This work contributes to research in robot-assisted learning (RAL) and robot-assisted language learning (RALL) for supporting the integration of refugee children and their families where existing studies do not consider specifically the social role of the robot in a learning task. Typically robots have assumed a traditional tutor role and directly taught social behaviors and skills to the learner. However, there is evidence that robots assuming a peer role can enhance learning in children. Does this result also apply to refugee children and their families, and are there other factors to consider? To answer this, as a first step, we conducted a system evaluation study with university students and found that a peer-learner robot could foster more meaningful learning than a tutor robot. Next, we interviewed professionals working with government-assisted refugee (GAR) families to understand their attitudes towards RALL for supporting refugee integration. The professionals interviewed think that robots can be a neutral medium to help resolve the dissonance in social norms and values refugees experience in everyday social scenarios. They preferred the peer-learner role, as it can demonstrate and serve as a practice partner for refugee children, and envisioned that several scenarios can be designed for the robot to enact with the refugee children to improve fluency in navigating real-life social scenarios in the host country. The findings inform this ongoing long-term research project, where the next step is to tailor the RAL system to refugees through further co-design and a user studies with this population.