A Reference-Dependent Computational Model of Anorexia Nervosa.

Influential accounts interpret anorexia as arising from perfectionism, dichotomous thinking, and poor control expressed in a variety of life domains, resulting in low self-esteem. In this context, restraining eating would allow patients to re-establish some control and self-esteem. Although this view has offered important insight, one shortcoming is that constructs such as perfectionism, control, and dichotomous thinking, remain poorly specified. To clarify these constructs, we propose a computational model of anorexia. This relies on previous theories of evaluation, which highlight its reference-dependent nature: when attributing a value to an outcome, our brain automatically assesses the outcome relative to its context. Following these theories, the model proposes that a high reference point explains general characteristics such as perfectionism, dichotomous thinking, low self-esteem, and low sense of control. These characteristics would result specifically in anorexia when the sense of control regarding body shape, compared with other life domains, is relatively high. The model raises the possibility that reference effects also might explain why patients pursue extremely low weight; exposure to skinny body images-one product of obsessive dieting-might change the reference point for their own body, hence leading to extremely low body weight, staunch refusal to gain weight, and body misperceptions. The model contributes to clarify key concepts adopted in the literature and their relation. Such computational formulation might help to foster theoretical debate, formulating novel empirical predictions, and integrate psychological and neuroscientific perspectives on anorexia.