TheOtherDave comments on How habits work and how you may control them - Less Wrong

64 Post author: Kaj_Sotala 12 October 2013 12:17PM

You are viewing a comment permalink. View the original post to see all comments and the full post content.

Comments (76)

You are viewing a single comment's thread. Show more comments above.

Comment author: pjeby 13 October 2013 06:14:29PM 11 points [-]

Seems like a real risk, but he seems to offer little support for it

Read the original book. Punishment is useless, you want negative reinforcement, and yes there is a difference.

"Punishment" is something bad that happens when you do something. "Negative reinforcement" is something bad that goes away when you stop doing something.

The trick is that brains have a kind of reinforcement kluge: instead of having an "avoid this, it's painful" circuit, we are reinforced by positive changes, including the removal of a negative stimulus.

So technically, the thing about punishment is, it's not really punishment. Animals and people don't learn to stop doing something in response to punishment, they learn to do whatever makes the punishment stop the quickest. If this happens to be avoiding the thing being punished, it's purely a matter of luck. They may also learn to say, hide their behavior from whoever's punishing it, run away, etc.

So the catch to all this is that self-punishment is useless because the fastest way to stop the punishment is just to stop punishing yourself in the first place. The only consistent self-punishment people can apply is the kind they've been trained to do by someone else -- i.e., the kind that they got rewarded (or negatively reinforced) for doing.

(Well, technically I suppose you might be able to train yourself to continue punishing yourself by rewarding yourself for punishing yourself, but...)

Comment author: TheOtherDave 13 October 2013 06:45:52PM 1 point [-]

If this happens to be avoiding the thing being punished, it's purely a matter of luck. They may also learn to say, hide their behavior from whoever's punishing it, run away, etc.

Indeed, they will often learn all of these at once, and then the punisher must do extra work to negate the latter set. So, yeah, negative reinforcement typically works better than positive punishment in the long run.
Negative punishment (that is, removing something good when I do something) can work OK too, though it has some of the same problems.
Training an incompatible behavior via positive reinforcement is often faster, though sometimes not an option.