Modeling action decisions in working memory-guided behavior

Abstract

When performing everyday tasks, we make use of our visual working memory (VWM) to temporarily store the information we see. In this study, we have designed the TCC-AT model which can predict naturalistic behaviour where VWM is used with only two parameters: memory strength (how well something is remembered) and action threshold (is the confidence in memory high enough to commit). We also designed and performed an experiment where people used VWM to remember colours and were tasked with reporting on those colours. For every memory task performed participants had to decide if they were confident enough in their memory to commit, either losing or gaining points depending on how well they remembered the colours. By making groups of participants remember a different number of objects at the same time, and making the strictness of the scoring different, the behaviour reflected different memory strengths and action thresholds in the TCC-AT model. The results showed that the model was consistent in predicting memory strength, but not in predicting action threshold. Multiple factors in the model and experiment were identified which could cause the problems in predicting action thresholds.