Positive Reinforcement = adding something to re-enforce
Negative Reinforcement = removing something to re-enforce
Positive Punishment = adding something to punish
Negative Punishment = removing something to punish
Punishment = something done to bring down a behavior in frequency and/or intensity.
Reinforcement is something done to increase a behavior in frequency and/or intensity.
So this person who wrote the following on an e-mail list I am on is using the terms incorrectly.
"I can get the other two dogs to go down the steps using mild negative reinforcement: stepping into their space and telling them to go down."
So I think what she is actually doing is using positive punishment. Stepping into their space (adding) to decrease their behavior (not going)?
-------------------------------------------------------------------------
What are R+, R-, P+, P-, and extinction?
You'll see these terms pop up on the list fairly often. They are terms used in Operant Conditioning. They stand for:
Positive reinforcement (R+): something is added to increase the occurrence of a behavior. For example, you give the dog a treat to reinforce the sit.
Negative reinforcement (R-): something is removed to increase the occurrence of a behavior. For example, in the forced retrieve (not a clicker technique), a dog's ear is pinched to force him to open his mouth to take the dumbbell. When the dog opens his mouth -- the desired behavior -- the ear pinch is released.
Positive punishment (P+): something is added to decrease the occurrence of a behavior. For example, a collar pop is applied (not a clicker training technique) to discourage a dog from pulling on the lead.
Negative punishment (P-): something is removed to decrease the occurrence of a behavior. For example, turning away (removing attention) from a dog who is jumping up on you.
Extinction: Lack of reward. A behavior weakens because it is not reinforced -- not because it is punished. For example, a dog wants a bone on the other side of a fence. It reaches its paw under and can't get it. It will try repeatedly, but because the "reaching under" behavior isn't reinforced (by getting the bone), the dog will eventually quit trying. Extinction will not work for self-reinforced behaviors.
Is it important to know these definitions? Yes, it is, because it helps us understand each other much better. For example, in everyday usage, positive and negative often mean good and bad. However, on this list, positive and negative mean something added or taken away. Punishment is another word that carries strong connotations in everyday language. In Operant Conditioning, punishment means strictly "reducing the occurrence of a behavior."
---------------------------------------------------------------------------
From an e-mail
Order of training strategies , always giving the animal a chance beginning with most positive first: R+ (Reinforcement positive bringing something into the environment like a reward), P-(Punishment negative such as ..take away, a time out), R- ( Reinforcement negative such as escape , moving away from something .. it increases behavior ) and last P+( Punishment positive ...adding an aversive)
Note: P- is the removal (or time out) of the chance to earn R+. With R-and P- , there are no aversive's which is something we want to avoid if possible first .
(Remember ,punishment, either + ( plus) or - (minus ) DECREASES a behavior.... and Reinforcement,either + (plus ) or - (minus ) INCREASES behavior)
Hey Lisa, I checked out your blog. Negative reinforcement and positive punishment both involve an aversive, something the dog doesn't like. They can be intertwined. You have to look at whether you are trying to encourage a behavior (reinforce) or discourage (punish). In the steps case, I am applying an aversive (body pressure) to encourage them to go down the steps. I walk into them, but immediately when they go down I move back (release the pressure). I left that part out (and it is crucial!).
ReplyDeleteOne characteristic of R- is that the dog's response is to do the minimum necessary to get away from the aversive. It doesn't make for a gleeful response to a cue. I had just decided today to work out a way to positively reinforce them going down the steps without my going every time. I'll have to toss treats down or use a treat delivery device.
Back to R-/P+: If the dog were trying to come _up_ the steps and I walked into her and pushed her back with my body, that would be P+ because I would be discouraging (punishing) the behavior of coming up the steps.
Most clicker trainers avoid R- and I certainly don't use it much. It is very hard to avoid it completely because it's easy to do unconsciously. I just try to be clear and honest about it when I do.
eileenanddogs
Hi Eileen, Thanks for commenting! I realized that this is Yet another part which I need to understand before I can start training Baron with a clicker.
ReplyDeleteI guess my confusion layed in the part where your actions of stepping in their space = them going down the stairs. I understand the stepping in is a positive (adding something) but since the action caused (for lack of better words) them to go down the stairs is where I mistaken it to be punishment.
Re-reading the notes about what a punishment is, you aren't trying to "bring down a behavior" but rather increase a behavior so that would be a re-enforcement.
Thank you for commenting and helping to clarify this for me!!
Hi again Lisa,
ReplyDeleteIt's fun to mull over the quadrants. You have a good point about "adding" the walking into them behavior and I had to think about that a moment. I think if there is an aversive being used, it has be introduced sometime. The adding in isn't the important moment, it's that they can and do escape from it, and in so doing perform the desired behavior.
I look for negative reinforcement to be the operator whenever there is escape, avoidance, release of pressure, removal of threat. (With humans, the threat of punishment can constitute negative reinforcement!) Horse training is full of it. Squeeze your legs together and the horse goes faster to avoid it. When horse speeds up you release the pressure.
Luckily these days there are a lot of people who are starting to train horses with mostly R+.
eileenanddogs
Thanks Eileen for your thoughts. I think you saw the list conversation as well. It seems like the whole negative/positive thing takes a bit to really think about.
ReplyDeleteGlad you posted a comment to help me start thinking about it because it would be good to consider my approach before I start training and to figure in if my approach is negative/positive punishment or reinforcement.
Well I'm really glad you started the discussion because today I started retraining my dogs positively to go down the stairs. I am also going to make a "before and after" video of it to show people the behavior difference in a cue trained by R- and one by R+.
ReplyDeleteI think you are wise to explore all this on the front end so you will literally know what you are doing.
eileenanddogs