17Energy homeostasis depends on behavior to predictively regulate metabolic states within 18 narrow bounds. Here we review three theories of homeostatic control and ask how they 19 provide insight into the circuitry underlying energy homeostasis. We offer two 20 contributions. First, we detail how control theory and reinforcement learning are applied 21 to homeostatic control. We show how these schemes rest on implausible assumptions; 22 either via circular definitions, unprincipled drive functions, or by ignoring environmental 23 volatility. We argue active inference can elude these shortcomings while retaining 24 important features of each model. Second, we review the neural basis of energetic 25 control. We focus on a subset of arcuate subpopulations that project directly to, and are 26 thus in a privileged position to opponently modulate, dopaminergic cells as a function of 27 energetic predictions over a spectrum of time horizons. We discuss how this can be 28 interpreted under these theories, and how this can resolve paradoxes that have arisen. 29
We propose this circuit constitutes a homeostatic-reward interface that underwrites the 30 conjoint optimisation of physiological and behavioural homeostasis. 31 32Keywords. reward prediction error, dopamine, hypothalamus, energy homeostasis, active inference
33. CC-BY-NC-ND 4.0 International license It is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.The copyright holder for this preprint . http://dx.doi.org/10.1101/242974 doi: bioRxiv preprint first posted online 3
The problem of homeostatic control
34A remarkable feature of physiological systems is their stability. Most physiological 35 variables are regulated within narrow bounds by operational and computational processes 36 collectively known as homeostasis (Cannon 1932). The mechanistic complexity of 37 homeostasis extends beyond simple negative feedback control and embodies a wide 38 spectrum of hierarchically organised physiological control structures, molecule to agent, 39 operating over a multitude of timescales, milliseconds to months (Carpenter 2004). 40Homeostatic control is often framed as the regulation of variables around a fixed set point, 41the achievement of which upholds a physiological equilibrium (Cannon 1932 states to fall within the ranges that afford organismal survival (Fig 1b; Sterling 2012). 55For all motile agents, effective homeostatic control results from the interplay among 56 automated physiological processes (henceforth referred to as physiological homeostasis) 57 . CC-BY-NC-ND 4.0 International license It is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.The copyright holder for this preprint . http://dx.doi.org/10.1101/242974 doi: bioRxiv preprint first posted online 4 and overt behaviour (henceforth, behavioural homeostasis). The coordinated mechanisms 58 of phys...