In this section they focus on one of the hypothesis: how would the new day instance affect the forget parameter. They think that it is highly possible that students could be forgetting the previously learned knowledge when there are several days interval between the practices on the ITS. The model they used to test our hypotheses is a new model built based on the Split-KT model discussed in the previous section. By adding a time node to the Split-KT model they are able to easily specify which parameters of the model should be affected by a new day. The new day node is fixed with a prior probability of 0.2, which is the overall proportion of the new day instances in the dataset. The forget node is only conditioned on the added new time node, so there is only one new parameter “forget_n” introduced in this KT-Forget model and represents the forget rate on a new day. They use “forget_s” to denote the forget rate on a same day, which we set to be 0 just as the forget parameter in the original Knowledge Tracing model implying that there is no forgetting between opportunities in the same day. When a new day response occurs, New Day=T, the probability that student forget knowledge is forget_n, P(Forget=T|New Day=T) and is 0, otherwise.