Robotic Phenomenology: Grammars of Behavior

Embodied cognition holds that behavior is shaped by the continuous coupling between internal bodily regulation and engagement with the environment. Robots likewise act under dynamically changing internal conditions; yet, such variables are typically hidden implementation details rather than seen as contributors to behavioral organization. We introduce a framework for robotic phenomenology to externalize how internal variables participate in shaping behavioral structure over time. Interoceptive and exteroceptive signals are jointly modeled as symbolic sequences, from which grammar-based representations of behavior are induced. These representations allow individual activity episodes to be summarized as structured motifs and interpreted using an LLM as first-person, phenomenology-like behavioral reports. We demonstrate the approach on an obstacle avoidance task, incorporating battery voltage change alongside environmental sensing. Rather than claiming that robots have individual experience, this work offers a tool for analyzing internal–external coupling as an organizational feature of embodied behavior.