Information Technology Reference
In-Depth Information
Then the scenario (
O
3
; D
) entails
(
) after [
(
)], hence
in
pt
insert
pt
also
)]. Moreover, the observa-
tion
:
in
(
pt
) after [
insert
(
pt
)
;
clear
] and, hence, the observation
:disq
(
ignite
) after [
insert
(
pt
)
;
clear
] are entailed. Mainly responsible
for this last conclusion is the causal relationship
inexecutable after [
(
ignite
insert
pt
:
in
(
x
) causes
:disq
(
ignite
) f
8y: :
in
(
y
)
^:
tank-empty
^:
low-battery
^:
engine-problem
Thus, in addition to the above, (
O
3
; D
) entails
after [
insert
(
pt
)
;
clear
;
ignite
]
runs
which shows how qualication gets revoked once the only reason for a
disqualication disappears.
3.5 Degrees of Abnormality
Our account of the Qualication Problem so far makes two simplifying as-
sumptions. First, all abnormalities are considered equally unlikely. Second,
complete knowledge as to all possible causes for an abnormal disqualica-
tion is implicitly assumed by equating the abnormality with these causes.
A natural extension of the current theory is to allow dierent degrees of
abnormality. In this way, it becomes possible to specify, for instance, that
having been running out of gas is more likely than a low battery, which in
turn is more plausible an explanation for being unable to start the car than
a potato being placed in the tail pipe. Dierent levels of likelihood are thus
representable without the necessity to provide precise probabilities.
The introduction of degrees of abnormality incidentally also provides a
ready means to resolves the problem associated with our equating abnormal
disqualications with their possible causes. In addition to all conceivable
reasons for an abnormality to occur, one fluent in the list of causes may
represent the
unknown
cause. Suppose a domain constraint
disq
(
a
)
W
i
F
i
is weakened to the eect that
disq
(
a
)
W
i
F
i
_
(
a
). Then it
is no longer violated in case
a
has been observed abnormally disqualied
while each known cause,
F
i
, has been proved false.
7
mysterious
Suppose further that
`abnormality' fluent
(
a
) has higher degree of improbability than
mysterious
7
Introducing the unknown cause allows to model, for example, the tendency of
some human beings to hit out at mechanical devices, e.g., dispensing machines,
when discovering their malfunctioning. The foregoing reasoning process there, if
any, attributes the encountered malfunction to some unobservable fluent which
is hoped to being manipulable by means of that very action.
