Human-robot interaction (HRI) is fundamentally constrained by the fact that humans and social robots are mismatched conversational partners. While social robots often exhibit human-like expressive features, they lack the underlying mechanisms that support human social reasoning, resulting in mismatches that disrupt users’ perception, interpretation, and adaptation, increasing the cognitive effort required to interact effectively. This paper explores how biomimetic insights from social behaviours in nature can help address these mismatches. To analyse these mismatches, this paper introduces a conceptual framework, the ‘Theory of Lenses’ (ToL), which identifies misalignment types in signal transmission and interpretation across interpersonal and intrapersonal dimensions. Drawing from communication strategies in nature, we argue that one promising direction lies in the design of honest signals that reliably reflect a robot’s internal states and limitations. Future work will focus on empirical validation and operationalising this principle by developing a framework to design and evaluate honest signals in diverse HRI scenarios. Ultimately, our goal is to build social robots that are not only expressive but also socially intelligible and interactionally sustainable.