Thursday, March 17, 2016

Bribery vs. reinforcement

"I want my dog to listen to me, not a cookie."

"I don't want crumbs in my pocket all the time."

"Not every distraction we encounter will be less appealing than food."

"He should do it because he has to, not because he wants a slice of hot dog."

These are just some of the many objections people have to training with food. They all have one feature in common: they refer to bribery-based training, not reinforcement-based learning.

The difference is both simple and subtle. Bribery breeds the attitude that "if I do this for you, then you will pay me a treat" whereas reinforcement builds the attitude that "when I do this for you, good things happen". Bribery is about cajoling dogs into doing what we ask by showing them the money and holding our breaths while they weigh their options. Reinforcement of a known behaviour is about surprising the dog with a wide variety of rewards as a thank you for doing what we've asked. Reinforcement of a new behaviour normally involves feeding a treat for each and every repetition, but it differs from bribery in that treat presentation comes after the behaviour as early in the training process as possible. If a lure is used, it is quickly faded into a hand signal. Once the hand signal is reliable, then a verbal cue is introduced before the hand signal. Once the dog understands what the behaviour is and is reliably performing it on a verbal cue only, then the treats are faded into a variable reinforcement schedule (so the dog is now gambling) while other rewards are introduced.

So what exactly is a reward, and why specifically food? Anything the dog likes to have or do is fair game, and this is why food is such a powerful component of a reinforcement-based learning system. Eating is our first love - our first appetite, if you'd like to be corny about it. Puppies' and human infants' first behaviours all revolve around the acquisition and consumption of mom's milk. Our carnivorous companions in particular are hard-wired to be opportunistic about their food, working hard to satiate themselves before the next fast. In addition, food is incredibly convenient to dispense, allowing for rapid and numerous repetitions, especially if kibble is used. Fellow raw feeders, bear in mind that kibble now comes in many forms other than the traditional extruded type: baked, freeze-dried, refrigerated cooked, refrigerated raw, frozen raw, and air-dried. I simply use breakfast to teach new behaviours, and whatever is left over goes into an enrichment meal toy, such as the Busy Buddy Barnacle for dry kibble or the KONG Wobbler for raw meat nuggets. And lastly, the biggest reason to use food in training is that most of us feed our dogs twice a day. Those are two major opportunities for us to reinforce something about our dogs that we like, each and every day!

This leads us into the concept of simply taking advantage of anything a dog likes as an opportunity for reinforcement. Does your puppy like to play with other puppies? Then don't allow him to meet people or other dogs while he's straining toward them. Instead, ask them to wait while you get his attention. Prompt him - or wait for him - to make eye contact with you on a loose leash, then release him to play off leash  in a safe area as a powerful reward! Does your dog like car rides? Then ask him to walk on a loose leash to the car. Does your dog like to play with toys? Then surprise him with a rousing game of fetch as a reward for a rocket recall during a squirrel chase. Reinforcement-based parenting is about thinking of the world in terms of opportunities for reinforcement rather than distractions.

Food is indeed invaluable in efficiently training new behaviours, and lure-and-reward in particular is the most basic technique that everyone thinks of in conjunction with positive reinforcement. Bribery plucks out nothing but the luring part, a mere blip in the reinforcement learning mosaic. Everyone uses food - you know, to keep our dogs alive. Most people just use it to make their dogs fat, many use it to bribe their dogs, and some actually use it to reinforce good choices and trained behaviours. Now that the distinctions are clear, which will you choose?

No comments:

Post a Comment

About the Harenbergs

Audrey, Aipa, and Steve Harenberg on the Blue Ridge Mountains

We build strong working relationships between dogs and handlers through the use of modern learning theory and the development of technological aids.

Audrey has been involved with numerous non-profit organisations, as foster mom, therapy dog handler, and obedience instructor. In addition to teaching private lessons, group classes, and in-home training, she's busy creating free blog articles and books to help make progressive dog training accessible to all.

Steve is finishing his PhD in computer science and will be taking over technological development once he is freed from academia.