Concerns about the well-being costs of being "always on" are widespread, yet the scientific evidence remains surprisingly inconclusive, in part because most studies still rely on people's own, often inaccurate, estimates of their screen time. Passively logged digital trace data are increasingly presented as the solution. But do they deliver on that promise? In my dissertation, I examined three choices researchers make when turning raw digital traces into measures: how traces are theoretically interpreted, at which level of granularity behaviour is operationalized, and which devices are logged. This dissertation draws on a two-week citizen-science study where experience sampling was combined with smartphone and computer logging (N = 1,315).

The studies show that behavioural indicators correspond only weakly with subjective states such as online vigilance, that fine-grained features can reveal associations that aggregate screen time obscures, and that smartphone-only logging underestimates and misrepresents people's digital engagement. In this talk, I will argue that trace data are not a more objective substitute for self-reports, but capture a distinct, behavioural-structural level of description that must be integrated with subjective data to become psychologically meaningful. I will close by discussing how these lessons inform DIGIWORK, a 100-day study on digital work communication that links trace data to employees' daily experiences at and outside of work.

About the speaker:

Kyle Van Gaeveren is a researcher at DISCOLAB, the digital well-being and disconnection lab led by Prof. Mariek Vanden Abeele at imec-mict-UGent (Ghent University, Belgium), where he recently obtained his PhD in Communication Sciences. His research examines digital media use and well-being by combining passively logged smartphone and computer data with experience sampling methods. He has also helped develop tools for collecting screen time data beyond Android, including a data donation procedure for iOS devices. His research interests include digital trace data, measurement validity, intensive longitudinal and multilevel modelling, and reproducible research workflows.