High Value vs High Stakes: The Ethics of Positive Reinforcement

You hear it often. We dog trainers say it often. Use high value “rewards” (reinforcers) when training your dog.  In sports it’s all about creating a high-drive performance, and in real life it’s all about getting reliable responses to life-saving cues (think: recall). In behavior work, it’s often about successfully altering a dog’s behavior so that he can keep his home. It makes sense; reinforcement drives behavior and it is the learner that decides what is reinforcing. So, obviously, using a reinforcer that is highly valued by the learner will produce responses that are highly valued by the trainer. No one disputes that a reinforcer must be something the learner wants.

11942126_10206716373541037_3850026086041281098_o

These high-value reinforcers may be very high-stakes for some.

But what happens when the trainer takes this concept to a different level and uses a reinforcer that the learner not only wants, but needs? What happens when a trainer deprives their learner of basic needs in order to elevate the value of these needs as reinforcement? Further, what happens when the learner must weigh the value of the reinforcer against discomfort, fear, or even pain?

Positive reinforcement crosses a line. It becomes it’s own form of coercion. That’s what happens. Put anyone (dog, human, horse, dolphin, iguana, whatever) in a high-stakes situation and you can produce drastic behavior.  Otherwise upstanding citizens would steal to eat (or feed their children).  Addiction drives people to dangerous means every day.

If you’ve read my writing before or know me as a trainer, you know that I use positive reinforcement-based dog training, and that I am always after high value reinforcers for my dogs and those belonging to my clients.  I want strong behavior that comes with enthusiasm and holds up in the face of competing motivators.  I am certainly not here to argue that using high value reinforcers is unacceptable and we should all just use carrot bits and cheerios (although for your average labrador that’s as good as steak!). Instead, let’s examine some scenarios in which considering a lower-value, or lower-stakes reinforcer might be a good idea. 

Deprivation 

It has long been accepted in animal training to deprive your learner of a primary reinforcer in order to make that reinforcer more usable to the trainer. Animals used for entertainment might only be fed during performances, dogs trained for work might live in a kennel when off-duty, and far too many sport dogs spend long hours in crates, only coming out for food or play when it’s time to train or trial.

Let me be clear, there is a difference between feeding a dog its daily caloric intake via training and making food contingent on work.  “We will train for food and I will be sure you are fed plenty, no matter how training goes,” is fine while “you will only get enough to eat if you perform as expected,” is not. Perhaps more sinister is the deprivation of social interaction, but the same rules apply.

A better idea than deprivation is to always provide the dog with better stuff than he already has for training.  If he eats kibble in a bowl twice a day (as so many dogs do) then boiled chicken and string cheese might be a great high value reinforcer to him. If he eats raw meat (like mine do) kibble might be the hot ticket. If I want my dog to work for toys (and I do) I probably reserve the best toys for training, while lower-value items are available all the time for him to enjoy.

High Stakes Scenarios 

When trainers talk about high-value rewards they are doing so because dog training is all about competing motivators.  At any given moment we’d like the dog to be most motivated to do as we ask while the rest of the world is motivating him to do otherwise. Call your dog in the park and the squirrel is competing with whatever motivator you’ve got. Ask your dog to sit and stay while you answer the door and your motivator (cheese, chicken, ball, whatever) is competing with his desire to greet the guest (or scare off the intruder).  Because we are in the business of motivating dogs to do what we want and not what their normal dog brain tells them to do, we must come up with highly-valued reinforcers to get things done.

The trouble arises when the motivator competing with ours is safety or comfort.  Finding something a dog will risk its own safety for and then using that reinforcer in what the dog might consider a high-stakes scenario is a double-cross; a blatant abuse of power.  A dog whose relationship to her tennis ball is akin to the relationship of the addict to her drug of choice will risk life and limb for that ball. She will allow herself to be placed uncomfortably close to things that scare her.  She will do things that  make her very uneasy, or even things that cause her pain. Telling her you’ve got the ball gives you all the power in the world; she will now do anything. And we, as trainers, need to consider whether we think that’s right.

12109780_10206934735039938_8933617181428051438_o

This dog is learning to relax and disengage from a trigger. He is fed for returning to his handler, but he is not given the tug toy under her arm. When the tug toy is used he will push himself too far, tolerating things that do make him uncomfortable.

In my opinion it’s acceptable to use these high-value reinforcers in only low-stakes scenarios. There is no problem with asking your dog to perform a few behaviors she knows in an environment where she feels safe and then paying up with the ball. I’d even argue that teaching her a few new skills (skills that in no way put her at risk) in a safe place by using the ball is a great idea. But start to ask the dog to tolerate close proximities she’d never choose (I’ll give you the ball if you let this child pet you), to approach triggers she’d rather run from (go sniff that big man with the hat and here’s the ball), or to complete tasks that are scary for her (run this agility course with a judge calling out scores on a loud microphone, there’s a ball game in store) and you have crossed a line. You have successfully used positive reinforcement as coercion.

When using these high-value reinforcers our dogs may appear to overcome their fears or anxieties, when in reality the reinforcer has simply overshadowed these emotions in the moment.  As soon as that ball, that bite of chicken, or whatever it happens to be is removed from the scenario the dog is afraid again, and it is clear no healing has taken place.

So use hot reinforcers. Bust out the garlic chicken for dog training class, and the udder tug for that agility seminar. Just never forget the power you have over your dog when you take control over what matters to him, and vow not to take advantage of this power.

 

 

 

 


One thought on “High Value vs High Stakes: The Ethics of Positive Reinforcement

  1. Well said. When I began training, I had come out of an abusive relationship and I recognized the “extinction of reward” and positive reinforcement/ negative punishment– withholding, etc. as relationship and emotional abuse.

    I had to come up with some hard thinking to distinguish positive-only dog training from abuse and feel that I was training in an ethical manner. Your explanation of coercion is great– also, withholding essentials is extremely abusive.

    It should be “I get treats” not “what I need.” And this also covers physical affection and the normal things dogs do need to feel connected to the pack. If you begin withholding normal interaction, dogs can become extremely anxious. That’s controlling behavior.

    Your example of toys as bribes to overcome anxiety is wrong in a second way– a good trainer should not use positive reinforcement when counterconditioning a dog to fearful stimuli present in the environment.

    The rule is never to try oppose P+ (in this case, scary/anxious stimuli) with R+. It’s sensory overload.

    Use R- instead to deal with a dog’s anxiety. Let the dog walk away, stop fixating, calm down, and bring back up to when the dog starts getting overaroused/anxious, then turn away again. Work on the various sensory stimuli in a total scenario in isolation– sound, sight, smell– incremental exposure.

    Once the dog knows he can retreat from a fearful stimuli and take its time to study and learn about it, anxiety may be replaced by curiosity. I then reinforce that curiosity by allowing it to be indulged whenever safe to do so.

    If the dog is having isolation issues (i.e. perceived P-) then adding R+ such as comforting music/sounds, treats, etc. can all work to ease anxiety, as well as neutralizing that particular stimuli as a signal of punishment. (Which means, never crate your dog as punishment if you want the dog to like being crated.)

    Above all, consistency is really the best way to stop whatever you are doing from becoming emotionally abusive and confusing to the dog.

Comments are closed.