Exploration of the Rate of Forgetting as a Domain-Specific Individual Differences Measure

Abstract

Learners differ in their learning aptitude. Modern computerized fact-learning systems take these individual differences into account by adapting repetition schedules to the learner's characteristics. Adaptation is based on monitoring responses during learning and using these responses to inform the model's decisions about when to introduce and repeat material by updating the model's internal parameters. Typically, adaptive systems start a learning session with a set of default parameters, with these parameters being updated and adapted to the learner's characteristics when responses are collected. Here we explore whether domain-general individual differences such as working-memory capacity or measures of general intelligence, which can be assessed prior to learning sessions, can inform the choice of initial model parameters. Such an approach is viable if the domain-general individual differences are related to the model parameters estimated during learning. In the current study, we asked participants to learn factual information, and assessed whether their learning performance, operationalized as (1) a model-parameter that captures the rate of forgetting, and (2) the results on an immediate and delayed post-test, was related to two common measures of individual differences: working memory capacity (WMC) and general cognitive ability (GCA). We failed to find evidence in favor for such relations, suggesting that, at least in this relatively small and homogeneous sample, executive functioning and attentional control did not play important roles in predicting delayed recall. The model parameters estimated during learning, on the other hand, are highly correlated with delayed recall of the studied material.