One of my favorite, and first (written April 20, 2012), Training thoughts was this one. This one is so important for understanding why our dogs appear to regress, and why it is so important to be patient with the learner and be accurate in our feedback.
“Hi, my name is Dr. Kelsey Ford. You might know me by my casual name of Kelsey (that’s what all my friends call me) or my fancier name of Trystyn’s Cat Burglar, but I thought I should formally introduce myself. You see, I’m a scientist, a real, professional scientist, and I thought if I told you about my side of the story, it might help you with your dogs (who are also certified scientists too, by the way).”
Okay, well maybe my dog Kelsey can’t really talk, but I did have an insight the other day that, along with some previous opinions, really helped to clarify some of what I notice with dogs and their trainers in the sport of agility. (Of course, this applies to all areas of dog training, but I shall keep the analogies to agility here.) I’m a behavioral scientist by trade (I research human behavior in organizations), and work in the realm of “science,” and I realized my furry family members were far more apt at science than the average human being (and definitely many scientists, myself included). So, while this story might seem a bit odd, I wanted to share it with you as it highlights why certain things in training are so darn important. So, on with my training thought….
Science can be defined as, “the methodical approach to the acquisition of knowledge” (Koning, 1994). To that end, our dogs are scientists as they seek to acquire knowledge of their universe (a.k.a. their life with you in your household and family). Dogs are phenomenally methodical in their acquisition of knowledge. Have you ever felt your dog was “testing you” to see where the “line was truly drawn”? I used to say that Brooke was always trying to renegotiate “our contract” every 6 months. That is because they have been methodically acquiring knowledge on their universe, and finding out what it takes to get their cookie, toy, ear scratch, cuddle, chase. In fact, it is my opinion that they are far more methodical and better scientists than us humans as it is their full-time job and passion. They aren’t distracted by kids, TV, the internet, spouses, the phone, books, etc. They are constantly doing science, except for when they are sleeping, perhaps.
How does a scientist acquire knowledge about their universe? Well, we hypothesize, intuit and guess what might be reality (a.k.a. the Theory), then we test it. The results of the test then either support the hypothesis or refute it. From that information, we either seek to re-test the hypothesis in another context to see if it still holds true (if it was supported) or we amend our guess or understanding of our universe and test again (if it was refuted). In good science, we continue that until we find the boundaries under which the hypothesis is no longer true. In other words, we are continually seeking to refute it. That’s how our theories (or understanding of our universe) come to be.
Now, there are two sides to the coin. There is the side in which our understanding is true, and the side where our understanding is false. In the dog world, the “theory” in this analogy is our criteria for our dog’s behavior, which is the “true” side of the coin. All other behavior that fails to meet our criteria is the “false” side of the coin. For example, for the tire jump, the criteria (theory) is that the dog must jump through the tire. The false condition is that the dog does anything but jump through the tire (e.g., sneaks under the tire, runs around the tire frame, stops short and refuses, walks on the tire, breaks the tire).
As the scientist, the dog tests their hypothesis by providing a behavior. In other words, Dr. Kelsey Ford asks me, “If I do this, does that get me the cookie/toy/praise/the chance to do more equipment?” The results of her test (i.e., my responses) tell her if her hypothesis was correct or wrong. My responses have two options: positive or negative. In other words, my “positive” responses tells her, “You were right!” and my “negative” response tells her, “Nope – wrong guess!” (To be clear here, the “negative response” is the lack of reward or in more severe cases, the removal of an inappropriate reinforcer.)
Thus, when we combine the actual criteria (theory) and the test results, we have four possible combinations: 1) the hypothesis is correct and the feedback properly indicates that; 2) the hypothesis is incorrect and the feedback properly indicates that; 3) the hypothesis is false, but the feedback indicates it is correct (false positive); and 4) the hypothesis is correct, but the feedback indicates it is wrong (false negative). (See the diagram below for the illustration.)
So, let’s get down to the important points. All excellent scientists will test both “the positive” and “the negative” – i.e., the conditions under which the criteria are true and the conditions under which the criteria are false. In dog-speak, all dogs will test their understanding by jumping through the tire, and all other possible behaviors, including going under and around the tire! That is, the dog will not fully understand and learn the criteria until they have tested both the positive and the negative/false. Just like humans and our science, we don’t have a valid theory until we test both Quadrants I and II. However, humans tend to be lazy scientists and we have a bias towards testing only Quadrant I (after all, it is hard for us to publish research for Quadrant II as it’s not “interesting”). Dogs don’t have the same level of bias, though, which is why I think they are better scientists than us!
As the dog trainer, you are the results provider – you are responsible for providing the correct matching test result. If the dog has met the criteria – the response should indicate that! That’s Quadrant I (the valid positive). When my dog jumps cleanly through the tire – she needs to get the result that indicates her hypothesis was correct! If my dog has not met the criteria (gone under the tire) – my response must indicate that too (no reward, no reinforcement)! That’s Quadrant II (the valid negative).
Now, I don’t know about you, but I’m human. This means there are days when I’m a bit klutzy with my feedback. There are times when I don’t even realize that Dr. Kelsey Ford is testing a hypothesis and I provide thoughtless / automatic feedback. These human errors mean my sweet scientist, Kelsey, has to deal with unreliable feedback. I sometimes (inadvertently or erroneously) provide false positive and false negative results to her tests (Quadrants III and IV). As you can guess, this makes it harder for her to accurately understand and predict her universe and my criteria for her behavior.
False positives lead to problems in training in that there can be chaining of inappropriate / undesired behaviors (like the dog turning their head after every piece of equipment), messy performances (paw paddling on the sit-stays), or even worse, trained non-responses (dog standing on the table because it has been rewarded to do so instead of doing a down). Even better yet, that it is a good thing to hop up onto the counter (getting food rewards because an absent-minded me left it out) – ARGH! (That’s definitely not a desired behavior or the criteria for her.)
False negatives aren’t as “bad” for dog training in that you don’t get as much reinforced messiness or behavioral problems, but it does lead to slower and less efficient learning, and can lead the dog to second guess the criteria unnecessarily. For some dogs, it can even lead them to completely abandoning their correct hypothesis, and stubbornly test a long series of incorrect hypotheses. This is a common problem for the untrained dog handler who is stingy with their praise and positive reinforcement.
Thus, as Dr. Kelsey Ford’s trainer and benevolent master of her universe [grin], it is my responsibility to do two things:
(1) Encourage her to “test the negative” and to not view it as a setback in her training. It would be wrong of me to think, “HUH! She knew it last week, and now she doesn’t know it at all!!!! Why is she going under the tire now!?!? Dumb dog! I must make it easier for her to prevent her from making this mistake!” Please don’t think that!!! Rather, embrace your dog’s scientific mind and simply provide her with the appropriate feedback of no reward (and no reinforcement including re-cuing), and in your mind say, “Feel free to adjust your hypothesis and test it again, my smart, little scientist!” Susan Garrett commonly refers to this as “embracing the failures” and she repeatedly argues the handler is responsible for pushing the envelope with the dog to provide the dog with the opportunity to learn “what the criteria is not.” I like this “providing your dog the opportunity” to test the negative, but I also think it is very important to highlight it is not a sign of “forgetting” on the part of your dog or stupidity of your dog or its failure to learn. It is the complete opposite of that!
(2) Do all within your power to ensure your feedback is accurate! Do all within your power to prevent false positives and false negatives. Just imagine for a moment what it would be like for a woman to get false positive or false negative, depending on her circumstances, on a home pregnancy test…. Dogs can experience the same type of frustration with false positives and false negatives from their trainers. To that end, a lot of what Susan Garrett, Bob Bailey, and Ian Dunbar promote for dog training is critical for this. Focus on your timing and delivery of your rewards to ensure that you are rewarding that which meets your criteria and not additional behaviors (e.g., false positives). Reward their choices and not the end of the sequence (e.g., prevent false negatives on their choice). Place your rewards in their ideal locations (on the line, not at random locations, again to prevent false positives). Be aware that your dog is likely testing hypotheses even when you are not doing a “formal training session.” Finally, keep your training sessions short so you are more likely to notice what the dog is doing and thus be correct with your feedback.
I hope this analogy helps you and your dog training. Just please, do be careful of providing sloppy feedback, and encourage your scientists to test the negative as that is what further refines their hypotheses and development of theory (i.e., understanding of your criteria)! It’s not your dog being “stupid” rather it’s quite the opposite – she’s being a methodical scientist testing her hypotheses!
References and Additional Resources:
Koning, Ross E. 1994. The Scientific Method. Plant Physiology Information Website. http://plantphys.info/plants_human/scimeth.shtml. (4-20-2012))
Garrett, Susan. Shaping Success (see www.clickerdogs.com)
Garrett, Susan, online courses: Shaping a Difference, 5 minutes to Brilliant Recalls, Puppy Peaks (http://www.puppypeaks.com)
Dunbar, Ian (see: http://www.siriuspup.com/about_founder.html)