Human-Robot Interaction (HRI) research is still largely shaped by the replication paradigm, in which social robots are designed as functional reproductions of human agents, based on computational and representational models of cognition [1, 2]. This perspective, however, struggles to account for the relational and embodied dimensions of interaction, often failing to support adaptability in dynamic, real-world contexts. This paper proposes a shift toward a dialogical model of HRI grounded in enactive cognitive science [3, 4]. Taking Learning from Demonstration (LfD) as a case study, I show how its current implementations often rely on Theory of Mind (ToM) assumptions and unidirectional models of imitation [5, 6]. An enactive reinterpretation can reframe LfD as a process of mutual adaptation through sensorimotor coordination and interactional dynamics [7] and provide a more comprehensive and dynamical view of human cognition which aligns with alternative effective and situated practices of LfD [8, 9]. Following this reconfiguration, novel ways of developing LfD will be suggested. This conceptual transformation aligns social robotics with broader goals of sustainability. Designing for dialogic co-creation and embodied reciprocity has the potential to enable the development of systems that are socially sustainable, i.e., adaptive and inclusive – able to integrate meaningfully into everyday life while promoting ecological inclusivity. By moving from replication to dialogue, this approach reimagines HRI as a domain in which novel forms of intersubjectivity emerge, essential for sustainable technological futures.