[This post is part 2 of a summary of Gualtiero Piccinini’s “Computational Modeling vs. Computational Explanation: Is Everything a Turing Machine, and Does It Matter to the Philosophy of Mind?” (2004)]
In part 1, we saw that Gualtiero Piccinini wants to give a formulation of computation that does not entail pancomputationalism, the view that every physical system is a computation. A notion of computation this broad would leave no role for the distinctive role that computation plays in explaining mental behavior.
Before giving his own account, Piccinini discusses to accounts that he finds deficient: one that links computation to representation, one which equates computational explanation and functional explanation.
The representational account says that “only processes defined over representations count as genuine computations, so that only systems that manipulate representations count as genuine computing systems”. This account excludes some things that we do want to exclude, like stomachs and hurricanes. But it is too restrictive. Compilers certainly compute but do not obviously represent anything.
The other account equates computational explanations and functional explanations. Functional explanations (very broadly speaking) explain systems or parts of systems by reference to their purpose or role. For example, understanding the circulatory system as serving a certain purpose – to circulate blood – and each component part of that playing a subsidiary role (pumping the blood, distributing the blood). One should be careful not to conflate functional and computational explanations. The heart case shows that there are functions which are not computational. Functional and computational explanations are not equivalent. Rather, we should think of computational explanation a distinctive subset of functional explanation. Specifically, Piccinini prefers to think of them as mechanistic explanations – explanations that appeal “to a system’s components, their functions, and their organization”.
Piccinini thinks that we should think of computational systems as those that have mechanisms whose purpose of manipulating a special kind of input in a special way. “Under normal conditions, these systems perform sequences of operations that depend on which strings of digits are present within the system and on the internal state of the system, so that by performing those operations, the system may generate different output strings of digits in response to different input strings of digits and different internal states.” We can characterize these systems as following rules. For example, we can characterize the calculator from the previous post as following certain rules to perform addition, division, and what not. The rule provides a certain kind of explanation of behavior – the output “8” is explained as a result of the input (“4+4”) and what the rule says the system does is supposed to do to that input. You can also give a mechanistic explanation of how the system follows that rule – for example, what different components of the calculator do in manipulating the input. Note that this system is also a functional system: it’s function is to add. Note also that this is not just a system whose behavior can be modeled computationally – the system has a function, and a mechanistic organism, that means that it transforms inputs in a particular way. It is performing computations.
This account of computationalism does not lead to pancomputationalism. And it makes it an open, well-defined empirical question of whether the brain is a computing system. Does it manipulate digit-like input states, based on the input and the internal state of the brain, in a way that can be described as following rules?
Experience Machines 