To learn the basics of magazine training and shaping, we used the computer program, Sniffy. First, I had to magazine train the virtual rat. The rat is magazine trained when there is an association between the sound of the magazine (where the food pellets are distributed from) and the distribution of food in the hopper. Bring the theories of learning to life using SNIFFY, THE VIRTUAL RAT! Instructors praise Sniffy, a realistic digital rat in an operant chamber (Skinner Box), for his ability to give students hands-on experience in setting up and conducting experiments that demonstrate the phenomena of classical and operant conditioning.
A couple of weeks ago, I had to learn how to train a virtual rat before I could begin making progress on/start training my live rat, Delilah. I magazine trained, shaped, and did a few reinforcement schedules on the virtual rat.
Magazine training involves teaching an animal to associate a neutral stimulus with a secondary stimulus. A secondary stimulus becomes a stimulus through the process of being paired with a primary stimulus. The neutral stimulus was the food magazine. The secondary stimulus was the sound that was made when the food magazine would deliver food.
The Sniffy software graphed Sniffy’s progress and produced cumulative records. Magazine training did not take very long, maybe 10-15 minutes. I had to press the lever to have food delivered and quite enjoyed watching Sniffy run across the operant chamber towards the hopper when he/she heard the noise.
Shaping took substantially longer to do than magazine training. I spent about 30 minutes to an hour trying to shape the virtual rat. Shaping is where the rat is actually taught to press the lever for food. Sniffy learned fairly quickly, for a computerized rat, that if he/she pressed the bar he/she would get food.
I do feel more prepared now after seeing an example of how magazine training and shaping are really done. It was good to experience this virtual world first before applying it to a real rat training situation. The training of the computerized rat was probably easier than trying to train a real rat. The computerized rat was very concentrated at the task at hand and I do not think most rats are like that.
I put Sniffy on a VR (variable ratio) 5 schedule. This means that Sniffy was rewarded after every five bar presses. I then went to VR 10 and then VR 20. Then, I played around with other reinforcement schedules. I put Sniffy on a FR (fixed ratio) schedule where she was rewarded every 50 times. I was just thankful there was a way to speed up Sniffy’s progress. That is something you cannot do with a real rat. I did notice that Sniffy was faster at learning the various reinforcement schedules compared to shaping. It was amusing to watch how Sniffy would learn the schedule. Sniffy would press the bar a couple of times and then give up/get frustrated and then would learn after a few tries how many times to push the bar before a reward of food would come.
February 17, 2016.As you may remember from my previous post, not only do I have a real rat named Remy for my Learning Psychology class, I also have a virtual rat named Sniffy (3.0)! Below are a couple of pictures of Sniffy eating from the food magazine (left) and rearing (right).
In order to prepare us for training our real rats, I and each of my classmates downloaded Sniffy.
My first task was to magazine train this little lady. Magazine training in the case of Sniffy is to train her to associatea clicking sound that is produced by me, the trainer, pressing the space bar with receiving food in the magazine (pictured above). This part is crucial to the training process because without this association, reinforcing Sniffy for behaviors I want would be meaningless.
After magazine training, my task was to train Sniffy to press the bar (silver; pictured above the food magazine). The bar dispenses food pellets without me having to do anything, but Sniffy is not aware of the association between the bar and thefood dispensing yet. That is what shaping will eventually accomplish.
Magazine training-
I chose to reinforce Sniffy whenever she came close to the food magazine. The only problem is that she just sort of ambles around the operant box, grooming herself, rearing, and doing other rat-like things. Only every once in awhile would she even look in the direction of the food magazine. In these moments, my timing of reinforcement was critical--the closer she was to the magazine, the better. Sometimes by the time I released the food pellet, Sniffy was already walking away and it would take a few more seconds before she even realized that there was food, and then she would amble over to the magazine again. These delays were frustrating.
Sniffy's curiosity and exploration continued throughout the magazine training task. Sometimes a breakthrough (at about the halfway point according to Operant Associations Mind Window on the computer program) would occur and Sniffy would linger by the the magazine after previously receiving a food pellet. I would take advantage of these time by rapidly releasing several more food pellets. This strengthened Sniffy's association of the clicking sound and receiving food from the magazine. The stronger this association was, the quicker Sniffy's response time to the magazine was, even if she was on the other side of the chamber.
It took about 30 minutes and way more than 30 food pellets to magazine train Sniffy, so I saved a file and returned to shaping another day.
Shaping-
In order to shape Sniffy to eventually press the lever, I had to reinforce her behaviors when she was facing the bar, standing near the bar, or rearing up on the bar. Because Sniffy had previously been magazine trained (above), she realized that she was doing behaviors I liked when she heard the clicking sound and was rewarded with food. The behaviors I liked, then, became more and more frequent until Sniffy began pressing the bar. When she pressed the bar, this behavior was automatically reinforced, as pressing it dispensed a food pellet.
Below is a video a few seconds long of Sniffy's behavior (in general) while first learning the result of pressing the bar:
Sniffy The Virtual Rat Macho
Then, Sniffy began to associate a bar presswithreceiving food. Once she figured this out, I just sat back and watched her continuously press the bar until the association was at the maximum level on the Operant Associations Mind Window. The whole process took about an hour and a half, but much of the time was spent just watching Sniffy press the bar over and over.
Sniffy post-shaping video:
Sniffy Pro 3.0
The cumulative record of Sniffy's shaping is below. As you can see, on a continuous reinforcement schedule (or FR1), the behavior is increasing over time because at EVERYbar press, Sniffy receives food.
Now that Sniffy has been shaped, I can put her on a schedule of reinforcement and eventually learn how to extinguish the bar pressing behavior. Stay tuned!
Sniffy The Virtual Rat Download
-Jessica