One of the primary concerns which we should all have when approaching the
question
of "design" and "design automation" is the ill-defined, sometimes terribly
structured nature of what "design problems" are. The process of refining the
design problem statement frequently reduced the problem to a well-structured
problem. Indeed, somewhere along the line this must be done if a "solution"
is to be found. But, unlike the direct problems involved in CAD,
such as "find all the pixels along the straightline between these two points,
or find the pixel where these two lines cross", the general design problem
is frequently cast in domains far afield from geometry or spatial consideration,
such as "we need to increase the sense of luxury in this car design".
Straight forward "rule-systems", ala mycin, are great at what is best called
"diagnoi/// diagnosis". Design is quite a different kind of problem. Its
"constructive" rather than "analytric". The possible paradigm of generating
all possible solutions and evaluating them to find the best one is simply
to time-consuming to be a meaningful approach. So the "rul-e-checking"
approach is not satisfactory iether.
I think that there is some sort of methodology that may lay between based
on a strong version of generate a prototype - find out what's wrong with it,
make another, and do it over. A protyping scheme.
I am currently reading _ Creativity in Invention and Design_ by
Subrata Dasgupta, Cambridge Press. It is the strongest analysis of
computational design that I have seen to date and I suggest to anyone
interested in the subject. Maybe, we can discuss it here on the DESIGN-L
if enough people are interested.
- Ray Lauzzana
question
of "design" and "design automation" is the ill-defined, sometimes terribly
structured nature of what "design problems" are. The process of refining the
design problem statement frequently reduced the problem to a well-structured
problem. Indeed, somewhere along the line this must be done if a "solution"
is to be found. But, unlike the direct problems involved in CAD,
such as "find all the pixels along the straightline between these two points,
or find the pixel where these two lines cross", the general design problem
is frequently cast in domains far afield from geometry or spatial consideration,
such as "we need to increase the sense of luxury in this car design".
Straight forward "rule-systems", ala mycin, are great at what is best called
"diagnoi/// diagnosis". Design is quite a different kind of problem. Its
"constructive" rather than "analytric". The possible paradigm of generating
all possible solutions and evaluating them to find the best one is simply
to time-consuming to be a meaningful approach. So the "rul-e-checking"
approach is not satisfactory iether.
I think that there is some sort of methodology that may lay between based
on a strong version of generate a prototype - find out what's wrong with it,
make another, and do it over. A protyping scheme.
I am currently reading _ Creativity in Invention and Design_ by
Subrata Dasgupta, Cambridge Press. It is the strongest analysis of
computational design that I have seen to date and I suggest to anyone
interested in the subject. Maybe, we can discuss it here on the DESIGN-L
if enough people are interested.
- Ray Lauzzana