Exploring the dynamic interplay between Learning and Working Memory within various cognitive contexts

The intertwined relationship between reinforcement learning and working memory in the brain is a complex subject and has been researched widely across various domains in neuroscience. A main part of the research is focused on identifying brain areas responsible for these two functions, understanding their contributions in accomplishing the related tasks and examining how they adapt when faced with changed conditions such as a cognitive impairment or ageing. Numerous models have been introduced to formulate either these two subsystems of reinforcement learning and working memory separately or their combination and relationship in executing cognitive tasks. In this paper RLWM model is selected as the computational framework that models the behavioral parameters of subjects with different cognitive abilities due to age or cognitive status. In this sense, related RLWM task is used to test a group of subjects in different ages and different cognitive abilities based on their scores in the Montreal Cognitive Assessment tool (MoCA). The results show that the overall performance accuracy and speed decline as the age group (young vs. middle-aged) differs. They also differ significantly on some model parameters such as learning rate, WM decay and decision noise. Additionally, among middle-aged group, subjects that were categorized as normal vs. MCI based on MoCA test, differs in speed and performance accuracy as well as decision noise as one of the RLWM model’s parameter that was higher in MCI middle-aged subjects.