Page 129 - Contribution To Phenomenology

122 TOMNENON

Before, I proceed to a few remarks about the relationship between
Phenomenology and Connectionism, I should say at least a word or two
about learning in connectionist systems. "Learning*' takes place not by
changing inputs, or changing initial activations, but rather by "changing"
or "shifting the weights" of the connections between various nodes for an
output that is considered a mistake, until—perhaps after a large series of
adjustments—a more appropriate correlation between input and output
is achieved. Up until now the decision about whether the weights need
to be changed and how much they should be changed has been made by
an external operator who is also the interpreter of the outcome in terms
of desired "real life" results. In the fourth part of this paper, I will
suggest what might be a more appropriate, though certainly more
complicated way for learning to be enacted, one which would follow
rather naturally from a different notion of representation inspired by
Phenomenology as transcendental philosophy.

II

From the standpoint of Husserlian Phenomenology, one might well ask
what all of this has to do mental Ufe of human beings. For as human
beings, who are in Husserl's view subjects (more correctly, as subjects,
who happen to be human beings as well), we learn of our mental life
not by studying machines or analyzing mathematical systems, or even by
learning how brains work, but rather by closely observing and analyzing
what is given to us through the immediate awareness of consciousness
and its products to itself. We know of ourselves as the bearers of mental
states and we know of the mental states (das Psychische) of which we
are the bearers directly and completely whenever we turn back away
from the objects that are the immediate focus of our intentional states
such as believing, desiring, loving, or hating, and focus our attention upon
that which is immediately given to us as such, namely those acts of
believing, desiring, etc. in which the objects present themselves to us.
The various procedures associated with the term "phenomenological
reduction" are really nothing other than ways to make sure that we avoid
making any commitments to the existence of the objects of such
intentional acts and focus simply upon the acts themselves and what is
involved in them, not as real events in the history of actually and
indubitably existing individuals with these or those characteristics, this or
that name, but rather as a realm of phenomena that are immediately