network. It matters, it really matters, how you connect the population, how ideas spread through the group. It isn't good to have the communication network too dense, or eveyone gravitates toward the first good idea anybody thinks of, and it's not good to have it too sparse, or new ideas never propagate out to the group. "Overflying" -- it is important that when somebody influences you, you go beyond that, you look at what they suggested and more. Who influences you? It is a big question, whether you are influenced by the most knowledgeable person you know, or by some kind of average of all the people you know. There is a big difference there, not that one's better than the other necessarily, but it affects the kind of communication structure that best supports problem solving.

It turns out that cooperation is not just a pleasant way to act, it's actually a very constructive thing when it comes to finding solutions to hard problems. If people can talk, if they can influence one another, if everyone is agreeable to make adjustments when they learn something new, the group can make progress toward a goal, even if it's a difficult goal. Sometimes you move away from the target, for instance the group may be attracted to a local optimum, but eventually, if the system parameters are good, somebody will stumble across the best solution or region of the problem space, and then the rest will follow.

I normally live my life as if this, the swarm business, and the sex-ed thing were separate. Never mind music, poetry, philosophy, my job, my family. I compartmentalize my life, just like you probably do. And I definitely don't want the CRC to start having opinions about whether to use the FIPS or canonical interaction strategies, and whether Shi and Eberhart's inertia weight is more "Christian" than Clerc's constriction coefficient, which affects the whole right-hand side of the velocity formula. But sometimes you realize, it all works together. Those ideas apply here, and the challenges that we face in trying to overcome bigotry and nuttiness here apply there.

The idea of a local optimum is a good one. Sometimes a problem has more than one solution. Sometimes, but not very often, all of the solutions are equally good. Sometimes there are a number of pretty good solutions, but one that is better than the rest. You want to find that one. In our situation, you can see that it's a pretty good personal solution to write off gay people altogether, call them sick and reject them, and then you don't have to deal with your own feelings, and you don't have to think about it. Since most people are straight, that solution pretty much works, most of the time; you just have to surround yourself with people who are adopting the same solution as you. Of course, gay people have forced the issue. They have made a decision to come out in public and say "I do exist." So now the decision to ignore them is challenged, it's not such a good solution any more, because they won't passively agree to being ignored and rejected.

So what do you do? One subpopulation of straight people says, OK, they're right, I don't really feel like they do, I don't get the gay thing, but so what? They seem like decent people, I don't know what they see in that, but whatever, it doesn't hurt me any. And then you're done. Another subpopulation says, that's wrong, it's evil, it's sinful, they must be rejected, they must be ignored, they must be stopped. In my other world, these are just local optima, they are two solutions to a cognitive problem. One, I'd have to say, is better than the other, because it's simpler to accept things you can't change and people create less of a disturbance when they are not being discriminated against, and really, a bunch of gay people aren't going to hurt anything. I mean, really, they are not going to destroy your marriage, unless you yourself are ... never mind.

I should point out that, in this cooperation thing, competition is really hiding in there. It matters that you are the best, in my programs. I wrote a program once where a person could be part of the swarm, and I tried it, and I found out that I really really wanted to beat those other guys, even though they weren't real. There's an inequality in the formula, a less-than condition, and you want that condition to evaluate as true, you want to find a better solution than you or anybody else has found so far.

I am ridiculously optimistic. I don't know why, but I keep thinking that people will do the right thing. It might be slow, but I figure people want to live sensible lives, people don't want to hurt somebody for no reason, people want to be able to trust each other. This sexual orientation thing is mainly something a lot of people haven't thought much about. Now that the CRC has forced the issue, you just about have to choose whether you want to be on the denial side or the magnanimous side; do you really think it's just the way some people are, or do you believe it's a plot -- an "agenda" -- to take over the world? People have to think about it, and given that none of us ever really learned anything about sexual orientation, we have to rely on our playground knowledge, and it will take a while to turn that corner. But for some crazy reason, I think people can do it.

Anyway, I will continue to compartmentalize my lives. This TTF part of me is definitely an important part, and a part I didn't know I had. I had never been anything like an "activist" before, and really, nobody who knows me, in all these years, has ever called me a "liberal." And here I am, a liberal activist, what a trip. It does seem that I'm just a little bit Irish, a little bit of a bulldog, I inherited a stubborn streak from my old man, and I think you can tell, I'm not going to let go of the CRC's throat till the blood's drained out of it. But, as I sit here sipping coffee on this beautiful spring -- is this it? Is this spring, or is there more snow coming? -- morning, I am thinking, even though my life is carved up into pieces, the pieces really do fit together, in a weird way.


Blogger grantdale said...

Fascinated to read of your work on "swarm intelligence". Well, I am!

Have also done some work on "solutions" that "groups" come up with (in a workplace environment).

Everything does work well if the group is committed to honestly working through the problems and the solutions. (And if they basically know what they are talking about, or are agreeable to learning if they do not, when technical or complex issues are discussed).

The horrid derailment occurs when the group attempts to maintain group cohesion (above all else) by compromising with -- frankly -- a lunatic, or someone with a deliberate agenda of causing trouble etc. If you compromise to remain on friendly terms with such people... rather than find the optimal solution they most often end up incorporating very wrong thinking and therefore coming to very bad outcomes.

Oh. CRC. Guess you know all about that problem already... and the solutions to that problem!

March 13, 2007 10:01 PM  

Post a Comment

<< Home