Comparing models for modeling subjective and objective measures for two task types

Abstract

Adaptive automation (AA) has emerged as a viable solution to improving human performance in complex environments. However, understanding when to prompt, pause, and terminate AA remains unclear. Augmenting the user with physiological sensors offers new insight into the user's state, and thus, offers insight into when and how to implement AA. The research presented investigates the efficacy of prediction algorithms for modeling physiological and subjective data in AA environments. A comparison of traditional and emerging modeling methods results in recommendations for algorithm selection, generalizability, and risks of over fitting data are provided.