In active inference, behaviour has explorative (epistemic) and exploitative (pragmatic) aspects that are sensitive to ambiguity and risk respectively, where epistemic (ambiguity-resolving) behaviour enables pragmatic (reward-seeking) behaviour and the subsequent emergence of habits.
Bayesian models of brain function such as active inference and predictive coding offer a general theoretical framework with which to explain several aspects of normal and disordered brain function. Of particular interest to a study is the potential for such models to explain the pathology of auditory phantom perception, i.e. tinnitus. To test this framework empirically, Hullfish et al., performed an fMRI experiment on a large clinical sample (n = 75) of the human chronic tinnitus population. The experiment features a within-subject design based on two experimental conditions: subjects were presented with sound stimuli matched to their tinnitus frequency (TF) as well as similar stimuli presented at a control frequency (CF). The responses elicited by these stimuli, as measured using both activity and functional connectivity, were then analyzed both within and between conditions. Given the Bayesian-brain framework, they hypothesized that TF stimuli will elicit greater activity and/or functional connectivity in areas related to the cognitive and emotional aspects of tinnitus, i.e. tinnitus-related distress. They conversely hypothesize that CF stimuli will elicit greater activity/connectivity in areas related to auditory perception and attention. They discuss this results in the context of this framework and suggest future directions for empirical testing 1).