Abstract
Digital phenotyping uses data from smartphones and wearables to extract behavioural and biosocial markers of psychopathology in situ. Traditional entropy-based measures capture static system properties that neglect temporal dependencies critical to psychiatric phenomena. We propose a "dynamic" approach to the modelling of digital data capturing the time-varying aspects of processes of mental disorders. We defend that the resulting dynamic digital markers better capture variability in regulatory mechanisms of psychopathology.