How do you reward/reinforce desired behaviors in cases when the desired behavior is normal rather than exceptional? If I give my child a reward every time they’re not hitting their sibling, isn’t this isomorphic to taking away their expected reward every time they do hit their sibling (a punishment)? Worse, if I only started this “reward basic lack of misbehavior” scheme when I noticed that a child was prone to misbehavior, isn’t that just going to be perceived as “[I/my sibling] got a new nearly-continuous reward for misbehaving enough to trigger parental notice”?
Then, since it’s significantly harder to apply reinforcements than punishments to badly-behaved children, wouldn’t we expect to see a strong correlation between reinforcement and good results (or between punishment and bad results) regardless of how effective each was at changing behavior?
You don’t reward the omission of bad behavior, you reward good behavior. For example you reward your child every time they are good to a sibling. Sure, it is harder to look for good behaviors and to reward that instead of simply punishing the child when they do something bad but punishment doesn’t work well as mentioned earlier and for example often leads to situations where the child decreases the undesired behavior but only when it is being watched.
Also whenever you can you should choose intrinsic or ever extrinsic reinforcements.
Talk about building a system that can be gamed! My big dog does not steal my little dogs food except when he can. And in between they play together pretty amusingly. I don’t think I would gain anything by rewarding them playing together in terms of a lower rate at which the theft of food occurs. The problem being, my big dog (and probably my little dog) don’t think its “wrong” to steal food, they just don’t have that gene.
Given that the food is usually the positive reinforcement when you are conditioning animals and that is one of their goals for which they change their behavior it is indeed quite hard to reinforce them not to eat when presented with the option.
You could reward at random time intervals if the behavior persists during the interval (for persistent behaviors like not hitting), or after a random number of repetitions of the behavior (for discrete behaviors). I’m not actually sure why random reinforcement works better than systematic, but I expect the effect to apply here.
The problem of what to reward is harder, but maybe you could make a list of every absolute demand and stick to that? Parents always drop some of their nice-to-haves because otherwise the kid can’t do anything right so you’re better off letting them draw on the wallpaper if that means they’ll stop sticking forks into plugs. (Also the Chaos Legion demands that you ask “Is that actually bad?” when a kid does something unexpected but not so obviously bad you didn’t think of it.)
Questions for both you and Tenoke:
How do you reward/reinforce desired behaviors in cases when the desired behavior is normal rather than exceptional? If I give my child a reward every time they’re not hitting their sibling, isn’t this isomorphic to taking away their expected reward every time they do hit their sibling (a punishment)? Worse, if I only started this “reward basic lack of misbehavior” scheme when I noticed that a child was prone to misbehavior, isn’t that just going to be perceived as “[I/my sibling] got a new nearly-continuous reward for misbehaving enough to trigger parental notice”?
Then, since it’s significantly harder to apply reinforcements than punishments to badly-behaved children, wouldn’t we expect to see a strong correlation between reinforcement and good results (or between punishment and bad results) regardless of how effective each was at changing behavior?
You don’t reward the omission of bad behavior, you reward good behavior. For example you reward your child every time they are good to a sibling. Sure, it is harder to look for good behaviors and to reward that instead of simply punishing the child when they do something bad but punishment doesn’t work well as mentioned earlier and for example often leads to situations where the child decreases the undesired behavior but only when it is being watched. Also whenever you can you should choose intrinsic or ever extrinsic reinforcements.
Talk about building a system that can be gamed! My big dog does not steal my little dogs food except when he can. And in between they play together pretty amusingly. I don’t think I would gain anything by rewarding them playing together in terms of a lower rate at which the theft of food occurs. The problem being, my big dog (and probably my little dog) don’t think its “wrong” to steal food, they just don’t have that gene.
Given that the food is usually the positive reinforcement when you are conditioning animals and that is one of their goals for which they change their behavior it is indeed quite hard to reinforce them not to eat when presented with the option.
You could reward at random time intervals if the behavior persists during the interval (for persistent behaviors like not hitting), or after a random number of repetitions of the behavior (for discrete behaviors). I’m not actually sure why random reinforcement works better than systematic, but I expect the effect to apply here.
The problem of what to reward is harder, but maybe you could make a list of every absolute demand and stick to that? Parents always drop some of their nice-to-haves because otherwise the kid can’t do anything right so you’re better off letting them draw on the wallpaper if that means they’ll stop sticking forks into plugs. (Also the Chaos Legion demands that you ask “Is that actually bad?” when a kid does something unexpected but not so obviously bad you didn’t think of it.)