Learning by Trial and Error -Part 3 – Negative Reinforcement (R-)

 

Trial and Error Learning – Part 3 – negative reinforcement.

Negative reinforcement – the removal (-) of a stimulus following a behaviour which causes that behaviour to increase.

As discussed in the previous blog, reinforcement increases behaviour. When we negatively reinforce a behaviour, we remove something unpleasant in order to increase the preceding behaviour. Now the stimulus we are removing needs to, by definition, be unpleasant enough that we are motivated by it to want it gone. It doesn’t need to be painful, although it often is, but it has to be unpleasant.

To give some examples

1. My dog gets off his bed. I stare hard at him and tell him to go back to bed, putting social pressure on him, until he goes back to the bed. When he gets back onto his bed, I stop staring at him. The stare down, is unpleasant for the dog, and it’s removal reinforces going back into bed. Staying in bed is increased.

2. I put an electric collar on my dog (I don’t, this is an example). I let my dog off lead and recall him and then press the button to shock him. When he returns to me, I release the shock. The removal of the electric shock reinforces the recall.

3. I am standing on the train and someone stands too close to me making me feel uncomfortable. I take a step back to feel more comfortable. Me taking a step back has been negatively reinforced be relieving the discomfort of him standing too close.

4. I have an abscess in my tooth. I go to the dentist and he removes it, relieving me of the pain. My going to the dentist has been reinforced by the removal of the pain.

5. I have an eyelash in my eye. I go  to the mirror and remove it. My actions of removing the eyelash have been negatively reinforced by the relief from the discomfort.

6. My dog barks at a stranger because he is frightened. I wait until he is quiet before moving away. My dog’s cessation of barking has been reinforced by me moving away from the scary stranger.

In horse training, this is referred to as “pressure/release”. Pressure is put on the horse and released when the horse performs the desired behaviour.

Now, I’d like you for a moment to consider the human examples above, especially the tooth ache one and ask yourself this

“Is the relief I feel from that pain worth the pain or discomfort in the first place?”

Would you rather not have had the pain at all? A few years ago I injured my back and attended a chiropractor. After going to see him, I felt so much better, less pain and better movement. But guess what, I didn’t want the pain to start with and then actively engaged in making my back stronger so that I wouldn’t feel that pain again.

So, any time we use negative reinforcement to change behaviour, we are by necessity introducing a painful/uncomfortable/annoying sensation into the scenario which the dog will come to associate with it. Negative reinforcement can be used successfully, but there is always the risk of fallout and in the vast majority of cases it’s unnecessary as the behaviour can be taught with positive reinforcement.

Constructive comments and questions are welcome as ever.

Next time, negative punishment – removing something the dog want in order to reduce behaviour.

 

Advertisements

First blog – republished from March 2012

The below is the original text from the blog I published in March 2012. I now no longer use no reward markers and absolutely do my best not to use any aversive training, whether physical or psychological. Constructive comments and questions are welcome as always.

 

Jesse

 

I started training my own dogs about 10 years ago. At the time we had Superstar Mollie, who was a tricoloured border collie cross. She was a little highly strung but sharp as a tack when it came to training. Bosco, my boy, came to us in January 2001 at 8 weeks old. He was a Dogue de Bordeaux, the most handsome example of the breed I have ever seen (although, admittedly, I am a little biased). Not too tall and built like a tank, he loved a ball, tug games and treats but unfortunately, I learnt compulsion training as my start and we didn’t have an awful lot of fun during the first few years of his life as I had been taught wrongly that he was being dominant and that I had to control him through rank reduction programmes and the like.

A year later, I got our sweet sweet girl Kitty, the Neapolitan Mastiff. A great example of the breed, tight skin on her body and only a little loose skin and she could move like a greyhound. I’ll post more about the dogs individually later.

Being a cross over trainer, I have the benefit of experience of both camps. I now use no physically aversive techniques and only occasionally use some techniques which may be mildly aversive psychologically. I can categorically state that aversive techniques do work in training and can have a short term effect on the manifestations of some problem behaviours. Do I agree with them any more? Absolutely not. I have had a few discussions both online and at conferences on the use of aversive training and am also very careful and deliberate to state that I no longer use nor agree with punitive training. Having said that, when I started training all those years ago I successfully taught all my dogs to walk on a loose leash, recalls and down-stays using a prong collar.

Skinner proved all those moons ago that positive punishment does work. I am not proud of using these techniques in the past but I am where I am today, in part, because of them and my knowledge of them, although my relationship with my three very special dogs unfortunately suffered as a result and due to the early death of Bosco, I didn’t get the opportunity to fully repair that but I hope I did go some way towards it.

Which brings me to the way I train just now. I’ll post later about the influences which brought me fully to where I am but basically I started to read everything I could. Ian Dunbar has played a massive roll in my learning, along with Stanley Coren, Jean Donaldson, Sophia Yin, Ray Coppinger and many others.

As I got better at training, friends, family and work mates would ask for advice and my girlfriend encouraged me to start my own business and make it a bit more official. I take on most cases from recall and general manners to separation distress and aggression cases. Most of my experience comes from dealing with dog-dog aggression as I have suffered through it myself (again, more later)

The most I now give to a dog is a verbal non reward marker and rely heavily on counter conditioning through positive reinforcement and am using Grisha Stewart’s Behavioural Adjustment Training which uses negative reinforcement, for aggression cases and am having good success with it.

That’s all for now folks, constructive comments are always welcome.

John McGuigan

Glasgow Dog Trainer and Behaviour Consultant

http://www.glasgowdogtrainer.co.uk

Part 2 – Positive Reinforcement – R+

Rico3

 

Learning by trial and error part 2 – Positive reinforcement.

Positive reinforcement (R+ for short) is defined by adding something (+) to the dog’s environment after he does a behaviour in order to increase (reinforce) the likelihood that the behaviour will re-occur.

The “positive” part is not to define good or bad, it is a scientific term which means adding, like in mathematics. R+ is good however. The dogs likes positive reinforcement because the the thing which we are adding has to be enjoyable for it to reinforce the behaviour. We cannot add something unpleasant after a behaviour and expect that behaviour to increase.

So a few examples for both us and our dogs.

1. Dog sits and gets a treat, sitting is reinforced by the treat.

2. Dog barks and we throw the ball. Barking is reinforced by the ball throw

3. Dog jumps and we tell the dog to get down and push him away. The jumping is inadvertently reinforced by the attention and touching (this is not desired)

4. Dog barks at the window and we shout at him. Barking at the window is unintentionally reinforced by the shouting/attention (again, not desired)

5. I eat a donut and like the taste, it makes me happy (Pink Jammie on a Friday for me). The taste and release of the feel good hormones in my body reinforces the eating of the donut (even though I know they are not good for me).

6. I watch my favourite TV programme. The enjoyment I get reinforces watching it, I am more likely to watch it again next week.

7. I do a good job and the boss gives me time off. Giving me time off (addition of free time) reinforces the work I have done, so I am more likely to do a good job in the future.

Now please remember, it is the effect that addition has on the learner’s behaviour which is important not the intention of the person who is applying it. To give some examples of intended reinforcement (reward) which do not reinforce behaviour –

1. I clap my dog on the top of the head after he recalls and tell him “good boy”. Since most dogs don’t like being clapped on the top of the head, even though I am trying to be nice and praise him, if it doesn’t actually reinforce the recall, i.e. make it more likely the next time, it’s not positive reinforcement.

2. My dog is feeling ill and I give her a treat after she sits. She does not want to eat, so the treat is not reinforcing.

3. My boss, who I don’t like tells me the work I have done has made his life easier and thanks me for doing it. I don’t like him, couldn’t care less about his life at work being easier and he can stick his thanks. He is meaning to reinforce my willingness to work but it doesn’t as it doesn’t mean anything to me.

So in the last three examples, something has been added which the person adding it thinks is pleasant, but if it doesn’t change behaviour then it’s not reinforcing.

On the other side of that, in the example of the barking at the window above, we are not intending to reinforce behaviour, but we may end up doing just that.

Last year I had a client with a big dog, over 35 kilos, who had trained 4 really obnoxious behaviours really reliably all by telling his dog “NO!”.

If your dog is still doing some unwanted behaviour then something is reinforcing it.

If your dog isn’t doing a wanted behaviour, then we are not reinforcing that behaviour well enough.

Questions and constructive comments are welcome as ever.

I will cover negative reinforcement (removing something unpleasant to increase behaviour) in the next part.

Learning by trial and error – part 1 – introduction

 

mewat

This is my first written blog in a a good while. I have been motivated to write again for a number of reasons, not least the amount of bad information out there on dog training, what works and what doesn’t and why it works.

Learning by trial and error, or Operant Conditioning, is the way organisms, including us and dogs, work out which behaviours work and which don’t. There are four factors at play during operant conditioning, positive reinforcement, negative reinforcement, positive punishment and negative punishment.

We also need to have a basic understanding of learning by association which I have covered in my previous blog

https://glasgowdogtrainer.wordpress.com/2014/07/27/learning-by-association-classical-conditioning-a-powerful-tool-in-dog-training/

 

I’ll take a step back briefly and explain a little about learning by association (classical conditioning). Learning by association is always going on every time we interact with our dogs.

In high school, I studied both French and Latin. My Latin teacher was an amazing man and teacher and I thoroughly enjoyed going to his classes. I did well in the exam and enjoyed the learning process. My French teacher on the other hand, although a good enough teacher, shouted at us, berated and belittled some of us. I passed French but did not enjoy learning French.

The method in which we learn is not only important, it is crucial. We can learn in many environments, but learning in an environment which promotes learning, makes us better learners.

Our dogs absolutely can and do learn using e-collars, shouting etc, but I’m pretty sure they don’t enjoy the process.

By teaching using only non aversive training, if done properly, the dog enjoys both the learning process and performing the task she is learning.

So to start, I’ll define what the scientific definitions of reinforcement and punishment are (remember, these are not the common uses).

Reinforcement – anything which increases the duration, intensity and/or frequency of the immediately preceding behaviour.

Punishment – anything which decreases the duration, intensity and/or frequency of the immediately preceding behaviour.

In short, reinforcement trains more behaviour, punishment trains less behaviour.

Examples
1. Dog sits and gets a treat – sitting is reinforced (hopefully)

2. Frightened dog lunges and barks at scary man with beard. Scary man goes away -lunging and barking is reinforced.

3. Dog pulls on lead and is corrected on a metal or shock collar – pulling is punished

4. Dog barks at us and we immediately leave the room – barking is punished.

Now remember, when doing this, it actually has to effect the behaviour, our intention is irrelevant. So, if your dog is barking and you tell him to shut up, the dog actually needs to reduce barking in the future, not at the time, for shouting to be an effective punisher.

In the next part, I’ll discuss positive reinforcement.

Constructive comments and questions are welcome as always.

John