>In responding to a description of QuestMaps, Hal Popplewell wrote...
>
>> This being done, do we not, then have knowledge encoded in "near" rule
>> form and couldn't this knowledge translated to executable rules for
>> permanent use in future decisions (intelligent decision support)?
>
>Will this work in real knowledge-worker environments (non-linear, crative,
>adaptive, etc.) like software/product development?
I'll jump into this one. :-)
At MCC, where the research was done that lead to QuestMap (gIBIS), there
was also research done on an "AI" version... As I recall, the focus was on
using AI to _make_ decisions, once the humans had input all the "details".
I objected to this approach at the time because I felt that, given all the
"details", humans were better at making decisions than computers. So why
automate the "easy" part.
But, another complaint was the amount of detail necessary to get anything
reasonable out of the AI system. If we encode enough to be "near" rule
form... it is a heck of a lot of work. And who does it benefit? Usually
not the person/people who have to encode it. Why should a worker who is
"temporary" encode the knowledge for someone else?
So my answer is I don't believe it will work in the "real world"... not
because of the the non-linear, creative aspects of many 'real' tasks...
but because of the _human_ aspects.
-- kcby
K.C. Burgess Yakemovic Group Performance Systems, Inc.
kcby@gpsi.com http://www.gpsi.com
"Helping people with people, through technology....
because the "soft stuff" is the _hard_ stuff!"
4776 Village North Court phone & fax 770-395-0282
Atlanta GA 30338 USA