From threat perception to behavioral decision: A computational framework for understanding approach-avoidance in social anxiety.

Abstract

Cognitive models of social anxiety disorder describe what people find threatening and what sustains those appraisals, but say less about the step at which an appraisal becomes a decision to approach or avoid. This review organizes evidence around a decision process with three components. Valuation weighs anticipated social threat against anticipated social reward, with perceived controllability proposed to set the balance. Policy converts the resulting value into an action. Learning revises value estimates from outcomes and, separately, governs the rate at which a perturbed affective state returns to baseline. Domain specificity operates as a boundary condition: these processes are engaged to the degree that a context places a self-relevant attribute under the judgment of others and makes the outcome follow from that judgment. Because relevance is a conjunction of conditions that can each hold partially, the boundary operates by degree rather than as a switch, though this has not been tested at the situation level. Stating each component as a set of parameters yields a rule for assigning findings to one component or another, and allows each to be manipulated separately. Support is strongest for reward valuation and for asymmetric updating, weaker for controllability and recovery, and weakest for Policy, where the evidence on the threshold is mixed and the three parameters have never been estimated together. The framework's contribution is not a new account of what individuals with social anxiety fear, but a statement of the intervening process precise enough to be tested one part at a time.