Why Use Predictive Processing to Explain Psychopathology? The Case of Anorexia Nervosa

Abstract

Predictive processing accounts are increasingly called upon to explain mental disorder. They seem to provide an attractive explanatory framework because the core idea of prediction error minimization can be applied to simultaneously account for several perceptual, attentional and reasoning deficits often implicated in mental disorder. However, it can be unclear how much is gained by such accounts: the proffered explanations can appear to have several weaknesses such as being too liberal, too shallow, or too wedded to formal notions of statistical learning. Here, we taxonomise the relatively unrecognised variety of explanatory tools under the framework and discuss how they can be employed to provide substantial explanations. We then apply the framework to anorexia nervosa, an eating disorder that is characterised by a complex set of perceptual, reasoning and decision-making problems. We conclude that the predictive processing framework is a valuable type of explanation for psychopathology.

Visual Self-Misperception in Eating Disorders

“…Nevertheless, they suggest that, in the case of eating disorders, vision might be modulated by non-perceptual states, such as emotions or the expectation of seeing oneself. This possibility warrants further investigation and discussion (Gadsby & Hohwy, 2020).…”

Can predictive processing explain self-deception?

“…The only way for the PEM framework to capture irrational or inadequate belief updating is in terms of the suboptimal generative models (see, e.g. Gadsby & Hohwy, 2021;Schwartenbeck et al, 2015). The significance of this claim will be discussed in subsections 3.2. and 3.3.1. expectations about their son's innocence could explain why the parents disregard PE (i.e., disconfirming evidence) as irrelevant.…”

“…Indeed, the proponents of PEM emphasize that this option is guaranteed to work by "the complete class theorems (…) that state that any behavior, no matter how apparently pathological, is Bayes optimal under the right set of prior beliefs (Parr & Friston, 2021, p. 172). Moreover, they consider this explanatory strategy as advantageous, because explaining things in terms of generative models and prior beliefs should enable us, among other things, "to identify the origins of aberrant inference in the cerebral hierarchy" and "to differentiate between different mechanisms that might cause (…) deviation" that characterizes various pathological conditions (Schwartenbeck et al, 2015, p. 111; see, also Gadsby & Hohwy, 2021).…”

“…More recently, this pattern of explanation has been generalized to account for other neurocognitive abilities and phenomena, including attention, emotion, action, and psychiatric disorders (e.g. Clark, 2016;Gadsby & Hohwy, 2021;Mathys, 2016). Indeed, according to its main proponents, PEM's successes in accounting for the various cognitive processes and the capacities of the brain, strongly support the idea that the single principle of prediction error minimization describes the brain's overarching computational task that underpins all of its subpersonal cognitive processes (Clark, 2013;Hohwy, 2013; see, also Sprevak, 2021a).…”

Is the biological adaptiveness of delusions doomed?

“…Although F&C's model has been developed with schizophrenic and Capgras delusions in mind, in principle there is no reason why the model should not be extended to delusions in other conditions, as predictive coding is an all-encompassing account of human cognition. In fact, beyond schizophrenia and Capgras, predictive coding models have also been developed for other disorders where delusions and delusion-like ideas are present, such as OCD(Levy 2018), depression(Badcock et al 2017) and Anorexia Nervosa(Gadsby and Hohwy 2020).…”