Educational robotics (ER) has emerged as a novel educational tool that enables students to improve their thinking skills. The study aims to compare the effect of a structured versus an unstructured ER curriculum on students' group metacognition during collaborative problem-solving with ER. The authors' hypothesis is that an unstructured ER curriculum might be more beneficial in supporting young learners' group metacognition in programming contexts. This study follows a quasi-experimental design with students (n = 35) split into two comparison groups -- a structured ER curriculum group and an unstructured one. The results show that students in the structured curriculum group demonstrated higher levels of group metacognition and better collaboration. Furthermore, using a micro-ecological approach, the study reveals that individual metacognitive contributions from students in the unstructured curriculum group had a systemic impact on the group work progress.