Stability of Individual Parameters in a Model of Optimal Fact Learning


We are using an algorithm based on a computational model of human memory to optimize the scheduling and repetition of individual items within a learning session. The model estimates the rate of forgetting for each participant to determine the order in which items should be repeated and to decide when previous items have been learned well enough to introduce a novel item. To improve the model further, we conducted an experiment to test how stable the parameter estimates are over time and across different materials. We have found that estimated rates of forgetting are stable over time within one type of material but not across different types of material. This finding has important implications for how information about a learner should be preserved between study sessions.