IEEE RO-MAN 2026@Kitakyushu, Japan [Link]
Along with the development of AIs and LLMs, Human–robot/AI/agent interaction is becoming increasingly complex, enabling rich, adaptive, long-horizon exchanges. Therefore, although final psychological outcome measures such as trust, acceptance, or satisfaction are still valuable, they are often not sufficient on their own to capture what actually happened during an interaction: they may not reveal how coordination emerged, where misunderstandings accumulated, or why particular subjective experiences formed over time.
To better interpret these outcomes, it helps to focus on the cognitive processes that drive interaction−that is, the moment-to-moment mental updates behind behavior (e.g., how people infer intentions, update beliefs, allocate attention, and regulate emotion). In real settings, humans cannot directly observe a robot/AI system’s internal states, and robots/AI systems cannot directly observe human beliefs, emotions, or cognitive context; both sides also operate under limits in attention, computation, and communication. Under these constraints, interaction is a time-evolving, coupled inference-and-action process shaped by reciprocal actions, misunderstandings, and repair.
Advancing cognitive and computational accounts of these hidden dynamics is not only a scientific goal but a practical pathway for improving human–robot/AI/agent interaction by providing model-informed awareness of human cognitive/emotional dynamics and interaction context to an artifact, supporting more robust coordination and, ultimately, symbiosis in real settings.
This session welcomes broad contributions across HRI/HAI. We aim to foster discussion on how empirical phenomena in interaction can inform—and be informed by—principled computational and cognitive accounts, including mechanisms shaping subjectivity, context sensitivity, and bounded rationality. Relevant work spans empirical/behavioral research, user and field studies, design frameworks, system implementations, evaluation methodologies, and computational modeling/theory. Submissions that illuminate interaction dynamics over time and yield implications for theory-grounded models, metrics, and robot/AI interaction policies are especially encouraged (e.g., Bayesian Theory of Mind, information theory, active inference, pragmatics, interactive POMDPs). The session aims to foster constructive discussion across HRI communities and to bridge empirical findings with computational perspectives that inform both scientific understanding and interactive system design.