Full text not available from this repository.

Abstract

Researchers in Human-Computer Interaction typically rely on experimentsto assess the causal effects of experimental conditions on variables ofinterest. Although this classic approach can be very useful, it offers littlehelp in tackling questions of causality in the kind of data that are increasinglycommon in HCI – capturing user behavior ‘in the wild.’ To analyzesuch data, model-based regressions such as cross-lagged panel models orvector autoregressions can be used, but these require parametric assumptionsabout the structural form of effects among the variables. To overcomesome of the limitations associated with experiments and model-basedregressions, we adopt and extend ‘empirical dynamic modelling’ methodsfrom ecology that lend themselves to conceptualizing multiple users’ behavioras complex nonlinear dynamical systems. Extending a method knownas ‘convergent cross mapping’ or CCM, we show how to make causalinferences that do not rely on experimental manipulations or modelbasedregressions and, by virtue of being non-parametric, can accommodatedata emanating from complex nonlinear dynamical systems. By usingthis approach for multiple users, which we call ‘multiple convergent crossmapping’ or MCCM, researchers can achieve a better understanding of theinteractions between users and technology – by distinguishing causalityfrom correlation – in real-world settings.

Item Type:	Article
Authors/Creators:	van Berkel, N and Dennis, S and Zyphur, M and Li, J and Heathcote, A and Kostakos, V
Keywords:	ubicomp, mobile, performance, interaction, causality
Journal or Publication Title:	Human - Computer Interaction
Publisher:	Lawrence Erlbaum Assoc Inc
ISSN:	0737-0024
DOI / ID Number:	10.1080/07370024.2020.1715221
Copyright Information:	© 2020 Taylor & Francis Group, LLC
Related URLs:	Google Scholar: Heathcote, A UTAS Profile: Heathcote, A
Item Statistics:	View statistics for this item

Actions (login required)

Item Control Page