[Comp-neuro] Modelling Philosophy

[Harry Erwin]

I'm concerned with modelling M-systems. Conceptually for me, an M- 
system is a system that maintains a model of its environment and uses  
that model to assess the current value of actions leading to future  
rewards and penalties. Those assessments are used to choose actions  
following some rule. The model of the environment need only have  
sufficient detail to support action assessment, and there are a number  
of possible ways that multiple rewards and penalties might be  
integrated into the value of an action.

Free-living eukaryotic cells embody M-systems, as do primitive neurons  
in nerve nets, and as do brains. (My standard example of an M-system  
is an echolocating bat hunting its dinner.) The complexity of an M- 
system seems to reflect a number of evolutionary processes concerned  
with the number and types of actions evaluated, how multiple rewards  
and penalties are integrated into the value of an action, how actions  
are chosen, and whether actions are integrated into plans. In humans,  
the mind (a complex M-system) appears to engage in a dialogue with the  
future to value actions and plans, and that loopy interaction with a  
space of possible futures that responds actively to plans and actions  
underlies the sense of free will.

I prefer to model brains as systems of neurone models. I do this  
because I don't think we understand M-systems well enough to model  
them in the abstract. By sticking fairly close to the biology and  
considering evolutionary processes, I think we are sufficiently  
constrained by reality to characterise at least some M-systems.

--
"an academic who listens to pleas of convenience before publishing his  
research risks calling into doubt the whole of his determination to  
find the truth." (Russell 1993)
Harry Erwin