Errorless Learning

Notes from the Errorless Learning session by Dr. Susan Freidman

• Errors punish effort (Paul Chance, 2003) They slow down learning and generate undesirable emotions (frustration, aggression, giving up)
• Prompts are information, not bribes
• A reward may or may not be a reinforcer (word use is important) It is understood in +R language that “reward” means reinforcer, but it’s not an accurate word. A Christmas bonus is a reward, but it doesn’t reinforce a specific behavior.
• “Errors are not necessary for learning to occur. Errors are a function of poor analysis of behavior, a poorly designed shaping program, moving too fast from step to step in the program, and the lack of the prerequisite behavior necessary for success in the program” (B.F. Skinner, 1968)
• To set up errorless learning:
  • Make the wrong choice more difficult (small, farther away, visually faded, etc.) and make the right choice as likely as possible.
• In testing example: In the errorless learning set-up, the animal made 25 errors out of 5000 attempts. In the trial and error set up, the animal made 3,000 errors out of 5,000 attempts. The trial and error sessions had many more stress and frustration behaviors in the learners.
• Shows that when an animal makes a mistake, the error is in the set-up, not the animal.
• It is deliberate and thoughtful redesign of the environment for the purpose of getting the learner to do the correct behavior (earn reinforcement) more quickly and without frustration. Can be achieved while working them toward a more complex goal.
• Reducing errors is the TEACHER’S responsibility.
• For behavior change/reduction (problem behaviors) you MUST look at the ABC’s.
• The source of behavior is always (at least in part) the environment (external stimulus or reinforcer) A or C of the ABC’s
• Know what signals, precedes and motivates and you’ll find what to change in order to change the behavior.
• When you don’t use prompts and lures, you are going back to Trial and Error learning. You still want to fade them quickly, but they are the best way to GET the correct, markable behavior.
• Free shaping only works with a highly skilled trainer AND highly skilled animal.
• If pressure is used, it must be information only and not aversive (which would impede learning). Always start with least possible amount and work up if needed.
• When an animal doesn’t do a well-known behavior, pause, then cue some easy and very likely to occur, well known behaviors that you can reward before asking for the original behavior again.