[ACT-R-users] computational basis of act-r

[Roman Belavkin]

> Connectionism is not the same as the dynamic system approach.

ACT-R model can easily be seen as an example of dynamic programming.  In my
opinion there clearly is a suspicion from connectionists towards sybolists
and vica versa.  It is, however, very similar to what happned in quantum
mechanics between the followers of wave and particle interpretations.  They
are two sides of the same coin, and brain cannot be fully understood using
only one interpretation.  In the end of the day connectionists have to
cathegorise (quantise) vectors before they can do anything.  Thus, they
begin to operate in discrete domain (symbolic).  Sybolists have to use
probability and statistics to overcome the constraints of deterministic
approach.  Thus, they begin to operate with continuum (subsymbolic).  You
may soon realise that two approaches meat each other.  The main problem is
how to put them together in one consise system and make it work.  This is
exactly what ACT-R has been trying to do, and they succeeded, perhaps, more
than others.  Whether it will remain to be the best candidate for the
unified theory of cognition in two years or not we don't know.  Let's hope
there will be more choice.

Roman Belavkin