Goal-Oriented Semantic Resource Allocation with Cumulative Prospect Theoretic Agents
Traditional Expected Utility Theory (EUT) fails to capture human subjective perceptions—such as context-dependence and risk sensitivity—in goal-oriented semantic networks, leading to suboptimal resource allocation. Method: This paper pioneers the systematic integration of Cumulative Prospect Theory (CPT) into semantic resource allocation, modeling agents’ loss aversion, probability weighting distortion, and reference-point dependence. We formulate a non-expected utility optimization model grounded in CPT preferences and redesign multi-channel wireless power allocation policies accordingly. Results: In realistic semantic communication scenarios, our approach improves task completion rate by 32% and perceptual quality consistency by 27% over EUT-based baselines, significantly enhancing robustness against human cognitive biases. The core contribution lies in exposing EUT’s fundamental limitations in human-centered networks and establishing the first CPT-driven semantic resource allocation paradigm.