Information Technology Reference
In-Depth Information
that an appropriate notion of causality is necessary when assuming away ab-
normalities. In the latter framework, event happenings are minimized while
taking into account the possibility that their occurrence is being caused (or,
in other words,
motivated
, hence the name). This excluding unmotivated
events and our minimizing uncaused abnormal disqualications are somehow
complementary while being based on the same principles. However, some me-
thodical criticism and limitations apply to MAT: An unsatisfactory property
of the underlying preference criterion, i.e., motivation, is its depending on
the syntactical structure of the formulas representing causal knowledge. As a
consequence, logically equivalent formalizations may induce dierent prefer-
ence criteria, of which only one is the desired. Moreover, the formal concept
of motivation becomes rather complicated in case of disjunctive, i.e., inde-
terminate information, which entails diculties with assessing the range of
applicability of MAT in general. Finally, events can only be `motivated' by
past events, which does not allow for telling apart caused events that occur
concurrently with the triggering event. (In passing, we mention that this last
restriction also applies to a version of the Event Calculus (see below) if events
are allowed to trigger other events, as proposed in [98]). In comparison, in
[115] it is illustrated that ramication can also be successfully applied to the
problem of minimizing event occurrences.
Talking about the
Event Calculus
(introduced by [61]; for an overview
and thorough analysis see [100] and the monograph [99]), dealing with the
Qualication Problem there requires to respect a fundamental dierence be-
tween this approach and, say, both Situation Calculus and Fluent Calculus.
Namely, the latter are based on a branching time structure where dierent,
hypothetical action sequences may fork left and right of the actual time line,
if any. In contrast, the Event Calculus grounds on a linear time structure
thus representing solely the actual evolution of the world. The occurrence
of actions is specied by assertions of the form
Happens
(
A; T
) stating that
action
A
is performed at time
T
. A consequence of the linear time struc-
ture is that once an assertion of this form is made, it cannot be withdrawn.
This entails diculties with formally representing observations of abnormal
action disqualications if these are interpreted as the action being physically
impossible|which is the interpretation we pursued throughout our analysis
of the Qualication Problem. An alternative view seems better suited for the
Event Calculus or, for that matter, generally for action theories based on a
linear time structure. Instead of tightly coupling an action with its success,
such as starting the car with the eect that the engine is running, performing
an action may rst of all be taken as the mere intention to achieve a certain
eect. The actual achievement then depends on both regular conditions be-
ing met and the absence of abnormal, disqualifying circumstances. We do
not wish to provide a detailed formal account of the Qualication Problem
in the Event Calculus here, but let us mention that this is accomplishable by
introducing a predicate
Succeeds
(
a; t
) meant to indicate that action
a
hap-
pening at time
t
produces the intended eect. Abnormal conditions denying
