Combodied Agents: A New Paradigm for Human-Centric AI
A recent study presents 'Combodied Agents,' a human-focused framework for Agentic AI that fills a notable gap in existing AI technologies. The research, identified as arXiv:2608.10915, contends that while Digital Agents alter software states, Embodied Agents change physical states, neither prioritizes the evolving state and agency of individuals for modeling, intervention, and assessment. For instance, if an elderly person misses a medication dose, a software agent may issue a reminder, while an embodied agent can deliver the medication, but neither clarifies whether the individual forgot, is confused, experiencing side effects, or chose not to take it, nor what assistance would be suitable. Combodied Agents track, model, and support personal human-state trajectories over time, utilizing software, sensors, wearables, robots, and human services as means to an end rather than ultimate objectives. This paradigm integrates diverse functionalities from personal assistants, health agents, AI companions, and adaptive human-AI systems into a cohesive loop, grounded in event-based modeling. The paper was recently submitted to arXiv as version 2608.10915v2, marking a significant move towards more human-centric AI, highlighting the necessity of understanding human states and agency in AI interactions.
Key facts
- Paper arXiv:2608.10915v2 introduces Combodied Agents, a human-centered paradigm for Agentic AI.
- Combodied Agents address a structural gap: Digital Agents transform software states, Embodied Agents transform physical states, but neither models human state and agency.
- The paradigm perceives, models, predicts, and supports individual human-state trajectories over time.
- It uses software tools, sensors, wearables, robots, and human services as action channels.
- It unifies capabilities across personal assistants, health agents, AI companions, and adaptive human-AI systems.
- The example of an older adult missing a medication dose illustrates the limitations of current agents.
- The paper was announced as a new submission on arXiv.
- The approach is event-based, forming a closed loop.
Entities
Institutions
- arXiv