Valence Computation as Higher-order Inference via Conceptual Self-Processing

Predictive processing models suggest that emotions arise from the hierarchical computation of prediction errors across signals. However, this framework alone cannot account for the subjective, evaluative quality of emotional experience. Here, we extend predictive processing by integrating higher-order theories of emotion, proposing that emotional valence emerges from value judgments grounded in conceptual self-processing over sensory and contextual representations. This framework offers a mechanistic account of how subjective emotional experience arises through inferences about the dynamic interplay between world- and self-models within a shared computational architecture.