Streaming entropy for dynamic modelling in psychological measurement

The contribution introduces an entropy-based method to explore the dynamic evolution in psychological measurements. The streaming entropy (Stre-GCE) approach is described and tested on a relevant clinical scenario, that is, the psychotherapy process, to investigate the dynamic behaviour formalised by simultaneous equations describing the mutual interactions between patient and therapist. Textual data are obtained from psychotherapy transcripts, whose word blocks have been processed and classified according to their positive, negative, or abstract valence. The Stre-GCE algorithm computes model parameters for each interaction component, then parameters get updated, describing the effect of the therapist on the patient and vice versa. The results show that Stre-GCE is able to take into account the modulating nature of the psychotherapy interaction.