Mental States and Processes
Clark distinguishes between three kinds of things we attribute to minds:
"Feelings" are related to what we've been discussing under the heading of "qualia," and Clark says he won't be focusing on these. (Why? Because we have much less idea how to explain them, and he proposes to look "where the light is" (p. 5).)
So the book will focus on 2 and 3, and the overall goal is to try to see how cognitive science has attempted to explain these within a broadly materialistic world view.
Cognitive Science and Cognition: The Basic Idea
The basic idea is simple, although the details aren't. Consider reasoning: starting from a set of premises, and drawing conclusions that are reasonable given the premises. How can this be done by a complex physical system? Physical processes seem mechanical, not rational.
Well, how does a calculator work? The calculator has been built in such a way that we can regard some of its inputs as representing numbers, and others as representing functions on numbers (like addition and multiplication). Hit the keys "2", "+", "3", "=", and the display shows the sum of 2 and 3, namely 5. This isn't because the calculator is intrinsically rational; its processes are just plain old mechanical processes. But given the way it is designed, those mechanical processes have inputs and outputs that can be interpreted as having meanings, and the outputs are reasonable given the inputs.
Now let's take a fancier scientific calculator that can evaluate complex expressions. Type "(2 + 3) * (3 + 1)". This requires internal states as well as inputs and outputs: before displaying "20" the calculator has to first add 2 and 3, getting 5, store that result in an internal memory location; add 3 and 1, getting 4, and store that result in a different memory location; then multiply the contents of the first memory location by the contents of the second one. We never see the "5" or the "4", so the calculator has internal states which are meaningful but which have only an indirect effect on the behavior of the calculator.
The basic idea of cognitive science is that we're sort of like that. Our inputs and outputs (and internal states) can be interpreted as having meanings, and the relations between them are reasonable. But how is this possible? In the calculator, mechanical processes make sense, produce reasonable results, because that's the way the calculator was designed. But why should internal states of an organism work this way? The basic idea is that this is a result of natural selection plus learning. We know that natural selection produces many organs that work as thought they were designed to perform a certain function -- the eye to transmit information about colors and shapes, the heart to pump the blood, the lungs to oxygenate the blood, and so on. Perhaps natural selection has also produced an organ which acts as though it were designed for reasoning, just as a calculator is designed for calculating?
Brains as Computers
So the idea is that the brain is sort of like a calculator (or, better, a computer), and thinking is sort of like the processes that take place in a calculator (or computer). So thinking is to the brain as running a program is to the computer -- crudely, mind is to brain as software is to hardware.
Clark mentions several key moments in the development of this perspective:
How can we make progress in thinking of cognition as computing? One strategy is to try to develop computer programs which do the kinds of things we associate with human reasoning: play chess, prove theorems, etc. This is the path that AI has taken.
Symbol Systems
So, AI Mach I: the idea that intelligence is realized in what Newell and Simon called "physical symbol systems."
Knowledge gets stored in a bunch of symbolic representations, and intelligence is a matter of being able to draw inferences from those representations, and to construct representations of a problem and search the problem space for a solution.
Clark mentions three kinds of objections to this general approach.
1. Searle's Chinese Room Argument: Searle argues that this approach to AI can only simulate reasoning or understanding, never actually implement it.
2. Dreyfus: Dreyfus argues that human reasoning and action is not rule-governed in most cases, and that rule-governed activity will never be as fluid and natural as whatever it is humans actually do.
3. Grab-bag: the idea that natural selection may have left us, not with one general capacity, but with lots of more specialized capacities.
Last update:
March 24, 2008.
Curtis Brown | Philosophy of Mind | Philosophy Department
| Trinity University
cbrown@trinity.edu