Objective Bayesian Reality 10

We should expect, by definition, that any system evolving through the operation of a Darwinian process produces adaptations: characteristics that will increase the system's fitness to survive.

Friston models much of the subject matter of behavioural biology and neuroscience as an adaptive system with three components: the state of the system, the effect of the system on the environment, and the effect of the environment on the system. [i] Using this model he is able to arrive at a number of conclusions:

1) The system will possess internal models portraying external causes in the environment.

2) The system will evolve, through natural selection, to reduce the discrepancy between the internal models and the external environmental causes.

3) A generic scientific law can be proposed that predicts brain functions in numerous contexts; the 'surprise', bounded by free energy, between internal models and external reality is minimized.

In summary, the free-energy principle can be motivated, quite simply, by noting that systems that minimise their free-energy respond to environmental changes adaptively. It follows that minimisation of free-energy may be a necessary, if not sufficient, characteristic of evolutionary successful systems. The attributes that ensure biological systems minimise their free-energy can be ascribed to selective pressure, operating at somatic (i.e. the life time of the organism) or evolutionary timescales (Edelman 1993). These attributes include the functional form of the densities entailed by the system’s architecture. Systems which fail to minimise free-energy will have sub-optimal representations or ineffective mechanisms for action and perception. These systems will not restrict themselves to specific domains of their milieu and may ultimately experience a phase-transition (e.g., death).

This finding illuminates not only the nature of adaptive systems but also the nature of the evolution of emergent levels of matter within objective reality.

The brains of humans and other intelligent mammals are low entropy structures which require explanation. The principle of Maximum Entropy tells us to expect that their existence is dependent on constraints in the form of scientific law.

Evolutionary origins of brain functions can be traced to the abilities of early single celled life to perform phototaxis and chemotaxis. Such abilities clearly provide a selective advantage as they promote adaptive exchanges with the environment.[ii] Natural Selection, the Darwinian process involved, has selected progressive improvements in these abilities, ever more optimal mechanisms for adaptive exchanges with the environment. This progression is evident in examining such things as the evolution of neural chemistry in a chain of life forms from yeast cells to more developed organisms, including vertebrates.[iii]

Natural selection operates through the evolution of adaptations such as sensing, perception and learning. Each of these adaptations has its own design details embodied in its functioning. The organisms possessing these adaptations are dependent on them for their survival. Superior adaptive designs are rare and, when found through the operation of a Darwinian process, tend to be adopted and copied in numerous guises. These powerful, widely-copied mechanisms must be rational; they must adhere logically to the objective nature of their environment. Generic essential details of a widely-replicated design, such as genetics within biology or optimal Free Energy structures in the brain, can sometimes be isolated and proposed as scientific law.

In this manner we are able to view adaptive systems, in accordance with the principle of Maximum Entropy, as moving to states of the highest entropy available, subject to the constraints of their design details, those applicable scientific laws. These laws evolve through the operation of Darwinian processes along with the design of their subject matter.

Successful designs arising in adaptive systems are those that can improve survivability through more optimal exchanges with the environment. Such designs specify lowered entropy states. The specific mechanisms discovered and selected by Darwinian processes to enhance survivability and lower entropy form the constraints (i.e. the applicable scientific laws) that are responsible for preventing the system from moving to states of higher entropy.

Friston's theory provides details of the operation of this process within behaviour and neuroscience, but the model it provides appears to be applicable to adaptive systems in general, as I have attempted to describe above.

[i] Friston K, Stephan KE. ,Free energy and the brain, Synthese. 2007. 159:417–458

[ii] Laura M. Grabowski, Wesley R. Elsberry, Charles Ofria, Robert T. Pennock: On the evolution of motility and intelligent tactic response. GECCO 2008: 209-216

[iii] Emes, RD; Pocklington, AJ; Anderson, CN; Bayes, A; Collins, MO, et al. Evolutionary expansion and anatomical specialization of synapse proteome complexity. Nat Neurosci. 2008;11:799–806.

Adaptive Systems