Day Two Hundred and Twenty Five!!!!!!!

Prediction Error


Learning is actually based on something called a prediction error. The theory behind the prediction error is; if you already know what you need to know then there is no point in learning something new. So, if you have already predicted that the food is coming, then there is no real need to change anything. On the other hand, if the food shows up when you weren’t predicting it  to, meaning that you have made an error in your prediction, then this is a set up in order to learn something new, in order to be able to make better predictions in the future. 


So, if you make an accurate prediction (just like the patients in the hospital Dr. Laurie had talked about), when the patients could hear the clinking of the dishes down the hall, and then the person walks in the door with the tray if food, there isn’t a prediction error. In fact their prediction has been strongly associated because the prediction was proven to be true as it had happened over and over and over again. The prediction is secure and there is no need for additional learning. 


So, if we look at this in terms of other cues like watching TV, or looking at the time, or driving by a fast food restaurant, and you reinforce these predictions with the reward of getting food then there is no prediction error because your prediction is correct. There is no need to learn anything new. However, lets just say you go to the fast food restaurant and they don’t have your favourite drink or meal because there is either a broken machine or it was a seasonal food offering and now no longer being offered, then there was a prediction error. This is a prediction error, as it means that you don’t end up with the expected victory but feeling cheated by a surprise defeat, then this is a prediction error and learning will therefore occur. This theory is called the Rescorla - Wagner model.


To set this theory in practice in order to be successful in weight loss, it would mean that you would have to do the opposite of supporting the reward by; not stopping in to get the food, not rewarding yourself with food when the time is right, not eating when you sit down before the TV. The reward would have to change from the food being the reward to something else being the reward, it would have to be the goal of weight loss. 


Learning from prediction error is based on a sequence of multiple events proceeding a reward. 


It is about figuring out how to behave in order to maximize the attainment of long term reward. Learning from prediction error requires a lot of sequence of actions that slowly moves you closer to the goal. In other words, the goal would have to change, the reward would have to become the attainment of the weight you want to be. It would have to change from the expected victory of the reward being fulfilled, resulting in eating. Learning by prediction error, the non attainment of the food would have to coincide with the reward of the weight you want to be. 


So, if you were to sit down and outline your prediction in which food was your reward and then change the goal to attaining the weight you want to be. Doing your usual routine without eating when predicted, will allow you to start a new path of learning. 

​