Prosociality by Coupling, Not Mere Observation: Homeostatic Sharing in an Inspectable Recurrent Artificial Life Agent
arXiv:2604.10760v2 Announce Type: replace-cross Abstract: Artificial agents can be made to ``help'' through explicit social rewards, hard-coded prosocial bonuses, or direct access to another agent's state. I isolate a narrower route: homeostatic coupling. Building on ReCoN-Ipsundrum, I add a scalar homeostat and a social coupling channel while keeping action selection self-directed: the planner scores only the actor's predicted internal state, with no partner-welfare reward. In a one-step FoodShareToy, an exact solver finds a switch from EAT to PASS at $\lambda^\star \approx 0.91$ for the defa
This research provides a foundational step in understanding and engineering prosocial behavior in AI, moving beyond explicit rewards or direct state access.
A strategic reader should care because designing AI agents with inherent prosociality through homeostatic coupling could lead to more robust and less adversarial multi-agent systems.
The approach to instilling cooperative behavior in AI agents shifts from external incentives or direct control to an internal, homeostatic mechanism.
- · AI ethicists
- · Multi-agent system developers
- · Autonomous system manufacturers
- · Developers relying solely on external reward functions
AI agents begin to demonstrate emergent cooperative behaviors without being explicitly programmed for it.
This enables the deployment of AI systems in complex social environments with greater trust and less need for constant human oversight.
Societies develop hybrid human-AI social structures where AI agents contribute to collective well-being through intrinsic mechanisms rather than imposed rules.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.AI