Operant Conditioning

[Luke GSP]

This may be a dumb question ;) but I have always been a slow learner. Am I correct in thinking that All forms of training ie influencing a behavior is operant conditioning? I have racked my brains and cannot come up with any training method that i know of that cannot be explained with either -/+P -/+R or E

[Luke W]

Nope Well...depends on your definitions of 'training'.

Habituation/Desensitisation?

http://en.wikipedia.org/wiki/Learning

[Luke GSP]

Nope Well...depends on your definitions of 'training'.

Habituation/Desensitisation?

http://en.wikipedia.org/wiki/Learning

i am talking about influencing a behaviour in the sense of

I want you to do this or i dont want you to do this "voluntary" action not classical conditioning in the form of a pavlovs dogs way or in a response/behavior way like desensitization although isn't desensitization extinction in a sense?

Edited May 12, 2010 by Luke GSP

[szafir]

Classical conditioning often features in dog training/learning as well where previously unconditioned stimuli become paired with an unconditioned response to result in conditioned responses to a conditioned stimulus. For example, lets say a dog inherently likes going for walks and gets excited on them. This is an unconditioned stimulus because it's the dog's natural reaction. Before you go for a walk, you always pick up your keys. Originally this is an unconditioned stimulus, because on its own picking up keys means nothing to the dog. But paired together enough times (picking up keys always results in a walk) means the dog will become excited upon hearing the keys being picked up because it has paired the action with going for a walk. The keys have become a conditioned stimulus for a conditioned response (getting excited).

[Aidan]

OC & CC are not completely independent. This is a really big topic and not all the questions have been answered. OC is the best way we have to describe how operants are changed.

[Luke GSP]

Nope Well...depends on your definitions of 'training'.

Habituation/Desensitisation?

http://en.wikipedia.org/wiki/Learning

i would have thought that other than bridge items such as charging a clicker classical responces would tend to be unusable as a behaviour you would wish to harness as an outcome? However, is drive training classed as harnessing a classical conditioned responce to then develop an opperant conditioned outcome? ie drive training toy used to evoke drive initiation (classical conditioning?) once drive peak is iminent toy removed or neutralised, command given behaviour/action displayed toy returns or brought back to life = drive satisfaction and reinforcement through +r. Please note Not a statement just typing as cogs grind in brain so probably utter KAKA!

Edited May 12, 2010 by Luke GSP

[corvus]

At the very end of the NDTF conference Steven Lindsay said he didn't think clicker training was really either OC or CC. He said clickers are a teensy bit startling and that's why they work abnormally well as a marker. He likened it to tracking, in which a dog is driven by an innate desire to gather information. The dog doesn't know what's at the end of the track it is following a lot of the time, so how can one say he's doing what he's doing for an expected reward? Rather, he's responding to the surprise of swinging his head around and finding the scent, then losing it, then surprise - there it is again! I think what he was getting at with clicker training is that the clicker adds a dimension that kind of mimics that inherent seeking of information, thus taking it beyond OC, although that's not to say OC is not fundamentally involved as well.

Similarly, there are arguments that social learning falls outside of OC and CC. There's a paper somewhere that studies how puppies can learn a lot of the basic scent detection stuff from watching their dam working. I believe it's common practise to let future herders watch accomplished herders do their thing as well. At least in some circles. However, whether this can be of benefit to humans training dogs without doggy demonstrators is questionable. I read a paper recently that reported on a study in which human demonstrators basically meant nothing to a dog, while dog demonstrators sped up the training of some dogs (the submissive ones, as it turned out, but not the dominant ones). We do know that dogs tend to look at us for direction when they get stumped, whereas wolves don't. So maybe there's something that can be done with it.

As far as CC goes, I taught my last dog to shake the water from her coat on cue using CC. I just said "have a shake" whenever she shook and in the end she'd do it when I said it. Always smart for a dog that likes to swim in the lake and then comes and stands right on top of you before having a shake. I taught obstacle avoidance on walks the same way ("this side" and "other side"). And directions (like "this way"). I suspect there were may have been very low level OC going on, but it was all very passive.

I think that drive training is a case of Establishing Operations. We're talking about arousal and motivation levels. And possibly Fixed Action Patterns as opposed to classical conditioning. I always get flamed whenever I open my mouth about drive, so maybe I'll leave that one alone (OC all the way!).

[Aidan]

Insight Learning is sometimes given as an example of learning that isn't OC, but I'll let you form your own opinion of that after watching this:

[Luke GSP]

At the very end of the NDTF conference Steven Lindsay said he didn't think clicker training was really either OC or CC. He said clickers are a teensy bit startling and that's why they work abnormally well as a marker. He likened it to tracking, in which a dog is driven by an innate desire to gather information. The dog doesn't know what's at the end of the track it is following a lot of the time, so how can one say he's doing what he's doing for an expected reward? Rather, he's responding to the surprise of swinging his head around and finding the scent, then losing it, then surprise - there it is again! I think what he was getting at with clicker training is that the clicker adds a dimension that kind of mimics that inherent seeking of information, thus taking it beyond OC, although that's not to say OC is not fundamentally involved as well.

Similarly, there are arguments that social learning falls outside of OC and CC. There's a paper somewhere that studies how puppies can learn a lot of the basic scent detection stuff from watching their dam working. I believe it's common practise to let future herders watch accomplished herders do their thing as well. At least in some circles. However, whether this can be of benefit to humans training dogs without doggy demonstrators is questionable. I read a paper recently that reported on a study in which human demonstrators basically meant nothing to a dog, while dog demonstrators sped up the training of some dogs (the submissive ones, as it turned out, but not the dominant ones). We do know that dogs tend to look at us for direction when they get stumped, whereas wolves don't. So maybe there's something that can be done with it.

As far as CC goes, I taught my last dog to shake the water from her coat on cue using CC. I just said "have a shake" whenever she shook and in the end she'd do it when I said it. Always smart for a dog that likes to swim in the lake and then comes and stands right on top of you before having a shake. I taught obstacle avoidance on walks the same way ("this side" and "other side"). And directions (like "this way"). I suspect there were may have been very low level OC going on, but it was all very passive.

I think that drive training is a case of Establishing Operations. We're talking about arousal and motivation levels. And possibly Fixed Action Patterns as opposed to classical conditioning. I always get flamed whenever I open my mouth about drive, so maybe I'll leave that one alone (OC all the way!).

But I thought that classical conditioning was an involuntary response to a stimulus so wouldn't shaking the coat be a vountary action and hence if praised afterwards be an example of +R???

that is why I would have thought that drive would be classical conditioning as I thought that it was an involuntary response to the stimuli ie the dog does not decide to become aroused but rather it depended upon when the stimulus met the dogs threshold?

[Luke GSP]

Insight Learning is sometimes given as an example of learning that isn't OC, but I'll let you form your own opinion of that after watching this:

Great video, hard to see what the bird is trying to get to, is it food? If it were I would have thought that it was still OC as their was an applied reinforcement at the end ie i did this and I got a good thing? I suppose that the main difference is that the animal is performing a task to solve a problem with no pointers or cues but I would have thought that the outcome was still +R????

[corvus]

Yeah, the coat shaking wasn't quite CC. But it was pretty close in that to begin with I was giving the cue (stimulus) after the event and then praise right on top of it. Once she was doing it when I said it I praised afterwards and then it was R+. Some of it was good timing, though, and probably snuck in some OC in the learning phase.

My caution with arousal in general is that it's a slippery thing and I don't think it's well understood. From what I can tell, arousal is thought to be affected by either exercise or emotional state. Emotional state currently sits outside of OC. Although it's important to note that it doesn't exist in isolation of associative learning. It's more like an added layer, or maybe OC is an added layer on emotional state. You can condition an animal to go into an emotional state on cue, but to me I think of it as a two-way street between behaviour and ES, not all the one thing. I dunno, I'm not explaining this very well.

Whatever the case, when you teach a dog to go "in drive" on cue, by pairing the cue with the drive reward that results in the level of arousal/motivation you want, I think it's debatable that this is CC because I think that what you're essentially doing is using a setting effect, as in, "prepare for imminent crazy fun" (classically conditioned). That in turn switches on the anticipation and that feeds into emotional perception and that feeds into arousal. So I'm hesitant about just saying it's classical conditioning. To me, it's a bit more than that. Don't think I explained that very well, either.

[corvus]

Oh, wrt the pigeon, I think the overall result might have been R+, but there were lots of behaviours the bird did that weren't reinforced in order to get the reinforcement. Assuming the bird had not solved this problem before, it solves it without much trial and error.

[Aidan]

Oh, wrt the pigeon, I think the overall result might have been R+, but there were lots of behaviours the bird did that weren't reinforced in order to get the reinforcement. Assuming the bird had not solved this problem before, it solves it without much trial and error.

That's right, novel problem for the bird. The reinforcer is a small banana, for the record. Lots of operants emitted without a primary reinforcer.

This is pretty old, I forget when (Kohler first demonstrated it in animals, not the dog training Koehler), but it is still taught and used in models of behavioural science. Perhaps these days we would look at things like Panksepp's model of the "SEEKing circuit" or EEG scans of the front brain and suggest that each step closer to the banana was an intrinsically reinforcing event?

How might this apply to teaching loose leash walking (or teaching a dog to pull on the leash), for e.g.?

[Aidan]

So I'm hesitant about just saying it's classical conditioning. To me, it's a bit more than that.

There's always a bit more than that. You can teach a dog to associate the reflex of shaking when wet with a command, but those operants (actually shaking) must be reinforced (by shedding water).

[Aidan]

Social learning, some of you would be familiar with this because it is very famous:

http://www.youtube.com/watch?v=lCETgT_Xfzg

[corvus]

So I'm hesitant about just saying it's classical conditioning. To me, it's a bit more than that.

There's always a bit more than that. You can teach a dog to associate the reflex of shaking when wet with a command, but those operants (actually shaking) must be reinforced (by shedding water).

True, but I meant "a bit more" in the sense that I think there are some essential components to the manipulation of drive that fall outside of associative learning. But maybe I'm making a distinction for drive training that exists in any training.