Information Technology Reference
In-Depth Information
Example 4.20
We consider the scenario in Example 4.4. Agent
b
's
belief
bel
b
is more accurate than agent
a
's belief relation
bel
a
, because
in
bel
a
thereisabeliefofpreference(
,
movie)
which is not in agent
b
's belief
bel
b
, and this preference is false. Sim-
ilarly, agent
a
's belief
bel
a
is also more accurate than agent
b
's belief
bel
b
,becausein
bel
b
there is a false belief of preference
{
a, b, c
}
,
movie)
b
(
{
b, c
}
(
{
a, b, c
}
,
movie)
a
(
{
a, b
}
,
movie)
which is not in agent
a
's belief
bel
a
. Thisisanexamplethatshows
that the
more-accurate-than
relation is generally not transitive and not
asymmetric.
Given two arbitrary beliefs
bel
1
and
bel
2
, it is possible that we
cannot conclude that
bel
1
is more accurate than
bel
2
, but we cannot
conclude that
bel
2
is more accurate than
bel
1
, either. A trivial example
of such a case is when all and only those preferences that are correct
are in both
bel
1
and
bel
2
,orwhenboth
bel
1
and
bel
2
are empty. In
these examples, we are unable to find any preference that is in one
of
bel
1
and
bel
2
but not the other, whether it is correct or incorrect.
Hence, we cannot say that
bel
1
is more accurate than
bel
2
,or
bel
2
is
more accurate than
bel
1
. In general, the same arguments hold for all
cases when
bel
1
=
bel
2
.
Definition 4.9 (Accuracy Relation of Agents' Private Belief
Profiles)
We define the
is more accurate than
relation of agents'
private belief profiles as follows. Given two belief profiles
B
=
{bel
1
,bel
2
,...,bel
n
},
B
=
bel
1
,bel
2
,...,bel
n
}
{
,
we say
B is more accurate than B
if there exists
i
∈
N
such that
is more accurate than
bel
i
, and either
bel
j
=
bel
j
bel
i
or
bel
j
is more
accurate than
bel
j
for all
j
=
i
.
In short, given two collections of agent beliefs (either from the same
agent or from different agents), we say that one set of agent beliefs is
Search WWH ::
Custom Search