Information Technology Reference
In-Depth Information
(
runs
;
:
in
(
pt
)
;
:
heavy
(
pt
)
)
(
:
runs
;
:
in
(
pt
)
;
:
heavy
(
pt
)
)
ignite
Res
1
:
(
:
runs
;
in
(
pt
)
;
:
heavy
(
pt
)
)
insert(pt)
(
:
runs
;
:
in
(
pt
)
;
heavy
(
pt
)
)
(
runs
;
:
in
(
pt
)
;
heavy
(
pt
)
)
ignite
Res
2
:
Figure 3.2.
The scenario
O
=
f:
runs
^:
in
(
pt
) after []
g
suggests two models.
Our model preference criterion is not sophisticated enough to distinguish between
them, although the one determined by
Res
1
is much more plausible.
inexecutable after [
(
)]
ignite
insert
pt
for instance, is not entailed, contrary to our expectations.
It should be stressed that the reason for the second model being counter-
intuitive is not that in general potatoes being found in tail pipes is more likely
than these being too heavy to lift. Rather we have encountered the general
problem that global minimization of abnormalities does not allow to distin-
guish disqualications which can be explained from the perspective of causal-
ity. Successfully introducing a potato into the tail pipe produces an eect
which
causes
the fact that the second action, starting the engine, is unquali-
ed. That is to say, while an abnormal disqualication of
insert
(
pt
) comes
out of the blue in the unintended preferred model,
M
2
, an abnormal disqual-
ication of
ignite
, as claimed in model
M
1
, is easily explicable. One even
tends to not call the latter abnormal since being unable to start the engine
after having clogged the tail pipe is, after all, what one can reasonably expect.
Thus we see that the dynamics of state transition possibly gives rise to
a special kind of preference among abnormal disqualications|a preference
which is not entailed by observations nor by
a priori
dierences in likelihood.
Our account of the Qualication Problem therefore needs revision.
3.3 Causing Abnormal Action Disqualications
When assuming away abnormal disqualications of actions in order to tackle
the Qualication Problem, care has to be taken that any such assumption
is withheld in case specic information hints at the presence of abnormal
circumstances. We have seen that hints of this kind may be revealed by the
dynamics of state transition, a fact which is ignored if the preference criterion
