Agency and Expectations in Pain Treatment: An Investigation of the Active Inference Model

Abstract

The fact that we cannot tickle ourselves is traditionally explained by the attenuation of somatosensation by predictions from a forward model of self-generated movements. Alternatively, it has been suggested within the framework of active inference that Bayes-optimal reduction of sensory precision can explain this phenomenon. Combining a pain paradigm with stimulus-related expectations allows to compare both models through predictions from the Bayesian account of expectation-based hypoalgesia, where pain is influenced by the precisions of somatosensation and expectation. In two experiments, heat pain was sham-treated either externally or by the subject, while a cue created higher or lower treatment expectations. Both experiments revealed greater pain relief under self-treatment and high treatment expectations. Electroencephalography revealed a modulation of theta-to-beta frequencies linked to agency and top-down modulations of pain perception. Computational modeling showed that this is better explained by an attenuation of somatosensation than a downregulation of somatosensation precision, favoring the forward model.