A Model of Human-Automation Team Performance: Incorporating the Influence of Concurrent Task Workload on Automation Failure Response in Ground Vehicles

Abstract

We first present a model of the influence of four properties of automation on human-automation team performance, dichotomized between conditions when automation functions as designed and when automation fails. These properties are reliability, transparency, degree of automation, and concurrent task workload during automation use. Mediating cognitive variables between properties and performance are trust, dependence, situation awareness, and mental workload. The strength of these influences is estimated from effect sizes in meta-analyses. We then present a meta-analysis of the effect of concurrent task workload on automation failure response as a function of non-driving related task (NDRT) workload within the context of driver take-over request time. Results from 11 studies revealed no overall penalties (mean effect size) for workload on failure response but found considerable variability across studies. Possible moderator variables, including task engagement are presented. These results are interpreted in terms of a model of optimal resource allocation between monitoring automation and the NDRT.