This paper explores how the active inference (AIF) framework can be applied to develop sustainable solutions within the field of social robotics, in the aim of offering a simple yet powerful application of AIF to support more responsible, transparent and sustainable social robotics. Building on previous work on a computational model where an agent learns to balance short-term needs with long-term resource availability, a possible extension is proposed: using AIF to guide the behavior of social robots towards a successful interaction with humans. This is done by outlining how this model could help robots manage “social resources” such as user attention, emotional well-being and trust. Hence, the potential benefits and limitations of this approach within human-robot interaction are explored.