Optimizing Performance through Stress and Induction Levels in Virtual Reality Using Autonomic Responses

Abstract

Virtual reality (VR) is now a consumer technology, but individuals' experience with systems or different applications varies enormously. This means that most consumer applications need to consider how to train naive users in the application's operation. We examine three different ways of imparting on-boarding instructions to users: first-person audio guidance, second-person non-player character (NPC) diegetic guidance or written instruction. Our primary hypothesis is that the second-person condition will induce a higher stress level on the user, given the perceived presence of a supervising NPC. Our secondary hypothesis is that there is a correlation between stress and performance, meaning that participants with elevated stress levels within a certain margin will complete their tasks faster and more successfully. By extension, participants whose stress levels are either above or below this optimal margin will under-perform on the same tasks. The tasks in question are an interaction test (IT), designed to test participants' abilities to pick up and manipulate virtual objects, and a mental rotation test (MRT), designed to place them under cognitive load. During these tasks we measure the users' level of stress from their bio signals via a mobile wearable device that tracks their heart rate (HR), galvanic skin response (GSR) and body temperature in real-time. Statistical significance was not found in the stress or performance levels between the instruction conditions, but the secondary hypothesis was supported and a correlation was found between stress and performance levels across the conditions in both HR and GSR.