Monday, March 14, 2016

Sniffy's Schedules of Reinforcement and Extinction

March 14, 2016.

After magazine training Sniffy to associate the clicking sound with receiving food in the magazine and shaped her to associate a bar press with receiving food in the magazine, my goal was to put Sniffy on a schedule (or a couple) of reinforcement.

As you'll remember from my posts about Remy, the purpose of a reinforcement schedule is to solidify the behavior in the animal even if it is not being reinforced for every correct behavior. That is, Sniffy has to learn that she will not receive food every time she presses the bar. I wanted to do a couple of different ones for Sniffy with the goal of comparing the different types of schedules based on what we have learned in our lectures.

My procedure was to put Sniffy on a variable ratio schedule of reinforcement (VR5) and then a fixed interval schedule of reinforcement (FI30). I don't have to do much here except watch her press the bar and see how her patterns of behavior change. After allowing Sniffy to run on each of these for awhile, I put her on extinction.

Results:
The characteristic I looked for/expected on a variable ratio schedule was a steady rate of responses, theoretically because Sniffy would not know when to expect reinforcement with a given number of bar presses and would work hard at all times.

Cumulative record (in part) of Sniffy's VR5 schedule


For a fixed interval schedule, I would expect the cumulative record to look scalloped, indicating the increase of bar pressing near the end of the 30-second interval.

Cumulative record (in part) of Sniffy's FI30 schedule

Finally, I put Sniffy on extinction, which I'm sure the cyber rat would find quite disagreeable if she had feelings. She was reinforced for no bar presses, and it was expected that after an initial increase in the behavior, it would drop off dramatically. The process of extinction is exemplified and illustrated very well below.

Sniffy after ~30 minutes on extinction


Discussion:
Training Sniffy was a dream overall compared to training Remy. It took awhile to magazine train and shape her because the program has high thresholds for establishing that Sniffy has learned something (i.e., getting the bar operant associations bars close to their max levels) which does not really seem to be necessary in a live rat. On the other hand, all the variances of a live rat are not applicable to Sniffy, which makes the process seem shorter because it is less frustrating.

I spent approximately 45 minutes on each schedule of reinforcement and about 30 minutes on extinction. Magazine training took 45 minutes and shaping took even longer--an hour and a half. Throughout each of the different schedules, I didn't notice many oddities about her behavior except that on VR5 and FI30, her bar presses became more insistent and she spent more time near the bar even if she was doing something else such as grooming (similar to Remy on her schedules).

Sniffy the virtual rat is the ideal method to learn how the operant conditioning procedures and helped me a lot in the beginning stages of training Remy. I enjoyed getting to put her on several schedules of reinforcement because it gave me a chance to put the concepts from our class into practice.

Who knows--I may keep playing around with the Sniffy software after the class is over!

-Jessica

1 comment:

  1. Approximately how many responses did Sniffy make and how much time elapsed before he reached the extinction criterion in the VR-30 schedule? Approximately how many responses did Sniffy make and how much time elapsed before he reached the extinction criterion in the continuous reinforcement schedule?

    ReplyDelete