Abstract Socially shared regulation of learning (SSRL) is essential for the success of collaborative learning, yet learners often neglect needed regulation while facing challenges. In order to provide targeted support when needed, it is critical to identify the precise events that trigger regulation. Multimodal collaborative learning data may offer opportunities for this. This study aims to lay such a foundation by exploring the potential for using machine-learned models trained on multimodal data, including electrodermal activities (EDA), speech, and video, to detect the presence of SSRL-relevant process-level indicators in successful and less successful groups. The study involves thirty groups of secondary students (N=94) working collaboratively in five physics lessons. Considering the demonstrated positive results of machine-learned models, the advantages and limitations of the technical approach are discussed, and further development directions are suggested.