Stimulating, indeed! I’m only halfway through this dialogue between Terrence Deacon and Tim Jackson and can’t help but share a few preliminary reflections.

They discuss the limitations of optimization models like the FEP, which capture the ways living systems actively resist dissolution by minimizing surprise. The active inference framework also includes an epistemic drive in an attempt to account for curiosity, exploration, play, etc., again in terms of uncertainty reduction. So these models aren’t entirely deaf to creative excess of life, but they do seem to me to try to explain said excess by swallowing it back into the same sort of uncertainty reduction story.

Drawing on Whitehead’s speculative scheme (see especially Process and Reality (1929/1978), pgs. 197-207), we could mark a distinction between two fundamentally different modes of evaluation: statistical judgment (basically, induction) and non-statistical judgment (akin to Peircean abduction); or, to put it in more evocative terms, between error minimization and eros maximization.

The former, statistical judgment, is what Bayesian inference formalizes. It captures the inductive moment of biological cognition rooted in the reinforcement of what already works, eg, the convergence of a lineage on a reliable adaptive strategy. Organism behavior can thus be modeled as the minimization of prediction error relative to a generative model.

Non-statistical judgment, on the other hand, operates by an entirely different onto-logic. In this case, the organism is not consulting a prior probability distribution to derive an optimal response. Instead, it feels thequalitative gradient of potentiae unique to each novel situation and relational context in a way that could not have been pre-specified, because it concerns a configuration of possibilities that has no prior statistical ground. Whitehead's name for this felt gradient is the “initial aim.” There’s a lot of background I’ve spelled out elsewhere (see the link in the comment below). But the gist is that Whitehead tries to account for the unexpected success of non-statistical inferences of probability (what Peirce called abduction, and what contemporary cognitive scientists address in terms of “the frame problem”) by showing how an otherwise infinite array of qualitative potentiae are prehended by an organism in each moment of its life history not as a quantitatively pre-defined menu of set states but as a qualitatively adjusted intensive lure toward relevance. He is thus introducing, alongside the need for error minimization, a cosmological drive toward eros maximization, ie, a felt attractor toward depth of experience and enhancement of meaning that exceeds what models derived from past regularities can account for.

FEP is an extensional framework. That is what makes it so powerful as a formalization. But it also means that its mathematical quantities are indifferent to what it is like to be in a given state. FEP treats organisms as if they were Bayesian calculators. This “as if” is quite powerful, but we ought not rush to ontologize it! For Whitehead, organisms are understood to prehend potentiae with a specific qualitative intensity that is not captured by assigning a probability to the corresponding state. Two occasions in the life history of an organism could occupy same position in state space, have the same measurable properties, and yet differ in their subjective form, their aesthetic depth, their degree of experiential realization, etc. Eros maximization is not simply the inverse of error minimization but a re-orientation of our attempts to think living organization so as to include qualitative intensities alongside quantitative extensities. The difference between an organism that merely survives via adaptation and an organism that seeks to intensify its experiential satisfaction is not best modeled simply as a difference in error minimization, but as a factor of eros maximization.

Terry discusses his idea of “inverse Darwinism” (which for Tim is really more like a faithful retrieval of Darwin’s approach prior to the neo-distortions). The idea is that biological creativity depends on the relaxation of selective constraints, allowing for the accumulation of variation before adaptive selection.

Terry also dwells on the way phosphate compounds on the early Earth were initially actively destructive of the organized molecular processes required for simple life to emerge. But at some point a transformation occurred and phosphate became a primary source of biological energy and genetic information storage. This struck me as a neat example of Hegelian sublation, a sort of enemy/friend dialectic clearly evident in this and in the frequent examples of evolution by symbiogenesis. The very reactivity that made phosphate destructive was sublated into a useful biological function. In Hegelian terms, the negative moment is preserved as productive force within the higher integration. Is this best understood as the minimization of surprise relative to a pre-existing model? It seems to me more like an example of creative incorporation driven by what Whitehead would call the organism's appetition toward a richer qualitative contrasts: a felt possibility of integration that no prior distribution over the system's state space could have encoded, because the relevant state space did not yet exist.

Whitehead claimed that “the art of persistence is to be dead.” If the development and evolution of life can be exhaustively explained by error minimization, why does life appear so excessive, so bent on creative advance?