09.43, Thursday 3 Jul 2008

Two kinds of training: respondent conditioning is when you perform two events simultaneously so the subject confuses cause and effect. So think of Pavlov and his dogs: the dogs salivate when he gives them food, and then he rings a bell whenever he gives them food and the dogs get used to that. Or rather, they get conditioned to that. Then they confuse cause and effect and end up salivating whenever the bell rings, whether the food comes or not.

(Pavlov cut the dogs' throats to find this out. His theory of conditional reflexes dominated institutional Soviet thinking for decades, in part leading to both the Soviet rejection of cybernetics and their late development of computers, and also to Lysenko rejecting Mendelian genetics. Lysenko directed farm policy under Stalin, and his misguided theory of agrobiology led to mass crop failure and starvation.)

Another kind is operant conditioning. It relies on consequences after the event to produce conditioning, and it's more suitable for voluntary behaviour. It's what they use on dolphins.

There are a few types: you can give a reward for good behaviour; you can remove pain for good behaviour; you can actively punish bad behaviour; you can remove a pleasant stimulus when bad behaviour occurs.

If you give me a biscuit every time I make you tea, I'll likely make you more tea.

The most powerful form is variable-interval reinforcement. That's when the reward doesn't happen every time, and you end up working harder to get it. It's as if you're trying to figure out the pattern, to get the reward to come more often. It's why email is addictive: you hit that 'get mail' button and get your reward, but not always, just sometimes, and that conditions you into checking more and more.

One weird thing that happens, in operant conditioning, is the extinction burst. There's a nice example I read, I don't recall where, about elevators. Imagine you live on the 10th floor and you take the elevator up there. One day it stops working, but for a couple of weeks you enter the elevator, hit the button, wait a minute, and only then take the stairs. After a while, you'll stop bothering to check whether the elevator's working again--you'll go straight for the stairs. That's called extinction.

Here's the thing. Just before you give up entirely, you'll go through an extinction burst. You'll walk into the elevator and mash all the buttons, hold them down, press them harder or repeatedly, just anything to see whether it works. If it doesn't work, hey, you're not going to try the elevator again.

But if it does work! If it does work then bang, you're conditioned for life. That behaviour is burnt in.

Or a baby, crying to get attention, will have one last huge attempt to get attention before learning that tactic isn't going to work.

I have a friend - again I can't remember who - who saw a talk from a fellow who trained dolphins - and I don't remember why or where - and he mentioned this extinction burst. You trail off the fish rewards for leaping through the hoop, and let extinction occur, and then when the extinction burst happens - you know, the dolphin is trying everything it knows, going crazy trying to get you to notice it and feed it fish - bang, that's when you get in with the big reward and there you go, the dolphin's hooked.

Anyhow.

It strikes me that dating, when successful, may produce operant conditioning.

It also strikes me that some people may have personalities that naturally produce operant conditioning to certain behaviours in the people around them, simply by acting with exactly the right balance of predictable/erratic or aloof/intimate.

Back to dating. It would naturally be most successful if a couple condition one another reciprocally. And it makes me wonder: could this be routinized? Or rather, could this be a pattern followed deliberately? And if so, could that be a product, for sale?

I don't believe that knowing the conditioning was occurring would interfere - my muscles still develop at the gym even though I'm working them artificially - but it would have to be done carefully.

How could you produce this artificially? I'm not sure how. Maybe a pattern of dates where one partner or the other is instructed not to show, almost at random? A system which means all communication is mediated through something which is unreliable, so it occasionally drops calls--and then that system is manipulated in order to produce the extinction, the extinction burst, and eventual pay-off?

That is: a couple dating should have available manufactured, reciprocal, variable-interval operant conditioning, with a pay-off timed to the artificially produced extinction burst, to trigger mutual addition, and they should be able to buy this in a shop.

It's an interesting design challenge. Here are my criteria: it has to be adopted knowingly by both parties (so no Rules of Seduction games); it has to be reciprocal and involve as little technology as possible; it has to be productisable--that is, it can't be the side-effect of another system: it has to be able to be actually or virtually packaged up and sold. And the usual product rules also apply: how are people going to understand and discover it; does it fit with natural flows (like, if the communication is mediated, won't they just swap phone numbers and use those instead, because it's easier); do all the halting states have ways out; how does use of this product act to expand the market for this product. Other than that, it's all open.

I am aware that talking like this makes me sound like a sociopath.

Follow-up posts: