Comments on dialogue design

Pedro Szekely ([email protected])
Wed, 14 Dec 1994 09:14:05 PST

Hi,

I looked again at the two documents Noi sent me: the FAX with the phone model
and the table of when tasks start and are deemed complete, and the document
with the dialogue design.

First some comments on the "Status of Dialog Modeling Design".

I am not clear on the dimensions of specification. All things that you
mention in the two dimensions affect sequencing. I am not sure why
interruptable and interleavable are not separate dimensions.

I would structure the first dimension differently, to more closely match the
model that we discussed with Kieras:

1) Which tasks need to be done
- All -- corresponds to all the children of a method. More formally, this
means that all the goal for the children have to be satisfied. In some cases,
the goal of a child is satisfied without executing the task, which is ok.
- One -- corresponds to multiple methods for achieving a goal.
I prefer the method way of thinking about it, ie, a Goal (Acomplish Goal
operator) can have multiple methods of achieving a goal.

2) Sequencing of method tasks (all goals for children need to be accomplished)
- There are multiple sequencing possibilities, with increasing levels of
sophistication:
-- Sequence: do all in sequence.
-- Unordered: order is irrelevant (except that preconditions need to be
respected).
-- Partial orders (eg like a PERT chart: A before B, C before B).
-- Preconditions.

3) How many times can or should each operator be executed:
- 1 time (default, assumes that doing it one time accomplishes the goal, no
need to formally define the goal).
- 0 or more: Goal is formally defined in a testable predicate. Task needs to
be done until goal is met.
- more fancy repetitions like in your document.

4) Re-entrant (boolean): is is possible to make multiple instances of the same
task? By default, if a parent is re-entrant, the child is too. This is a
tricky issue.

5) Interleavable (boolean): can other tasks (that are not children) start
while this task is being performed?

6) Concurrency: I am not sure what this means, is this different from
interleavable (I don't remember what we decided in our meeting).

7) Interruptable (boolean): can the task be interrupted. What happens then?

Comments on the "FAX":

Actually I think many of the previous comments apply to the FAX too.

Pedro Szekely
USC/ISI, 4676 Admiralty Way, Marina del Rey, CA 90292
Phone: 310/822-1511, Fax: 310/823-6714
URL: http://www.isi.edu/isd/HUMANOID-HOME.html