Monday, April 2, 2012

Rate of Reinforcement and Tracking

I just finished writing an article for the The Total Rottweiler Magazine on luring vs reinforcing behaviors that we are trying to train. I have talked about this on the blog before - are you creating a behavior with praise/food/toy/etc or are you reinforcing the likelihood of the behavior from occurring by "paying" the dog after he has completed the behavior.While I can't say that I never use luring to teach a behavior I generally try to avoid it for a variety of reasons - handler and dog both get dependent on the lure and I believe that it prevents the dog from learning to value the work itself.  Carrot and stick approaches in humans have been shown to produce immediate increases in performance but a long term decrease in productivity. I think it is likely the same with dogs.

In scent work there is a place for what I would define as luring (others are free to disagree with me but only if we also agree on the definition of luring!). With our detection dogs we initially place the reward object with the scent. We do not hide them together but rather let the dog see the reward object in the scent box and then let him go to the box and get his fight at the source of the odor. This allows him to make a connection between the fight/reward and the target odor. Usually we only have to do this a couple of times before the lightbulb comes on. That is if the dog has the drives we desire to do the work. We then very quickly (sometimes in the first training session) switch over to reinforcing the behavior with the reward object. Of course we keep our criteria low at first, rewarding the dog the second it shows an interest in the target odor.

Tracking, although still scent work, is another challenge though. With detection work we have much more control and knowledge of where the scent is when we start the dog. While that may be true with simpler surfaces and less aged tracks, when the dog gets onto more challenging surfaces and/or the track is hours old we can no longer say for certain where the scent is. Like with detection work, initially I will lure the dog on the track using food and targets, the first time out I am not totally concerned if the dog is finding them visually. They are still getting the reward with the source of the odor. I will quickly fade my food and targets as lures though and make them become rewards/reinforcers for tracking. However I want to be careful with my rate of reinforcement - like with any behavior I need to start out at a high rate of reinforcement and decrease it as the dog builds skill and stamina.  The frequency of the food/targets is the rate of reinforcement for the track. In addition I also want to be certain that when my dog is challenged by something that he is rewarded for it once he solves the problem. If I place my reward too close to his challenge then it serves as a lure, placed past the challenge and it serves as an immediate reward for a good performance.

Here is an example of what I mean. Steel is doing a challenging hard surface track in a local shopping center (I am the heathen out tracking on Sunday morning when everyone else in the bible belt is at church...). While Steel is familiar with normal changes of cover, changes in elevation initially seemed to be confusing. He would work through them but only after he had double and triple checked and quadruple checked etc all of his other options for directions. I like this clip because he checks scent in either direction, determines that it doesn't go anywhere and then makes a decision without second guessing it. He commits to the parking lot and then there is reinforcement for making the right decision.  You will see him spot the food from a few feet away. Even without it being on a target it is impossible in this situation to keep the food from being visual. However I know it was not visual until after he committed to the track based on his body posture. Thereby making it a reward for the behavior rather than a creating the behavior. I am also careful to avoid allowing my leash tension to provide information. Steel is pretty hard pulling so he is not particularly sensitive to line handling but I am still careful not to use the line to prevent him from leaving the track. If I need to do that then I need to back up a bit in the degree of difficulty and create a more solid founation.

Steel did such nice work problem solving and learning on this track that I decided that we deserved something from the corner bakery (for me) and petco (for him). He got a duck and potato natural balance roll (which does not contain eggs, more on that in another post) and I got one of these:

200 calories of apricot and almond deliciousness. I did have to laugh at the lady behind the counter though - "Just one?" "Yes. Just one." "You only want one?" Just because I can eat more than one does not mean I should.... <roll eyes>

No comments:

Post a Comment