Representing Self-organization and Nonstationarities in Dyadic Interaction Processes Using Dynamic Systems Modeling Techniques

Dynamic systems modeling techniques provide a convenient platform for representing multidimensional and multidirectional change processes over time. Central to dynamic systems models is the notion that a system may show emergent properties that allow the system to self-organize into qualitatively distinct states through temporal fluctuations in selected key parameters of interest. Using computer vision-based measurement of smiling in one infant-mother dyad’s interactions during a face-to-face interaction, we illustrate the use of generalized additive modeling techniques to fit multivariate dynamic systems models with self-organizing, time-moderated dynamic parameters. We found evidence for systematic over-time changes in the infant → mother cross-regression effect, which provided a glimpse into how the dyad self-organized into distinct states over the course of the interaction, including periods where the mother’s positivity was reinforced and strengthened by the infant’s positivity, as well as periods where the mother’s positivity was inversely related to the infant’s past positivity levels.