Lecture 11                                                      Philosophy of Cognitive Science


Three Levels


1.  Last time we talked about the division of mental function into three levels: ecological, computational, and physical.  This is a conceptual division that now allows us to organize a number of philosophical positions with regard to it.


2.  We already saw that functionalists may wish to forgo discussions of the physical level as unimportant to the purposes of cognitive science.

     One might be tempted to think of functionalists as physical level eliminativists, but this is too strong.  Functionalists think the physical level exists as a viable explanatory level, but that it is uninteresting in explaining cognitive phenomena.  Moreover, in order to have a complete explanation of a mental phenomena, we will want to know the mechanism that implements it.


3.  Sterelny discusses another kind of critic - one that wants to eliminate the computational level.  For many phenomena, e.g. the operation of an odometer, a thermometer, or the behavior of e. coli in the human digestive tract, it seems we don’t need to make reference to ‘computation’.  Rather, we specify the problem and then develop a direct account of the mechanism involved in implementing it.


Saying just why this is wrong headed is quite difficult.  Sterelny suggests that if the system shows informational sensitivity, we will want the computational level.  I suggested that certain kinds of processes are ‘representational’ or ‘representation hungry’.  The idea is that it would be difficult to understand the operations of the device without positing a representational, computational level.


4.  There are also those who would want to eliminate ‘the ecological level’ in explanation.  Of course, we must specify the problem to be solved, but beyond that, reference to this high level of generality plays no causal role in producing an organism’s behavior.  Thus, it plays no explanatory role either.



5.  The computational level gives rise to other problems as well, the first among them being how we should specify when two mechanisms are employing the same algorithm.  Without a way of individuating algorithms, we don’t really know when we’ve correctly specified the computational level.


6.  This kind of problem has led some to believe there really isn’t a particular computational level, but rather many computational descriptions that have more or less equal status.  Pylyshyn, in contrast, argues there is a primitive computational level and offers a number of criteria for how to go about distinguishing it.  You should read Sterelny and familiarize yourself with this discussion.