Information Technology Reference
In-Depth Information
Control is useful and can be a remedy in both cases, but with different roles. Control means
in fact (see Chapter 7 for a more detailed analysis):
a) The possibility of improving
Y
's performance, or of intervening in time - on the basis of
new evidence - for remedying, revoking the delegation, etc. Control is a remedy to low
expectations about
Y
.
b) The possibility of acquiring, through
monitoring,
additional evidence and information
run-time.
intervention) is also a remedy against uncertainty-based
lack-of-trust: '
In case something is bad I will know, and I will be able to do something
'. I
use monitoring in order to
update
and complete my beliefs, since I know that my prediction
of the future and my knowledge is incomplete. This is different from a low probability-based
lack-of-trust: '
In case some prediction is wrong I will know, and I will be able to do something
'.
I use monitoring in order to
revise
my beliefs that were wrong.
Of course, the two problems and the two functions of monitoring can co-occur.
In sum, control (monitoring
+
4.8.2 Conditional Trust
Searching or waiting for additional and determinant evidence also means having to formulate
a form of 'conditional' trust.
X
trusts
Y
but if and only if/after
Y
has performed a given action,
provided a given assurance, or proof
. 'Only if he swears on his sons'; 'Only if I can check',
'Only if has this documented experience',
and so on.
IF
(
Bel
X
(
Predicate Y
))
→
(
Trust
(
XY
))
(4.7)
Where,
Predicate
represents either an act or a feature of
Y
,ora
sign
of such a feature.
The difference between this 'conditional' trust and normal 'evidence-based' trust is just that
X
is waiting for such evidence; but in a sense she has already decided to trust
Y
, provided that
the expected evidence will be true.
4.8.3 To Give or Not to Give Credit
The second remark is that it is extremely important for there to be the possibility to mentally
ascribe that gap of ignorance (that part without evidence; what is 'possible' but not grounded)
infavororagainst
P
or
Not-P
.
In fact, given this model, the part that Dempster and Shafer call 'plausibility', the empty
part, is in fact ascribable to
P
or to
Not-P
.
P
is probable 55% but
possible
and plausible (not
against evidence) up to the 85%! While
Not-P
is probable 15% but plausible up to the 45%. In
our opinion, applied to the prediction of the behavior of
YX
counts on, this gap represents what
we call to 'give credit'. If
X
trusts
Y
beyond supported evidences and grounded probabilities,
and gives in favor of
Y
all the 'plausibility' space (till the limit of the opposite evidences),
then she is 'giving credit' to
Y
; she is trusting (believing) beyond evidence (see Figure 4.5 and
Section 4.2.1).
Search WWH ::
Custom Search