Understanding socially shared regulated learning (SSRL) in collaborative problem-solving (CPS) is vital for fostering effective group learning and improving educational outcomes. In CPS, team members regulate cognition, metacognition, emotion, and behavior together, with this study specifically targeting socially shared metacognitive regulation (SSMR). Conducted over a 12-week Human-Computer Interaction course with 15 participants, the study focused on automating the identification of SSMR triggers during CPS. Currently, identifying these triggers through verbal interactions is a manual and challenging process. In this study, the speaker diarization method was used to identify the speaker, success-fully detecting speakers in most cases. Following this, a machine learning (ML) model was employed to identify the triggers of SSMR episodes from the transcribed data. By synthesizing understanding from SSMR literature and conducting a preliminary data analysis, criteria were developed to identify triggers of SSMR episodes. The ML model identified 1,396 triggers by analyzing 21,256 turns taken by learners from four teams. To further validate the ML model’s findings, a manual verification process was conducted on a 25% sample of the identified triggers. Manual validation indicated a moderate level of reliability in the automated identification process. The findings suggest that the criteria for identifying triggers are important and that the automated approach has the potential to streamline the process of identifying triggers from teams’ verbal interactions. The automatic identification of SSMR triggers shows promise in reducing the manual data analysis workload in SSMR research.