Tuesday, September 06, 2005

It's not about race...

... it’s about class, culture, and civilization, and it's about feedback… or rather the fundamental lack thereof.

I haven't said much about Katrina, and I'm not going to say much because ... well basically everyone else has.

But this one subject we need to talk about.

Feedback: In systems/logical analyisis terms feedback is the measurement of the results of a particular action or decision set. This is distinct from other inputs, because it involvs empirical evaluation of results of prior actions.

In a regulated system, there are several different mechanisms for acting on feedback, the most basic categories of which are referred to as "bang-bang", and "Proportional response".

In a bang-bang system, you are always either full on, or full off, with no throttling. You attempt to reach an acceptable average behavior in the system by setting reasonable tolerances, and than averaging out the time spent below and above the median.

In a proportional system, more complex measurements are required, because the aim of the system is to present a response proportional to the deviation from the median, and thus damp out the oscilations inherent in a bang-bang system.

In realtively simple systems like home heating, bang bang is considered fine enough control; but the more complex a system gets, the more complex it's controls must get to achieve the desired results.

Now there are two types of feedback. Negative feedback tells you that you have gone outside your boundary conditions (turn around, slow down, stop), and positive feedback tells you you need to move more towards them (speed up).

Negative feedback is the only type of feedback in bang-bang systems, and it is the easist feedback to implement, as well as the most effective in producing large results. Too much negative feedback however tends to result in wild oscilations of the system, and it's eventual failure.

Positive feedback is necessary for proportional systems to function, however too much positive feedback will also result in out of control systems as they quickly exceed their bounday conditions and fail.

There is actually a third type of feedback useful in more advanced systems known as steady state feedback (it's actually a variant of positive feedback), but that's out of the scope of this discussion.

Now, feedback has the same meaning, in a very different, but related field of systems analysis called Game Theory.

Game theory is the study of optimal decision sets. The object of game theory is to analyze what decisions will produce optimal results for a given situation. Often this is logically represented as a game, thus "game theory"... but game theory applies to MANY situations; and absolutely critical to game theory is the concept of feedback.

Let me illustrate. The classic game theory construct is called "The Prisoners Dilemma", illustrated like this (I'll jsut cut and paste, because it's a standard example used by everyone):

The classical prisoner's dilemma (PD) is as follows:
Two suspects A, B are arrested by the police. The police have insufficient evidence for a conviction, and having separated both prisoners, visit each of them and offer the same deal: if one testifies for the prosecution (turns King's Evidence) against the other and the other remains silent, the silent accomplice receives the full 10-year sentence and the betrayer goes free. If both stay silent, the police can only give both prisoners 6 months for a minor charge. If both betray each other, they receive a 2-year sentence each.

It can be summarised thus:


Prisoner A Stays Silent Prisoner A Betrays
Prisoner B Stays Silent Both serve six months Prisoner B serves ten years; Prisoner A goes free
Prisoner B Betrays Prisoner A serves ten years; Prisoner B goes free Both serve two years

Now, in the prisoners dilemma, there are a few sets of assumptions you need to make to play the game. The first, is that every player is trying to minimize their negative result (imprisonment), and maximize their positive result (freedom). The second, is that every player should understand that all the other players are attempting to reach the same result set.

In the first eample, let's assume you are each playing in isolation, with no way of knowing what the other players choice is before the end of the game.

Now, in a single instance of this game, the optimal choice may LOOK like ratting out the other player... but that looks like the optimal result to the other player as well so he would likely choose that and you BOTH serve two years; the second most negative result for each player.

The optimal decision set here is that both players stay silent, and thus both end up doing six months, which is the second most positive result for each player.

The problem is, you can't necessarily count on the other player to keep his mouth shut. He may jsut go for main chance, and hope he goes free (and you get screwed).

If you think he's going to do that, your optimal decision IS to talk.

So how do you make that decision?

Feedback.

Let us say that you are both in the same room and can hear each others decisions.. and look in each others faces when the options are presented to you. You're getting feedback from the other player, even if they dont say a word, and then after they DO you get more feedback, thus can potentially alter your choice; which will tend to produce closer to optimal results.

That's a bang-bang system right there BTW. No proportionality of response, either shut up or talk; and hope the result is closest to optimal.

Now lets say you run this game ten times with the same player. both in isolation again, and the one who spends the least amount of time in prison wins.

How do you decide how to play?

You could always be "the good guy", and not talk; in which case every time your opponent talked (and he would talk every time once he figures out you're "the good guy") he would go free and you'd do ten years.

You could be "the bad guy", in which case you just talked every time, and pretty quickly your opponent would figure out what you were doing, and you'd both spend 2 years

Finally, you can play "Tit for tat". In tit for tat, you start by making the decision set most likely to be optimal for all players, then after each iteration of the game you do to your opponent exactly what he did to you the last time. You give him "tit for tat".

Eventually, your opponent should figure out that you are going to hit him back with whatever he hits you with, and therefore he will over time choose the decision set that produces the optimal result for both players.

This is the concept of feedback memory, or stateful memory. You retain information about (at least) the current state of the system, the last state of the system, and the starting state of the system; and the decision set that made you reach this point.

The more times you play, the more collective feedback you recieve about a player; until in theory, you should always have the same result every time you play. You both end up doing six months each time; and it becomes clear that "Tit for tat" is the optimal decision strategy.

But critical to this assumption, is that you are recieving good feedback in response to your actions. If you don't know what the consequences of your choices and actions were, you cannot judge how best to make your next choice.

And that is the problem we face in society today. Certain groups within our society are attempting to make it entirely free of negative feedback. They wish to create a consequence free society with regards to our personal interactions.

But without negative feedback, a system will either remain the same, or accellerate out of control.

Worse, they also want to create a MEMORY free society, or in game terms they want us to create a steless society (would that were true in political terms).

Essentially they want us all to play the prisoners dilemma with out lives, but they don't want us to be able to play "tit for tat".

The chaos in New Orleans is a vivid and appalling example of what happens when social feedback is distorted by the actions of the state.

Okay so how can I possibly be reducing these people lives (and deaths) to game theory?

Quite simple really: The underclass of New Orleans (which comprises more than half the city) has had much of the feedback removed from their decision set evaluation, because of governement "assistance". Their poor choices (for whatever reason they made them, and I'm not talking about reasons I'm talking about consequences) did not have the proper proportionality of negative result, and therefore the poor choices were made more often. As these poor choices require far less effort than do good choices, out of proportion with the positive results produced by good choices; objectively poor choices were moved closer to the optimal decision set.

Without a proportionality of negative feedback, the second most negative choice can easily appear to be the second most positive choice.

Okay translating into the real world, what does this mean?

Andy and Barbara are Boyfriend and girlfriend. So are Charlie and Donna.

Charlie and Donna get married, and have a child together. Charlie has to work two jobs to support his wife and child, who aren't living very well. Charlie is tired all the time, and Donna doesn't see him very often, and neither does the baby. When Charlie isn't working, he's taking care of the baby and Donna is off working as well. Because of all their hard work, they are all well fed, the baby is being properly taken care of, and they are building some savings.

Andy and Barabara also have a baby, but they don't get married. Barbara goes on welfare, and gets a publicly subsidised apartment. She doesnt work because she would lose her welfare and subsidised apartment. Andy works, but not very hard, because if he works too much then the government takes it away from him to pay back the welfare Barbara is getting. Barabara IS able to stay home with the child, and they are all well fed because of the food subsidies she recieves. They have no savings, and no advancement from this position; and eventually the welfare will run out, but if they put themselves in a better position for when that happens, they will lose the welfare anyway. Plus, they'd both be tired all the time, and they wouldnt get to have any fun.

In a natural system (without government distortion), decision set A-B (andy Barbara) would result in Barbara and the baby starving to death, and Andy going and impregnating some other woman to start the cycle over again.

This would result in the eventual destruction of our society, as no one still alive in it would be willing to breed.

Actually, the more likely result is that all three individuals in the A-B decision set would become criminals.

In the government distorted system, the worst possible decision set, is spared the worst possible result, instead given a marginally acceptable result. ou keep doing that, and there will be a percentage of folks who find that suboptimal result acceptable, and will live in that manner.

But what happens when the government money shuts off?

New Orleans happens