Humans collaborate with a large number of people in order to create and accomplish incredible feats. We argue that rich coordination dynamics underpin our capacity for collaborative creativity. These dynamics characterize the ways in which people are able to covary their thoughts, actions, behavior, etc. for functional purposes. We investigated the coordination dynamics of improvisation as a special case of collaborative creativity using two openly available data sets: a movement-based mirror game and jazz piano improvisation. By focusing on improvisation, the tasks elicit the need for real-time adaptation and mutual prediction based on information exchange between interacting individuals, with the creative ‘product’ being the behavioral performance itself. For each data set, we performed a transfer entropy analysis as well as an estimate of prediction decay. The combination of these two methods allows us to understand the dynamics as information-driven coordination flow and to differentiate unidirectional influence from mutual influence as well as the predictability of signals exhibited during collaborative creativity. We observed that for the mirror game, experts and novices exhibited unidirectional and bidirectional influence on each other’s movements largely independent of their improvisational experience level. Further, movement improvisation signals generated by experts were generally more predictable than those of novices. In terms of the jazz improvisation, our results showed evidence of bidirectional influence between the onset densities of coupled and one-way improvisational dyads, and the predictability of the signal did not vary systematically across these conditions. We discuss these findings in terms of differences between improvisational contexts, methodical challenges, and future directions.