Queueing models serve as effective proxies for the allocation of socially scarce resources, where customers sequentially receive limited services. To address this, service providers aiming to maximize social welfare can use quality-related signals to deter low-priority customers from queueing, thereby reserving access for high-priority counterparts. We conduct a behavioral economics experiment between the service provider (the environment) and customers (AI agents, or “suspects”). Our proposed framework integrates human-algorithm interaction in a transparent “white-box” setting. To explain the experimental results, we construct a theoretical model of information design to regulate scarce services allocation. Our research emphasizes the interaction between AI decision-making and its environment, highlighting how AI can overcome behavioral inefficiencies (e.g., selfish queueing) to foster cooperation and improve welfare.