Learning Uncertainty: Predictive Coding in Anxiety Formation

Generalized Anxiety Disorder (GAD) is among the world’s most prevalent psychiatric disorders. Affecting an eighth of the world’s population, it often manifests as persistent apprehension which is difficult to control. Despite its prevalence, neuroscientific efforts to understand the cognitive mechanisms of GAD formation remain sparse. This has resulted in a fractured theoretical landscape, lacking a unitary framework. While prior theories of anxiety describe the cognitive, affective and behavioral dimensions of anxiety, a unified theory outlining the formation of GAD remains elusive. Here, we point out that postulates derived from the Free Energy Principle (FEP) and predicting coding may allow for a unified theory to emerge. After outlining the FEP and its suppositions, we review extant work which applies a Free Energy framework to a range of psychiatric disorders. Subsequently, we outline existing theoretical approaches to Generalized Anxiety Disorder (GAD), and how they describe its etiology, cognitive architecture, and cognitive machinery. We end with an extension of existing efforts to understand GAD formation, rather than symptomology. We conclude that an approach toward GAD formation may benefit from a focus on predictive modelling, and may afford opportunities to re-conceptualize anxiety within the framework of working generative models, rather than static beliefs. In sum, we suggest that a biological system—having had persistent uncertainty in its past—will iteratively form posteriors in line with uncertainty in its future, irrespective of whether that uncertainty actually persists. After discussing the FEP, we explain how anxiety develops through learning uncertainty before suggesting predictions for how the model can be tested.