Abstract

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.