What is micro-shaping?
The first time I heard the term was at Clicker Expo when Kay gave a presentation on shaping where she talked about the importance of clicking for very small movements that would lead directly to the final behavior, and also on the importance of having a high success rate. I can remember being in a learning lab and taking data on the behavior of a golden retriever who was supposed to be orienting to a cone, but was staring intently at his owner’s feed pouch instead. His success rate was very low!
Kay started with some general rules for micro-shaping: (Note: later she said these are good rules for any kind of learning)
Setting up for success:
- 95-100% of behaviors are successful
- Clear base/opening behavior
- Use different reinforcement patterns
- Repetition to build up myelin
- Errors are at the micro-level
- Puzzle moments to check what the dog has learned
Micro-shaping is self directed learning where the dog is both given choice and provided with limited choices at the same time. Kay says that the boundary between choice and no-choice is where skill is built. This means setting up your training so that the dog learns the skills it needs in an environment that supports the right choices, but also provides enough direction that the dog doesn’t end up making wrong choices. I think I know what she means but I might have to think on this one a bit….my guess is that it’s a fine line.
Let’s look at the individual components of micro-shaping:
95-100% behaviors are successful:
- Not using extinction to attain approximations (that would not be fair)
- Process builds self-confidence and motivation in the learner
- Learner trusts and is confident when working with the teacher
- Learned trusts and is confident in the training process
Clear Base/Opening behavior:
- The base position is the starting position for the behavior (stand, sit, down, trot, etc…) It should be clearly defined and well known by the dog (begin with success).
- The base position should be refreshed intermittently (reinforced at various times throughout the session).
- Return to the base position if there is a loss of confidence.
- You can think of the opening position as the first step building in building a new behavior from the base position, think of extending or adapting from base position.
- Here’s an example of a base position/opening position combination. If you wanted to train a paw lift, you would start with a stationary base position like sit, stand or down. The opening behavior would be a change in balance in preparation to lift a paw.
- Different reinforcement patterns can be used to build flexibility (not variability) in a behavior.
- These patterns can be use the same delivery technique but vary the location (throw or place in different directions) or include variations in other qualities like pace, origin, distance, etc…
- Varying reinforcement patterns build in adaptability – can you easily land on your platform from here? How about from over here?
- Returning to the same location after reinforcement helps build a visual map of the behavior. Can you find the same position if approaching from different angles?
- It can be helpful to think of this as taking a skill and making it flexible. What are all the ways in which a behavior might need to be adjusted under different conditions?
She had some video of teaching her dogs to go around a cone. This behavior is shaped by starting with the dog (Merrick) on a platform. Kay is sitting in a chair with the platform in front of her and she has a cone on her left, near the platform. Kay is clicking Merrick for getting on the platform (facing her) and she is delivering the food by tossing it so that Merrick leaves the platform to get her reinforcement. Kay throws the treats in different directions so that Merrick learns to approach the platform from many different angles, at different speeds, etc…
Once she can easily return to the platform from any direction, Kay starts to introduce the cone in a more significant way. She sets up her throws so that Merrick is going to the left of the platform and moves the cone forward so that Merrick is more aware of passing by the cone as she returns to the platform. This is the beginning of shaping the dog to keep the cone on her right.
The video showed a lot of repetition. The dog comes to the platform, gets clicked and runs off to get her reinforcement. But even though there is repetition, it is not mindless repeating of the same behavior because of all the little variations that Kay weaves into the process. This kind of repetition is what builds myelin and myelin is what is responsible for learning to do a skill well.
Kay shared some information from Daniel Coyle’s book “The Talent Code.”
- Every movement, thought, feeling is a precisely timed signal that travels through a chain of neurons
- Myelin is the insulation that wraps around nerve fibers
- The more we fire a circuit, the more myelin there is, and the more it is optimized
- Repetitions are needed to build myelin, but these need to be repetitions of correct behavior. It is easier to learn to do a behavior correctly if you slow things down. Doing things at speed covers up errors. “It’s not how fast you can do it, it’s how slow you can do it correctly.” – Tom Martinez.
Back to the list of components of Micro-shaping:
Errors at the micro-level:
- Allow learner to fully explore and make changes at the micro-level
- Allow learner to find its own solutions and answers that improve its skills.
- Kay said that “micro-shaping is about building the skill of finding micro-solutions.” Let small errors lead to a greater understanding of the behavior.
- The more practiced you are at finding solutions, the faster the signals travel
- Errors are information to the teacher. Many errors or large errors mean the teacher needs to slow down and make finer slices.
- These are moments in which we seek information about what the dog has learned. They are not about testing the dog, but about testing the training.
- Can be done by changing one element of the exercise and observing the dog’s behavior. She had some nice examples of this. She asked one of her border collies to back up when it was standing with its hind legs up against a wall. The dog marched in place. She had one of her Gordon setters to back up when it was lying down and it scooted backwards without getting up. To the border collie, back up meant “move your feet,” to the Gordon setter it meant “increase distance from Kay.”
- Someone asked if the puzzle moment was like proofing. Well, sort of… in that one type of puzzle moment is when you cue the dog to do a behavior under conditions that are slightly different than normal. But a puzzle moment can also be a point in training where the dog has to make a choice or show you what it has learned. In the example of teaching the dog to go around the cone, Kay would continue the training by delivering most of her reinforcers to the left of the platform (still tossing them). The dog gets used to going to Kay’s left, and passing by the cone on its right (the dog is going to the left of the cone), as it returns to the platform. Once this is going well, she starts moving the cone a bit farther out a little bit at a time until there is a gap where the dog could go to the right of the cone to reach the platform. She is setting up a puzzle moment where the dog has to choose between passing the cone on the right (heavily reinforced pattern) or on the left (shorter route). The dog’s decision tells her if the dog has learned to keep the cone on its right or not.
- If the puzzle moment shows a lack of understanding of the behavior, the teacher can provide some guidance in the form of a hint or limiting the choices. In the example with the dog learning to go around the cone, Kay can direct the dog in the correct direction by gently blocking access to the opening between the cone and the platform with her foot or hand. That subtle block shows the dog that the short route is not an option and the dog has to find the solution which is to go back to the more heavily reinforced behavior of going to the platform with the cone on its right side. This reminds me of a talk I attended about backchaining where the trainer said that it is normal for the dog to test and see if they can skip elements of the chain. If you let them do this and they learn that skipping behaviors does not work, you will end up with a stronger chain. If you never let them experiment, then they will probably test it at some point when you don’t expect it and you may not be in a position to ensure there is no reinforcement. It’s better to set it up so the dog experiments a little bit when you are prepared to handle it.
- When you set up a puzzle moment, you should always have a plan for how to handle the different possibilities that can occur.
She finished up with a reminder that micro-shaping is about teaching skills and teaching dogs to find micro-solutions.
There were some questions from the audience about other training strategies and puzzle moments.
Her answer was that you can take the same ideas and apply them to other training strategies like luring, targeting, prompting and moulding. There would be differences in application and timing, but you would still be looking for a high rate of success. She uses the 6 points from her “Set up for Success” in her training, regardless of how she is teaching the behavior.
For example in luring, the puzzle moment would be when you ask for the behavior without the lure. One problem with luring is that the trainer often thinks the animal has learned the behavior before the animal has actually learned it. So learning when to set up the puzzle moment is an important skill and different training strategies would require different puzzle moments. Ideally you set up a puzzle moment when you think the animal understands it and the puzzle moment builds confidence.
Thank you to Kay Laurence for permission to post these notes. If you would like to learn more about Kay, her website is http://www.learningaboutdogs.com.