Multifractality as a lens on embodied human computer interaction

Abstract

This thesis investigates the application of multifractal analysis to user behaviour during interaction with technology.
Multifractal analysis is an analysis technique which quantifies features in behavioural signals which have been connected to embodied, control structures which emerge from interactions between processes across brain, body and context. It has been widely observed in human movement and behaviour, and variation in multifractal signatures has been associated with a range of flexible adaptive behaviours.

Through experiments and data analyses, I investigate the value of this approach for Embodied Interaction research, and HCI more broadly. I find the approach offers potential in operationalising key concepts in Embodied Interaction, which often rely on similar assumptions about the role of contextual interactions.
In particular, my results show that multifractal signatures in behaviour are informative about dimensions of ready-to-hand tool use — an influential construct in Embodied Interaction, drawn from phenomenological philosophy, which describes fluid, engaged tool use, in which the tool "disappears" for the user.

I also investigate the potential to apply multifractal analysis to understand behavioural constructs of wider interest in HCI, beyond Embodied Interaction. I find that the approach has significant potential as a convenient-to-deploy measure which can be used to infer behavioural constructs including fatigue, engagement and skill. To support future work, I develop a framework and make methodological contributions to which can guide future applications of multifractal analysis in HCI.

Date of Award	21 Mar 2023
Original language	English
Awarding Institution	University of Bristol
Supervisor	Oussama Metatla (Supervisor) & Anne Roudaut (Supervisor)