Allostatic Self-efficacy: A Metacognitive Theory of Dyshomeostasis-Induced Fatigue and Depression

Abstract

This paper outlines a hierarchical Bayesian framework for interoception, homeostatic/allostatic control, and meta-cognition that connects fatigue and depression to the experience of chronic dyshomeostasis. Specifically, viewing interoception as the inversion of a generative model of viscerosensory inputs allows for a formal definition of dyshomeostasis (as chronically enhanced surprise about bodily signals, or, equivalently, low evidence for the brain's model of bodily states) and allostasis (as a change in prior beliefs or predictions which define setpoints for homeostatic reflex arcs). Critically, we propose that the performance of interoceptive-allostatic circuitry is monitored by a metacognitive layer that updates beliefs about the brain's capacity to successfully regulate bodily states (allostatic self-efficacy). In this framework, fatigue and depression can be understood as sequential responses to the interoceptive experience of dyshomeostasis and the ensuing metacognitive diagnosis of low allostatic self-efficacy. While fatigue might represent an early response with adaptive value (cf. sickness behavior), the experience of chronic dyshomeostasis may trigger a generalized belief of low self-efficacy and lack of control (cf. learned helplessness), resulting in depression. This perspective implies alternative pathophysiological mechanisms that are reflected by differential abnormalities in the effective connectivity of circuits for interoception and allostasis. We discuss suitably extended models of effective connectivity that could distinguish these connectivity patterns in individual patients and may help inform differential diagnosis of fatigue and depression in the future.

Keywords: active inference; allostasis; computational psychiatry; dynamic causal modeling; effective connectivity; homeostasis; multiple sclerosis; predictive coding.

Figure 1

(A) Bayes theorem provides the foundation for a generative model m. This combines the likelihood function p(y | x, m) (a probabilistic mapping from hidden states of the world, x, to sensory inputs y) with the prior p(x | m) (an a priori probability distribution of the world's states). Model inversion corresponds to computing the posterior p(x | y, m), i.e., the probability of the hidden states, given the observed data y. The posterior is a “compromise” between likelihood and prior, weighted by their relative precisions. The model evidence p(y | m) in the denominator of Bayes' theorem is a normalization constant that forms the basis for Bayesian model comparison—see main text. (B) Suitably specified and validated generative models with mechanistic (e.g., physiological or algorithmic) interpretability could be used as a computational assay for diagnostic purposes. The left graphics is reproduced, with permission, from Garrido et al. (2008). (C) Contemporary models of perception (the “Bayesian brain hypothesis”) assume that the brain instantiates a generative model of its sensory inputs. Perception corresponds to inverting this model, yielding posterior beliefs about the causes of sensory inputs. The globe picture is freely available from http://www.vectortemplates.com/raster-globes.php.

Figure 2

(A) A graphical summary of predictive coding. See main text for details. Figure reproduced, with permission, from Rao and Ballard (1999). (B) A possible neuronal implementation of predictive coding. See main text for details. Figure reproduced, with permission, from Friston (2008).

Figure 3

A graphical summary of computational and physiological key components of hierarchical Bayesian inference. Computationally, prediction errors are conveyed by ascending or forward connections, while predictions are signaled via backward or descending connections. Critically, both experience a weighting by precision. Physiologically, the currently available evidence suggests that, in cortex, prediction errors are signaled via ionotropic glutamatergic receptors (AMPA and NMDA receptors), predictions via NMDA receptors, while precision-weighting is either implemented through neuromodulatory inputs (e.g., dopamine or acetylcholine) or by local GABAergic mechanisms. The figure is adapted, with permission, from Stephan et al. (2016b).

Figure 4

(A) Principles of classical feedback control. Figure is reproduced, with permission, from Powers (1973). (B) A graphical summary of allostasis and its dependence on predictions about future bodily states. Figure is reproduced, with permission, from Sterling (2012).

Figure 5

A graphical summary of a homeostatic reflex arc and its modulation by allostatic predictions. Blue lines: sensory inputs; red lines: prediction errors; green lines: predictions.

Figure 6

A simulated example of allostatic regulation of homeostatic control, based on Equations (8)–(12). The upper panel shows the temporal evolution of a fictitious physiological state x (Equation 11) which is affected by environmental perturbations (➊,➌,➏; all with a magnitude of 1.5). The middle and lower panels display an approximation to interoceptive surprise—i.e., squared precision-weighted prediction error (pwPE²; compare last line of Equation 8)—and the associated action signal (Equation 10), respectively. Following the timeline from left to right, the homeostatic setpoint or belief is initially specified with a prior mean and prior precision of 1 each. Please note that even before the first perturbation (➊) occurs, sensory noise (zero mean, 0.25 standard deviation) leads to ongoing actions of minute amplitude which lead to (very small) deviations of x from the setpoint. Following a first perturbation (➊), the homeostatic reflex arc emits corrective actions that are proportional to precision-weighted viscerosensory prediction error (middle panel). As the actions are successful, x returns to setpoint and viscerosensory prediction error decays. ➋ indicates the beginning of allostatic control: here, the prediction of imminent future perturbations (by some generative model not specified here) leads to an anticipatory rise in the homeostatic setpoint (a shift in the prior mean to 2). As a consequence, in the absence of any change in sensory input, actions are elicited to change the value of x to the new setpoint. This ensures that the following perturbation ➌ does not bring x anywhere near the critical threshold. At ➍, a safe period is predicted, and allostatic control resets the homeostatic setpoint (prior mean) to 1. At ➎, another perturbation in the near future is being predicted, however, this time the direction of the perturbation is uncertain. Therefore, changing the mean or setpoint is not a viable option and allostatic control takes a different form: instead of changing prior mean, the prior precision of the homeostatic belief is increased from 1 to 4. As a consequence, when a perturbation occurs at ➏, this yields a considerably larger precision-weighted prediction error and hence greater interoceptive surprise (see lower panel), leading to a significantly more rapid corrective action (compare the slope of signal rise between ➊ and ➏), putting the agent at less risk, should another perturbation occur shortly after ➏. It is also noteworthy that the increased prior precision enhances the effect of sensory noise (compare the roughness of the three signals just prior to ➊ and ➏, respectively).

Figure 7

A proposed circuit for interoception and allostatic regulation of homeostatic reflex arcs, together with a metacognitive layer (MC). See main text for details. Blue lines: sensory inputs; red lines: prediction errors; green lines: predictions.