Training Conversations

Algorithmic Training vs Conversational Training

When starting out in dog training, or even just starting out in training something new that you haven’t trained before, you likely heavily rely on instructions. Those instructions tell you exactly where to train, with which equipment to train, which rewards to have on you, what to do to initiate the exercise (whatever this might be – from a sit to running contacts), what to do if the dog gets it right and what to do if the dog gets it wrong.

This is what I call “algorithmic” training. You are given a flowchart, and now you just have to run through the steps. Basically, at this stage the dog might just as well be trained by C-3PO. This is not bad. Everybody needs to start at this point. The problems arise when we cannot progress past the algorithmic stage.

Ideally, at one point we have a sufficiently good idea of the exercise, of what the dog is required to do and how we want to encourage this. In this moment, the algorithm becomes a conversation. We are no longer following the steps of someone else to speak to our dog – we are directly talking to him.

What is the problem with using an algorithm?

Let’s think of an algorithm we encounter every day of our lives: Siri. Everybody who has ever tried to have an actual conversation with Siri knows that there is a big difference between talking to someone following instructions about having a conversation (Siri) vs talking to a real thinking being (actual people, for example).

Ask Siri where the next Starbucks is for example. She will gladly answer this question. Now ask which drink to get at Starbucks. Siri cannot reply and does a web search for “Which drink should I get at Starbucks?” Gee, thanks for the great conversation, Siri.

Imagine a scenario in dog training, maybe teaching a puppy to sit stay. The instructor (if he is good, hopefully) will tell you: reward for sitting until released, ignore and put back to his place for breaking.

What an algorithmic training would look like if you follow exactly what your instructor says:

Puppy sits for 3 seconds, you release, give a cookie

Puppy sits for 2 seconds then breaks, you put him back

Puppy sits for 3 seconds, you give him a cookie

Puppy sits for 3 seconds, you give him a cookie

Puppy sits for 1 second, then breaks, you put him back

Puppy sits for 3 seconds, you release, then give him a cookie

Don’t you think that puppy feels like it is trained by a robot?

What a conversational training would look like:

Puppy sits for 3 seconds, you release, give a cookie

Puppy sits for 2 seconds, then breaks, you put him back

(At this point, you ask yourself: Why did he break? Was he distracted? Did I change my position? Was he less engaged? You pick an answer, modify this detail, and have him try again)

Puppy sits for 1 second, you release, big party and running around like crazy “You are a Superstar puppy! You are so freaking smart!”

Puppy sits for 2 seconds, release, have another party “I cannot believe you solved this problem so fast! There has never been a better sitting puppy than you!”

You ask your puppy for a dynamic trick or maybe just a game of tug and end the session.

Didn’t this puppy feel like it actually talked to a thinking being in the session? Who understood his questions and helped him answer them?

There is nothing wrong with starting out with algorithmic training. If you don’t have an idea of how to talk to your dog about a specific behavior yet, it is a good start to use the words someone else put into your mouth. Just don’t forget to switch to your own words eventually – this is when the real joy starts.