Here are some fairly miscellaneous thoughts about comparing alternative
approaches to Reasoning about Action. (Please forgive me for a few fairly
blatant attempts to advertise my own work at the same time.)
(1) When comparing and evaluating formalisms, we need to be careful
not to form too strong associations between particular methodologies
(e.g. deduction and entailment methods, default reasoning techniques)
and particular ontologies . I can think of a few occasions when this
has happened. For example: (i) At least until Murray Shanahan's
1995 paper
[8] .)
the ontology of the Event Calculus was irrevocably linked in many researchers'
minds with Logic Programming (and in particular with negation-as-failure),
and thus dismissed or attacked on "semantic" grounds. (ii) The methodology of
using "action languages" (i.e. the Language A, etc.
[1] .)
has become
overly associated with the ontology that the original Language A
inherited from
the Situation Calculus. (Tony Kakas's and my "Language E" paper
[2] .)
has shown that, for better or for worse, the methodology can actually be
applied in a wider context.) (iii) It seems to be a commonly held belief that
"narrative time" ontologies, such as that of the Event Calculus, demand that
planning be done using abduction (as opposed to deduction). (I've shown
that this is not the case in
[6] .)
(2) As a community, we should be encouraging work on comparing action
formalisms and ontologies, and we should be critical of papers which don't
contain adequate comparisons with other work (and especially with work
based on different ontologies). There is now a fair body of work exploring how
the Event Calculus and the Situation Calculus correspond, so there's really no
excuse for lack of comparisons in this case at least. (For formal results,
see for example
[3] ,
[5] ,
[7] ,
[9] , and
[2]
- the last shows a
correspondence between Languages A and E.)
(3) In
[2]
Tony Kakas and I wrote:
"We believe that the use of, and comparison between, different
ontolgies is vital in the study of reasoning about action. Central issues
such as the frame problem, the ramification problem and the qualification
problem all take on different flavours when set in different ontological
contexts. Comparisons between different approaches can help reveal
which aspects of these problems are fundamental, and which are merely the
product of a particular method of representation" .
I stand by this view. A good example of a (nevertheless interesting) problem
which is the product of a particular ontology (rather than being
fundamental) is the difficulty of distinguishing between observations and
causal rules in the Situation Calculus and in the Language A (i.e. in the
context of the Language A, the difficulty in distinguishing the roles of
value and effect propositions). Vladimir Lifschitz presented a technically
interesting solution to this difficulty in
[4]
but neither the problem
nor the solution translate to other (ontologically different) approaches.
We need to be careful to distinguish between this type of issue and more
fundamental problems such as dealing with ramifications or continuous
change.
(4) We need to keep the role of "action languages" (the Language A, etc.)
in perspective. To quote Vladimir Lifschitz
[4] :
"Originally, action languages were meant to play an auxiliary role. The
primary goal was to represent properties of actions in less specialised
formalisms, such as first-order logic and its nonmonotonic extensions, and
the idea was to present methods for doing that as translations from action
languages" .
Well, we shouldn't loose sight of that primary goal. There are many good
reasons for using a general purpose logic to represent properties of
actions. Perhaps the most important is that it allows us to link in with
work on other aspects of common sense reasoning (reasoning about space,
shape, beliefs, contexts, etc.).
REFERENCES
[1] Michael Gelfond and Vladimir Lifschitz,
Representing Actions in Extended Logic Programming ,
Proceedings of the Joint International Conference and Symposium on Logic
Programming, ed. Krzysztof Apt, MIT Press, page 560, 1992.
[2] Antonios Kakas and Rob Miller,
A Simple Declarative Language for Describing Narratives with Actions ,
to appear in the Journal of Logic Programming: Special Issue
on Reasoning about Action and Change (scheduled for publication by
Elsevier Science early 1997).
[Abstract, Postscript and DVI Versions]
[3] Robert A. Kowalski and Fariba Sadri,
The Situation Calculus and Event Calculus Compared ,
in Proceedings of the International Logic Programming Symposium (ILPS'94),
1994.
[4] Vladimir Lifschitz,
Two Components of an Action Language ,
in Working Papers of the 3rd Symposium on Logical Formalizations of
Commonsense Reasoning (Common Sense '96), Stanford University, 1996.
[5] Rob Miller,
Situation Calculus Specifications for Event Calculus Logic Programs ,
in Proceedings of the Third International Conference on
Logic Programming and Non-monotonic Reasoning, Lexington, KY, USA, pub.
Springer Verlag, 1995, pages 217-230.
[Abstract, Postscript and DVI Versions]
[6] Rob Miller,
Notes on Deductive and Abductive Planning in the Event
Calculus , July, 1996.
[Abstract, Postscript and DVI Versions] .
[7] Alessandro Provetti,
Hypothetical Reasoning about Actions: From Situation Calculus to
Event Calculus ,
Computational Intelligence, volume 12, number 2, 1995.
[8] Murray Shanahan,
A Circumscriptive Calculus of Events ,
Artificial Intelligence, vol. 77, pages 249-284,
Elsevier Science Publishers, 1995.
[9] Kristof Van Belleghem, Marc Denecker and Danny De Schreye,
On the Relation Between Situation Calculus and Event Calculus ,
to appear in the Journal of Logic Programming: Special Issue
on Reasoning about Action and Change (scheduled for publication by
Elsevier Science early 1997).