>
> Given the previous discussions on plan, as in plan of action, and plan, as in
> schema or pattern, I am a little confused.
> Are you using plan in both senses of the word here?
A plan can be discussed as a knowledge structure, or it can be
executed. I can tell you how I make breakfast, for example, or I can
execute my plan and actually get something to eat.
I define a plan as a backward slice through the code, tracing back
first through the data and then through the control flow. This defines a
tree structure, where each branch of the tree is a plan (and the whole
tree is a plan). This algorithm "pulls out" from the average program the
plans average, count, sum, and read loop.
When a plan is created, the creation order follows the plan
structure backward from the goal. The data and control flow dependencies
identified by slicing define a non-linear tree structure, and the actions
(lines of code) in this structure then have to be placed in a linear order
to be executed on a serial computer. After the procedural order has been
specified (and there are often many valid orders), the whole structure can
be described as a plan schema, where the actions (lines of code) are
connected by data flow and linear control flow links. This structure can
be described as a plan or as a plan schema; it's just one structure. I
usually refer to the structure of a plan, where the plan has both plan
(non-linear dependeny) structure and procedural (linear) structure. This
distinction can also be seen as the problem (plan) versus solution
(procedure) structure but of course a plan has both structures. The
execution structure is similar to the procedural structure with, of
course, only one branch of a selection taken and the actions in a loop
repeated.
There are many ways to create a plan, many types of
planning. Experts can retrieve a complete solution (plan / plan
schema) and show top-down design: code appears in procedural
order. Experts in a new domain appear similar to novices; they show a
mixture of backward plan creation and forward plan retrieval. Pure novices
show backward plan generation from the goal.
Opportunistic planning (Hayes-Roth) shows how a plan can be
followed and then abandoned to sieze an unforeseen
opportunity. Serendipitous design (Guindon) is even less
structured. Situated design (Suchman) says that planning is always local,
with global evaluation. Changes in strategy during design have been
documented by Visser and Rist, as well as by many others.
Plan generation is complex, but a principled account can be
given. Building a computational model to support this complexity is the
tricky bit ...
Rob Rist
- Automatic footer for [EMAIL PROTECTED] ----------------------------------
To unsubscribe from this list, mail [EMAIL PROTECTED] unsubscribe discuss
To join the announcements list, mail [EMAIL PROTECTED] subscribe announce
To receive a help file, mail [EMAIL PROTECTED] help
This list is archived at http://www.mail-archive.com/discuss%40ppig.org/
If you have any problems or questions, please mail [EMAIL PROTECTED]
