Cognitive capacity limits in information delivery

User interfaces are intended to deliver necessary information to users, to make tasks easier and optimise task performance. However, delivering excessive or intrusive information can tax users’ workload capacity, making tasks more difficult and onerous. This thesis investigates the effect of cognitive workload on user experience and performance by addressing three research questions, the first of which focuses on how workload capacity has been empirically measured in previous research. A systematic review identifies the various theoretical constructs and empirical measures of workload capacity used across a range of disciplines and identifies the detection response task (DRT) as an appropriate measure of workload in a human-computer interaction setting (Chapter 2). The second research question addresses ways in which workload measures can be applied to information delivery systems. In two novel studies, the DRT is then assessed alongside a computer-based tracking task (Chapters 3 and 4). The DRT is shown to be capable of estimating participant workload, but its presence is associated with decreased performance on the tracking task, indicating interference. Mathematical models are then applied to DRT and tracking task data (Chapters 5 and 6); the latter is modelled using a novel transformation of continuous data to facilitate the use of Systems Factorial Technology (SFT). This analysis shows high task load is associated with lower processing efficiency on both the DRT and tracking task. Mouse movement data is analysed using techniques previously used to assess workload with arm-reaching data sets (Chapter 7). Mouse movements are shown to start later and exhibit more submovements when task load is high, indicating these variables can be used to assess workload. The final research question addresses the extent to which a workload measure can be used as part of an adaptive system. The potential use of the DRT in this context is assessed, where the difficulty of a primary task is adjusted based on users’ cognitive state (Chapter 8). The DRT is shown to be sufficiently sensitive to drive an adaptive system, though it is no more successful than primary task performance data in adjusting difficulty. This research extends the applicability of methods from cognitive science, such as SFT, to human-computer interaction settings and contributes novel analytic techniques, such as the transformation of continuous data to response time-like distributions, to both domains. Novel applications of the DRT are also presented, extending its use in HCI and software development settings.