Dynamical Motifs in Temporal Networks

In this paper, we explain the connection between information processing by complex systems and recurrent activity sequences in their dynamics. We argue that an understanding of information processing pathways in terms of these dynamical motifs is important for designing effective interventions. We then describe a recently completed study, where we video recorded 37 shared book reading (SBR) sessions, and thereafter annotated each of these sessions for 26 activities (reading the book, comments and questions, management talk by the teacher, and responses from the children). For all SBR sessions, the annotations consisted of sequences of one activity followed by another (transitions). We tested the empirical data against a null model where activities occur randomly, to identify 34 transitions that occur more frequently than by chance, and visualize these transitions that are statistically significant at the confidence level of p < 10−3 in the form of a static network. We then chose six significant transitions, and tested their extensions against the same null model to identify statistically significant length-3 sequences. This extension procedure was repeated to obtain length-4, length-5, and longer sequences until no further statistically significant extensions can be found. Finally, we organized the longest significant sequences into five families of dynamical motifs, and discuss their implications on the effectiveness of SBR.