Exploring the Time-Based Resource-Sharing Model of Working Memory Through Computational Modeling

Working memory is the fundamental component of cognition that allows us to temporarily maintain information needed for concurrent processing. An existing theory from the literature, the time-based resource-sharing (TBRS) model, posits that working memory is a serial, rapidly switching, attentional refreshing mechanism. While others have sought previously to formalize the TBRS model into a computational process model, I go further, using ACT-R to model the influence of working memory on an entire task from end to end. I leverage ACT-R\u27s existing base-level learning mechanism, typically used to model recency and frequency effects in long-term memory, to enact the attentional refreshing and temporal decay central to TBRS. I also use a novel combination of existing inhibition and association theories to implement a functional list representation. The model replicates trends in human memory spans and response times across six experimental conditions from a previously published study. These efforts reveal that areas not traditionally associated with working memory research directly, particularly item representation and response strategy, are necessary assumptions of any such process model despite being underspecified in TBRS and other theories. I discuss future experiments to further constrain these ancillary assumptions and conclude by proposing various directions for expanding the model in subsequent work