[Paleopsych] Freud and traumas

Werbos, Dr. Paul J. paul.werbos at verizon.net
Fri Jul 2 01:36:21 UTC 2004

Hi, folks!

There are a lot of cognitive psychologists who are as traumatized by 
Freud's ideas as
Freud's patients were by childhood experiences. This may be due in part to 
many of them
associating of Freud with sneaky girls who really did use Freud out of 
context to try to traumatize
any easy victims like young nerds. But it is also due to the diificulty of 
"making sense" of Freud's
ideas in more operational, functoinal terms.

Before you hit the traumas -- there was a famous, beautifully clear 
critique of psychoanalysis
(including Freud and others) by Baret and Yankelovi(t)ch years ago, a nice 
cheap paperback.
It described Freud's theory of psychodynamics and the flow of cathexis -- 
and characterized it as hopelessly
mystical. But in fact, for my Harvard PhD thesis, I translated it into 
mathematics -- and it maps
into a very general useful theorem, and to the most powerful and widely 
used algorithms available in the field
of artificial neural networks.  (Unfortunately, most folks only use a 
stripped-down version
of backpropagation, whose power is limited, but something close to the 
original more complex translation of Freud
has been proven out on some very tough engineering control tasks; see Si et 
al, eds, Handbook of Learning and Approximate Dynamic
Programming, literally in press at Wiley and IEEE Press.) The history and 
links to Freud are discussed in
my book The Roots of Backpropagation: From Ordered derivatives to Neural 
Networks and Political Forecasting,
especially chapter 10, Wiley, 1994. (A good university library should have 
it, as it contains the original material
on backpropagation, for which the popularized versions are generally very 


But... well... it has taken a long time to catch up to that one part of 
Freud, and really understand what's going on,

The trauma part I discussed 'way back in my paper in the 1977 General 
systems yearbook, in a small section entitled
"syncretism." And I have tried to revisit it, in a simpler way, many times 
since. But I haven't had time to
really hammer home all of the key points to the general community. (There 
is a book on supervised learning edited by Roychowdhury,
and the IJCNN proceedings from 1994...).

Basically, global generalization requires many iterations through 
experience. Learning "from memory" is critical to
our ability to generalize without needing to relive bad experience over and 
over again in real life. That's one
reason why we need these kinds of memories, which might be called 
subsymbolic episodic memories.
"Assimilation into the ego" basically means adapting our more coherent 
"egoic" generalizing model
of the world so as to be able to "backcast" or "explain" the memory, 
retroactively. But for the most accurate expectations,
we need to COMBINE our "egoic" predictions together with a kind of "nearest 
neighbor" kind of
allowance for memories we have not been able to assimilate, which point 
towards weaknesses in our egoic model.

Maturity does NOT consist in reducing the influence of the "id," the 
collection of unassimilated engrams;
rather, it involves a process of better assimilation of existing memories, 
COMBINED WITH teh formatoin
of new memories or experiences which "expand the being space" (in 
Heiddegerian terms).

Of course, people vary in their drive to assimilate (intolerance of 
cognitive dissonance, driving need for ego ...)
and their drive to accumulate new experience (novelty seeking).


All for now...



More information about the paleopsych mailing list