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

Reinforcement Schedules

Techniques > Conditioning > Reinforcement Schedules

Description | Example | Discussion | See also

Description

A reinforcement schedule is a planned sequence of reinforcement activity with a goal of achieving a certain level of training.

A typical reinforcement schedule has the goal of reducing the need to reward a particular action, leading to the action being performed perfectly on cue, immediately, every time.

Once the basic action-reward pattern is established and the cue is attached, find a way to remove the reward. A very common method is to use randomness, not giving the reward every time, then reducing the chance of the reward being given over time.

Another method is to give the reward for a whole set of actions, for example chasing, catching, retrieving and then releasing, or for doing an action three times in a row. Long sequences of action can hence be trained by first rewarding for each action, then for action groups, then for the whole sequence (and then randomly for the whole sequence).

You can also use reinforcement schedules when working to eliminate an action. When doing so, make sure the reinforcement schedule is as short as possible.

Example

As a part of a whole approach that includes a deliberate reinforcement schedule parent gives a gold star on an award chart for a child who does not play truant, leaving school inappropriately. They also give hugs and play time. Before long, they child is attending regularly. In discussion with the child they agree to give the gold stars less often, though they do keep up positive rewards. Over time, the child voluntary says they now like school and don't need stars any more.

A dolphin trainer gives a fish for performing a trick. When the dolphin does the trick every time, the trainer starts giving fish randomly. Steadily, this is reduced to a maintenance level where the dolphin only needs an occasional fish to keep them going.

Discussion

To go from rewarded action, where the subject acts in order to gain a reward (or avoid a discomfort), to a conditioned action, where the action is performed on cue, there needs to be a period during which the reward is removed. Important in this is that the performance does not tail off.

When the subject becomes accustomed to reward for action, they expect it and may start to degrade their performance in order to find the minimum they need to do to get the reward. However, if the reward is given on a random basis, the subject becomes unable to predict whether it will be given. This can make them anxious about the reward. If they have sufficient hope, they will try harder than ever, including acting more quickly. It is hence important to sustain that hope. Too quick a decline in reward can lead to them giving up.

Shorter schedules are important when eliminating an action such as smoking as just one more cigarette acts to sustain the habit.

See also

Reinforcement, Variable Reinforcement