PsyBer-Agent: A Psychology-Driven User Agent for Recommender Systems

PsyBer-Agent is a psychology-driven user simulator for recommender systems that models evolving latent user states such as fatigue, affect, and shifting preferences through its Psy-Engine. The framework calibrates simulated behavior against real interaction logs using Unbalanced Gromov-Wasserstein optimal transport and introduces the WebSim environment and PsyBer Benchmark covering multimodal scenarios and seven recommender backbones.