Critique: A Model for Functional Reasoning in Design
Freeman, Newell
*
It didn't really hit me how old this paper was until examples were
given that talked about drum memory. (This is the earliest
publication of AI in Design we've read for this course.)
This is a sign of how good a paper this really is. The concepts
discussed are still interesting and not many are directly addressed
by current research despite outdated examples. The authors do have
the advantage of just speculating about functional reasoning rather
than presenting a system that incorporates all the ideas mentioned.
Three classes of structure is listed, physical objects, abstract objects
and relations. I saw no good reason for differentiating between
abstract objects and relations. A definition of function might have
been useful. Given a definition it might be more obvious why
a specific representation of relation would be useful.
While discussing the basic model for the design task environment,
proposition P2: "For each function it provides, a structure requires
a set of functions." a hierarchical representation seems like an
obvious choice. I did not find much talk of this though. The paper
did address a different concern I had. The end of the "Structures
provide functions" section states "In general there is no limit on the
number of structures than can use a provided function." The "Capacity
laws" section assures this was not intended to mean at any one time.
the "Input-output relations" section's example seemed erroneous or at
least confusing. The number of users logged on does not depend on the
amount of memory available. Maybe the maximum number of users does.
The amount of cutting (for a knife) does not depend on the sharpness
of the blade. Maybe the quality of cutting, or the materials it can cut
does.
After the description of the model, using it for design is discussed.
Asymmetry between the desired functions and the provided functions,
there possibly being more provided since there is no requirement to
use these. While not required to use all provided functionality,
having a close balance might serve well as an efficiency goal.
When discussing the system being built based on the presented model
it is very unclear what is making the decisions and why in the system.
Why is one structure is chosen over another? Why is effort being
considered to automate the creation of an initial structure when the
process only takes a few minutes for a human designer?
*
P. Freeman & A. Newell,
A Model for Functional Reasoning in Design.
Proc. Int. Jnt. Conf. on AI, IJCAI-71, 1971, p. 621-640.
|