This paper addresses the query scheduling problem for goal-oriented semantic communication in a pull-based state update system. We consider a system where multiple sensing agents (SAs) observe sources characterized by diverse attributes and, using the received information, provide updates to multiple actuation agents (AAs) that act to achieve heterogeneous goals at a final destination. A hub acts as an intermediary, queries the SAs for updates on observed attributes and maintains a knowledge base that is then broadcast to the AAs. The AAs utilize this knowledge to perform their tasks effectively. To quantify the semantic value of updates, we introduce the Grade of Effectiveness (GoE) metric. Furthermore, we integrate Cumulative Perspective Theory (CPT) into long-term effectiveness analysis to account for the system's risk perception and loss aversion. Using this framework, we compute an effect-aware scheduling policy that maximizes the expected discounted sum of the CPT-based total GoE provided by the transmitted updates while respecting a given query cost constraint. To achieve this, we propose a model-based solution based on dynamic programming and a model-free solution utilizing state-of-the-art deep reinforcement learning (DRL) algorithms. Our results show that effect-aware scheduling significantly improves the effectiveness of communication updates compared to benchmark scheduling methods, especially in settings with strict cost constraints where optimal query scheduling is crucial for system performance and overall effectiveness.
