Understanding Negative Reinforcement
Lili ChinThe "negative" in Negative reinforcement (R-) doesn't mean "bad" but R- is still often conflated with punishment (P+) and ranked LOW in hierarchical models of ethical training procedures. There is a persistent view of R- as a "bad" quadrant, which does not take into account that when caring for sensitive animals in uncomfortable situations, giving them the freedom to say "No", "Stop", "Help" and for us to listen and remove/reduce the thing that is causing stress is (R-) and this is a good thing. Trust and confidence are built on R- as they are on R+.
So I've been working on a new infographic with Caroline Crevier-Chabot and Sam Griffin - who suggested examples to illustrate, with help from Dr. Eddie Fernandez. We discussed what makes R- coercive or non-coercive. The difference comes down to 2 things: the aversive itself and choice. What happens if the animal doesn't do the desired behavior? Can the animal choose to terminate the interaction?
Free to download HERE, where hopefully, the words in the images are clearer and more legible!
11/19 Edit to add (in response to some comments received):
This infographic does NOT at all promote the coercive use of reinforcement (R- or R+) in training and communications. The purpose here is to clarify the concept of R- and show that good welfare/ethical interactions is not always simply defined by quadrants. For clearer explanations of what is meant by "choice" and "non-coercive" interactions, I found these webinars & presentations really helpful: Degrees of Freedom and Constructional Approaches (blog posts)
1 comment
This is good food for thought. I am, of course, in favor of letting animals opt out in ways to make sense to them and I recognize that that practice can negatively reinforce the escape behavior.
I am wondering about the definition of “coercion” you arrived at, though. Years ago, we were taught that when we held a puppy and it squirmed, we should not let it down no matter what until it got still. So stillness (after what could be a very long burst of activity, sometimes verging on panic), was negatively reinforced.
I believe that was coercive, yet it fits your definition of non coercive since “the animal had a way to terminate the interaction.”
It seems like the difference between coercion and non-coercion (if we can, indeed, make that distinction given the fact that we set up the session in the first place) has more to do with what it takes for the animal to terminate the interaction. The animal doesn’t have to be still or guess some externally applied requirement. The behaviors described are those we recognize as discomfort or escape. If the handler is “allowing” those to terminate the session, they are not setting a contingency. They are not, for instance, requiring that the dog turn away from the Dremel. The dog could also yelp, jump, lick their lips, etc. The bottom line is that you are describing automatic negative reinforcement rather than socially mediated. The handler is not putting escape on extinction.
I don’t think you have described the difference between choice and no choice. They almost always have a choice, even if it’s between unpleasant alternatives.
I think the animal being able to terminate the behavior is a great concept to have in the front of our minds. But by defining “non coercive” as the animal having the choice to do that, with no further distinctions, you have allowed in some coercive methods.
Thanks for contributing to this important topic.