How we change what others think, feel, believe and do

Shaping

Techniques > Conditioning > Shaping

Description | Example | Discussion | See also

Description

Shaping is the process of incremental change towards a desired action, quite possibly after a given cue. The general principle is to reward even the smallest movement in the right direction. The complete action is hence build up one small step at a time.

Shaping is particularly useful for complex actions or where the subject is not particularly willing or able to perform the desired action.

The basic principle of shaping is that a perfect response is not required. All the trainer needs to do is to reward the closest response. This generally needs time and patience, but a lot can be achieved by this, even with an unwilling or unintelligent subject.

Where reasonable, shaping may be speeded up by using other methods that help lead and guide the subject in the right direction.

Other tips for shaping include:

• Focus on one action at a time. Condition it so it is repeatable, then move on.
• Make sure each step along the way is achievable by the subject. Do not try to move in big steps that may be unattainable.
• Consistently use the minimum reward that works. Avoid any punishment.
• Be consistent in how you act, speak and reward. Be comfortably predictable, not confusing nor scary.
• Use appropriate reinforcement schedules, including fading out of rewards for learned actions.
• Move as fast as the subject learns. If they are ready, move more quickly. If they seem stuck, slow down.
• Know what you are doing. Always be ready for whatever happens.
• If they 'forget' or do not complete the whole sequence learned so far, step back as far as needed.
• If they forget a just-learned routine when you move on to the next one, give them time to recover before stepping back.
• Use what works. If your methods for shaping do not working with a given subject, try something different.
• Keep the same trainer for the duration, rather than confusing the subject with a change of trainer.
• Avoid making sessions so long they get bored or over-anxious. End on a correct response, with a big, happy fuss.

Remember that what seems easy and obvious to you can be very hard for your subject. This is true of the difference between any teacher and their pupils. Try to get into their heads and understand just how they are not understanding. Likewise, remember how things you think you have learned can slip away and need more time and practice to be made instinctive.

Example

A dog trainer wants their subject come when called. Initially they reward just turning the head. Then they reward a movement in the right direction. Gradually they increase the distance the dog has to travel. They also work on sitting to attention when the dog comes as called.

An art teacher starts with getting pupils just wash blue paint consistently down a page to create a sky, praising improvements in this. Then they gradually add more complex elements. They end each lesson with a motivational speech.

Discussion

It can be difficult to learn a complex action all in one go, sometimes even for humans. It can also be difficult to train an unwilling subject who can resist when they think they are being forced against their will. Shaping addresses these by the principle of 'softly, softly catch the monkey', gradually nudging the subject in the right direction. A not dissimilar method is seen in fishing, where a larger fish is allowed to pull away a bit, but is still reeled in, a bit at a time.

One of the benefits of shaping is that, while you are paying attention to each tiny action, your subject picks this up and also sustains attention when they might otherwise become bored.

Shaping uses the fact that how subjects behave is variable. Unless they have already been conditioned, dog and a human behave differently in the same circumstances. All the shaping does is to catch the actions that are the best match available to what is required, even if this is initially a long way from the desired action.

See also

Reinforcement Schedules, Variable Reinforcement, Fading