The question of how emotions are triggered remains obscure. A recent influential theory is built on predicting coding, which states that the brain makes predictions and constantly compares these predictions with the actual incoming sensory information. The active inference theory of emotions posits that emotions arise when there is a mismatch between the bodily (interoceptive) sensations and what was predicted by the brain, i.e. when interoceptive prediction errors occur. This theory is extremely relevant for emotional disorders linked with infections (e.g. post-COVID), since infections can trigger various and intense interoceptive signals that can be difficult to predict by the brain. In this project, we will apply the active inference theory of emotions in an ecologically relevant model of infection, using an innovative experimental approach. We will go beyond the existing correlational designs by applying a causal intervention eliciting real sickness symptoms, and manipulating interoceptive prediction errors directly, in 240 participants. We hypothesize that emotional responses will arise during sickness when interoceptive prediction errors occur, i.e. when interoceptive sensations violate predictions. This project will therefore provide critical information to understand how individuals' predictions shape sickness behavior, and how emotions are generated in infection-related conditions, which could open new therapies for mood disorders associated with immune activation.