This post goes with today’s podcast episode:
Imagine 4 slider controllers you can manipulate once your dog has learned the basics of a behavior: environment, arousal, distraction and freedom. Whether all 4 are relevant for a certain behavior will depend on the behavior in question and on your goals.
Personally, I like imagining 3 settings on each controller for most (not all) behaviors. For example, an environment can be easy, intermediate or difficult, arousal can be low, moderate or high, distractions can be absent, mild or strong, and in terms of freedom, a dog can be on leash, drag a leash or off leash; a dog can be in a small enclosed space, in a larger enclosed space or out in the real world …
You do, of course, not need to think in terms of 3 levels of difficulty for any of the slider controllers: you can have 2 or more than 3 settings; whatever makes sense for a certain behavior and for a certain slider.
A good rule of thumb is to only manipulate one setting at a time. This gives you, depending on what you are working on, a specific number of possible training conditions. Once your dog has successfully trained through all of them, your behavior should be VERY well proofed and generalized!
I am not a math person, but I wanted to think through an example (because why not).
Let’s assume that technically, our four sliders have 3 possible difficulty settings each. Let’s further assume we are working on a sit and want it to be generalized anywhere and everywhere.
Environment: we’ll choose at home (easy), on the sidewalk (intermediate) and at the mall (difficult).
Arousal level: we’ll choose a lazy dog after dinner (low), our dog after a brief afternoon walk (moderate) and our dog when looking forward to toy play (high – our dog is a ball junky mixy mix).
In terms of distractions, we’ll work without any distractions present (none), with food items on the ground (intermediate – our dog likes food but is crazier about toys) and with toys on the ground (our dog’s most difficult distraction).
In terms of freedom, we will hold on to the leash close to the collar (no freedom), hold on to the end of the leash (a little freedom) and then try off leash (full freedom).
I guessed that the number of possible unique training conditions would definitely be higher than 20 but likely lower than 100. I Googled aand found a formula that gave me an answer that was far higher than my frame (way above 100) and therefore likely wrong. My math PhD friend Jana came to my rescue and helped me determine that the correct answer is 81. Thank you, Jana!
If we wanted our dog to perform their sit (without duration or distance, which are not on my sliders) in all kinds of environments and with all kinds of distractions present and in different states of arousal, we could ask for it in 81 different ways and then I bet you that that sit (again, not a great example because it’s a behavior we probably want duration and distance on) can be carried out VERY well.
This is not impossible to solve without being a math person. If I hadn’t been lazy, I could have brute-force-thought my way through this problem by writing out the options. I’d have started like this:
- at home + low arousal + low distraction + no freedom
- at home + moderate arousal + low distraction + no freedom
- at home + high arousal + low distraction + no freedom
- at home + low arousal + intermediate distraction + no freedom
- at home + low arousal + difficult distraction + no freedom
- at home + low arousal + low distraction + some freedom
- at home + low arousal + low distraction + complete freedom
… and I would have ended up with 81 different training conditions that way. It’s doable; it just takes a bit of time.
If we added more sliders or more than 3 difficulty settings each, things would rapidly become un-brute-force-think-able (unless you have A LOT of time or are a lot faster than I)!
Update: Jana just made this amazing visualization (which feels as meditative as like looking at an aquarium to me) because … I don’t know, the dog people I know are amazing and also really good at this stuff?1 THANK YOU! I didn’t figure out how to embed the original interactive Python visual, so I screen-recorded it instead. Enjoy!
- Let me use this as an opportunity to thank the wonderful folks who have offered to help me with stuff like this just over the course of the last year! People have made awesome graphics for me (Chris Cernac), done math and made this visualization (Jana Fuchsberger), helped me figure out how to run online classes without knowing about programming (Zane Selvans) and offered to help me build an online learning platform if I wanted to have my own (Sukrita Sundar). THANK YOU to all of you!
Also, a big thank you to the amazing students and colleagues who have brainstormed with me, inspired me and agreed to be my guinea pigs as I try out new things … even when these things start out janky because I’m learning how to do them as I’m doing them! Thank you for sticking with me and concincing me that my time and my way of engaging with dogs and humans are valuable, and for sharing your dogs and your thoughts with me, inspiring me, surprising me and making me laugh and cry and everything in between!
Since I’m apparently writing a whole paragraph about who I’m grateful for, let me not forget the most important ones: MY PEOPLE on all the continents who I can rely on and reach out to anytime, no matter what and unconditionally. Moni, Helmut, Pablx, Rachel, Chris H. – know that I feel lucky to have you in my life. Hugs to you all except for Chris H. who doesn’t like hugs – a bottle of Veracruz sun and warm weather for you instead! You are a big part of what makes my life good, and you have unfalteringly been there when it wasn’t good and allowed me to be there for you. I love you all lots. ↩︎