Introduced in machine learning as a statistical tool for computing the weights of artificial neural networks during training (see, for example, Hinton and van Camp [1993]; MacKay [unpublished]), the scope of application of the information-theoretic notion of free energy has been ballooning. Over the past twenty years, theoretical neuroscientist Karl Friston and his collaborators have turned this notion into a substantive principle for studying and understanding brains (Friston et al. [2006]; Friston [2009]; Parr et al. [2022]), organisms with and without a nervous system (Friston [2013]), and everything else (Friston [unpublished]). When applied in neuroscience, this principle states that the brain optimizes a free-energy function. This function, according to Friston and collaborators, measures how closely the brain’s representation of the world approximates the true state of the world; and by optimizing this function, the brain enables animals to select actions that further their chance to survive and to preserve their physical integrity.

Partly because of its breadth, partly because of its association with popular predictive processing accounts of the brain (see, for example, Hohwy [2013]; Clark [2015]), and partly because of the multifarious philosophical claims put forward by its architects, the free-energy principle has received enormous attention in several fields, including cognitive neuroscience, biology, physics, and philosophy. Some have even suggested that it might be considered a ‘Schrödinger’s equation for the brain’ (Dehaene [2008]) or a transcendental principle for the life sciences (Colombo and Wright [2021]).

Surprised by this level of attention, a philosophically inclined computational neuroscientist friend of mine once told me: ‘This debate on the free-energy principle looks like an unfortunate situation to me, where philosophers make misinformed (pseudo)scientific claims and scientists make confused (pseudo)philosophical arguments’. Kathryn Nave’s A Drive to Survive is a welcome intervention in this debate—and my friend should read this book, as it will change their mind.

Written with flair and accessible to an interdisciplinary readership in philosophy and science (but sadly containing several typographical errors too), Nave’s A Drive to Survive has two main goals, one negative and the other positive. The negative goal is to show that the free-energy principle cannot constitute an adequate theoretical foundation for understanding the nature and dynamics of living systems. The positive goal is to demonstrate that a bioenactivist approach fares better than the free-energy principle in affording such a foundation. By combining insights from phenomenology and biology, bioenactivism attempts to ground attributions of function, purpose, and normativity to living systems in ‘biological autonomy’, that is, in biological processes of self-generation and self-sustenance. Such processes would explain the emergence and constitution of biological individuality, agency, and ‘intentionality in terms of the striving of a system towards some non-reconstructive goal or norm’ (p. 7). This means that the processes constituting organisms and their mental lives are aimed not at producing accurate and detailed mental representations of some perceiver-independent world based on inferences of distal causal structures from impoverished sensory information. Instead, these processes are aimed at adaptive and autonomous self-production. Without a bioenactivist theoretical foundation supplying non-inferential, non-reconstructive conceptions of intentionality and normativity, accounts of mind and life relying solely on the free-energy principle and its associated active-inference framework cannot be complete or satisfactory.

In focusing on both the free-energy principle and bioenactivism, Nave does a double service to her readers. First, she helpfully elucidates the core commitments of the free-energy principle in its various, evolving formulations, as well as its relationship with the predictive processing and active-inference frameworks (especially in chapters 2–5 and the appendix). Second, she spells out, motivates, and illustrates the often elusive arguments and ideas behind enactivist approaches to life and mind (especially in chapters 1, 10, and 11).

A Drive to Survive develops a sustained argument that weaves together reasons against those views—such as W. Ross Ashby’s ([1956]) cybernetics and Friston and collaborators’ free-energy principle (Pezzulo et al. [2015])—that conceive of organisms as, fundamentally, persisters aiming to find homeostatic stability amid external perturbations. It also provides reasons in favour of views such as bioenactivism, which conceive of organisms as networks of energy constraints ‘whose precarious existence is dependent upon their own activity in working towards their continuous self-production’ (p. 7).

According to Nave, if free-energy theorists conceive of living systems as predictive controllers seeking homeostatic stability, then their conception is too general to distinguish the living from the non-living. After all, this conception is applicable to any physical system whose dynamics can be represented as trajectories through attractive, non-equilibrium steady states in a fixed phase-space governed by rules of evolution, pre-defined over some set of characteristic variables (see, for example, Colombo and Palacios [2021]). Free-energy theorists’ conception of a living system as homeostatic predictive controller is also too narrow, according to Nave. This is because ‘the principles of stability that [the free-energy principle] presents as necessary [for survival and functional integrity] are exactly those principles that biological systems are uniquely prone to violate’ (p. 155) in virtue of distinctive properties that they possess, such as their metabolism and capacities for ongoing change and material turnover. At best, the free-energy principle can provide us with formal tools helpful for approximately representing and studying ‘a contingent, if common, feature of biological dynamics’, namely, one way that an organism, ‘for so long as the organism’s organization does not change’ (pp. 208–9), can regulate its states to remain within the limits of viability.

The survival, structural identity, and functional integrity of living systems over time does not on stability, but rather on processes of ‘constraint closure’ (see, for example, Montévil and Mossio [2015]). These are processes of continuous, constrained energy transformation in an organism that produce a certain kind of structural organization, which in turn enables and constrains the very metabolic processes that bring about and constrain that structural organization. For Nave, it is this dynamic, reciprocal dependence between an organism’s metabolic processes and their structural constraints that make the organism an autonomous, intentional agent, differentiating it from other self-maintaining systems, such as hurricanes or candle flames. There is no stable set of mechanisms of homeostasis in living things. Instead, there is a kind of active maintenance of homeostatic systems in living things (which are indeed stable) that is not present in candles.

While Nave’s critical argument against the free-energy principle as an encompassing foundation for understanding life and mind is carefully laid out, the bioenactive approach she advocates is comparatively less detailed. Assuming mind–life continuity—that is, assuming that where there is life there is mind—this approach would ideally ground teleological attributions of purposiveness and intentionality to any living systems. But these attributions risk ascribing mental actions where there are none, based on the apparent flexibility exhibited by the behaviour of organisms such as cells, fungi, and plants. Constraint closure might ground the kind of self-maintenance distinctive of living systems without offering a convincing basis for explaining the patterns of flexible and inflexible behaviour that they display. Even if constraint closure explained patterns of flexible and inflexible behaviours in an organism based on whether the organism’s behaviour is dependent on its metabolic state, this would still not be enough to warrant ascriptions of intentionality. If a bacterium’s behaviour is flexible because it is sensitive to the bacterium’s metabolic state, it does not obviously follow that that behaviour is also intentional or goal-directed, unless we dilute the sense of notions such as ‘intentionality’, ‘goal-directed behaviour’, and ‘meaning’. At any rate, more needed to be said to convincingly bridge the gap between, on the one hand, constraint closure and meaningful ascriptions of intentional agency and normativity to, on the other, the amazing behaviours of organisms such as bacteria, fungi, and plants. Nave’s A Drive to Survive makes a helpful step in this direction. But future attempts to work out a detailed bioenactivist account of life, intentional agency, and normativity should pay closer attention to the evidential sources and explanatory practices of contemporary scientists in many fields in the sciences of life and mind.

In the face of the promise and hype surrounding both the free-energy principle and enactivism, Nave’s A Drive to Survive should be of interest to anybody who wants to make better sense of these two ambitious approaches purporting to lay bare the deep continuities between the traditional domains of physics, biology, and psychology.

Acknowledgements

I thank Claudia Cristalli and María Jimena Clavel Vázquez for their helpful comments on previous versions of this review.

Matteo Colombo
University of Tilburg
m.colombo@uvt.nl

References

Ashby, W. R. [1956]: An Introduction to Cybernetics, London: Chapman and Hall.

Clark, A. [2015]: Surfing Uncertainty: Prediction, Action, and the Embodied Mind, Oxford: Oxford University Press.

Colombo, M. and Palacios, P. [2021]: ‘Non-equilibrium Thermodynamics and the Free Energy Principle in Biology’, Biology and Philosophy, 36, available at doi.org/10.1007/s10539-021-09818-x.

Colombo, M. and Wright, C. [2021]: ‘First Principles in the Life Sciences: The Free-Energy Principle, Organicism, and Mechanism’, Synthese, 198, pp. 3463–88.

Dehaene, S. [2008]: ‘The Brain’s Schrödinger’s Equation’, Edge, available at www.edge.org/response-detail/10566.

Friston, K. [2009]: ‘The Free-Energy Principle: A Rough Guide to the Brain?’, Trends in Cognitive Sciences, 13, pp. 293–301.

Friston, K. [2013]: ‘Life as We Know It’, Interface, 10, available at doi.org/10.1098/rsif.2013.0475.

Friston, K. [unpublished]: ‘A Free Energy Principle for a Particular Physics’, available at doi.org/10.48550/arXiv.1906.10184.

Friston, K., Kilner, J. and Harrison, L. [2006]: ‘A Free Energy Principle for the Brain’, Journal of Physiology–Paris, 100, pp. 70–87.

Hinton, G. E. and Van Camp, D. [1993]: ‘Keeping the Neural Networks Simple by Minimizing the Description Length of the Weights’, in L. Pitt (ed.), Proceedings of the Sixth Annual Conference on Computational Learning Theory, New York: Association for Computing Machinery, pp. 5–13.

Hohwy, J. [2013]: The Predictive Mind, Oxford: Oxford University Press.

MacKay, D. J. C. [unpublished]: ‘Ensemble Learning and Evidence Maximization’, available at www.inference.org.uk/mackay/nips.pdf.

Montévil, M. and Mossio, M. [2015]: ‘Biological Organisation as Closure of Constraints’, Journal of Theoretical Biology, 372, pp. 179–91.

Parr, T., Pezzulo, G. and Friston, K. J. [2022]: Active Inference: The Free Energy Principle in Mind, Brain, and Behavior, Cambridge, MA: MIT Press.

Pezzulo, G., Rigoli, F. and Friston, K. [2015]: ‘Active Inference, Homeostatic Regulation, and Adaptive Behavioural Control’, Progress in Neurobiology, 134, pp. 17­–35.