I just got done reading this post from Athol (and watching the fantastic video) and it got me thinking about a slight variation on the point he’s making. For decades the conventional wisdom about training (animal training, training children, military, self-training, whatever) has been that the best results come from consistently applied positive feedback. Somebody does something right, give them a reward (praise, a treat, a toy, whatever). The early research definitely pointed in this direction, and showed that positive reinforcement was more effective than negative reinforcement (the carrot is more effective than the stick), and the best results occur if you reward the good behavior every time.
Modern research shows that this is baloney.
Yes, positive feedback is more effective than negative feedback. But you get better results if you mix the two. And consistent application is actually not the best way. The very best results come when you reward the subject randomly (about 25% of the time good behaviors get rewarded), and apply the negative reinforcement 100% of the time.
Why does this work better? It makes more sense to think that rewarding good behavior 100% of the time would give better results because you’d get a clear connection between doing something good and getting a reward. And that’s true – you do develop a connection, which is why that kind of training regime works. But the other way works better.
When somebody is rewarded for something 100% of the time, typically they will converge to the minimum. That is, they will seek out the minimum possible behavior that generates the reward. This is analogous to the idea of profit maximization in the business world. People (and animals) seek the maximum benefit at the minimum cost.
If, on the other hand, we reward them only occasionally then each time they’re not rewarded the little hamster wheel (to borrow a phrase from Roissy) starts spinning. The subject starts wondering why they didn’t get rewarded this time. Maybe the good behavior wasn’t good enough. Maybe they need to try harder next time. “If only I’m a little bit better, maybe I’ll get the gold star and the pat on the head next time.” Intuitively, we’d think that this would work best if stronger rewards are linked to better behavior, but animal research shows that it actually works best if the rewards (both timing and intensity) are completely random. It also works best if you make sure that the negative reinforcement comes down every time. And every now and again you should mix in a major, major carrot – and every now and then a massive stick (not literally; I am not encouraging domestic violence here).
The application of this to Game and LTRs should be pretty obvious. If your wife is exhibiting poor behavior (nagging, ignoring the kids, whatever), find an appropriate negative reinforcement and apply it every time. But also find an appropriate positive reinforcement when she does it right, and apply that… about a quarter of the time. Every now and then, really shower her with the positive reinforcement (the beta move) even if she hasn’t done a particularly special job. And once in a blue moon, blow something way the hell out of proportion (the alpha move).
