Homeostasis and causal control

Abstract Homeostasis is a core biological concept concerning the way organisms maintain quantities within a given range. Mathematical models of homeostasis build on mechanisms for feedback control from engineering. The concept of control also figures centrally in causal analysis, where it is typically linked to interventions. This paper uses dynamic causal models (Iwasaki and Simon, 1994) to bridge these three domains, leading to the correction of a misunderstanding common in all of them. In each, control is regularly understood as aiming to maintain certain variables at constant values. Yet paradigm control feedback and homeostatic mechanisms do not maintain these variables constant, and intervening on them would destroy the mechanisms’ functions. Instead of focusing on how feedback loops allegedly maintain constancy, dynamic causal models reveal how control systems exploit such lower-scale feedback loops to produce higher-scale causal relationships. The discussion points toward a novel engineering-inspired approach to causal inference, and clarifies how biological and engineered systems combine existing mechanisms to perform increasingly complex functions.