Many lab animals, and many person-hours, were recruited to help understand exactly how rewards and punishments could be arranged to influence behaviour. One surprising finding is that if you want to train an animal to do something, consistently rewarding that behaviour isn't the best way. The most effective training regime is one where you give the animal a reward only sometimes, and then only at random intervals. Animals trained like this, with what's called a 'variable interval reinforcement schedule', work harder for their rewards, and take longer to give up once all rewards for the behaviour is removed. There's a logic to this. Although we might know that we've stopped rewarding the animal, it has got used to performing the behaviour and not getting the reward. Because 'next time' might always be the occasion that produces the reward, there's never definite evidence that rewards have stopped altogether.
No comments:
Post a Comment