Exploring the Role of Predictive Coding and Active Inference in Pain: A Bayesian Perspective

Abstract

In this dissertation, we have explored the role of Bayesian inference in the perception of pain, using Predictive Coding and Active Inference as theoretical frameworks. We have examined how the brain uses experience and sensory information to make predictions about potential pain experiences, and how this information is integrated to form a pain percept in a Bayesian Pain Model. By considering pain from a Bayesian perspective, we aimed to deepen our understanding of the underlying neural mechanisms of pain perception and to contribute to the development of new pain management strategies. Overall, this work highlights important elements that contribute to the understanding of the Bayesian Brain in pain.

Our empirical work demonstrates that the brain processes pain via Predictive Coding mechanisms, by utilizing recurrent transmissions of top-down expectations and bottom-up prediction errors, encoded in specific temporal and spectral patterns. The investigation of an affective-visual modality provides evidence that these oscillatory processes are reflected in modality-specific oscillatory patterns. Our data support the notion that a pain percept is generated in a Bayes-optimal manner, as explained by a Bayesian Pain Model. This is demonstrated in two experimental studies where we performed bi-directional modulations of the pain percept by placebo and nocebo expectations. The Bayesian Pain Model can explain contextual modulations via a mean shift in intensity expectations, for example, during a reduction in pain by a feeling of agency. This is in contradiction to ideas of Active Inference, which posits that there should be an attenuation of sensory precision by agency. This dissertation demonstrates the application of Bayesian principles in pain processing, framing the brain as a statistical machine that performs optimal inferences about the world.