Multinomial processing models in visual cognitive effort diagnostics

Abstract

The pupillary response has been used to measure mental workload because of its sensitivity to stimuli and high resolution. The goal of this study was to diagnose the cognitive effort involved with a task that was presented visually. A multinomial processing tree (MPT) was used as an analytical tool in order to disentangle and predict separate cognitive processes, with the resulting output being a change in pupil diameter. This model was fitted to previous test data related to the pupillary response when presented a mental multiplication task. An MPT model describes observed response frequencies from a set of response categories. The parameter values of an MPT model are the probabilities of moving from latent state to the next. An EM algorithm was used to estimate the parameter values based on the response frequency of each category. This results in a parsimonious, causal model that facilitates in the understanding the pupillary response to cognitive load. This model eventually could be instrumental in bridging the gap between human vision and computer vision.