Information Technology Reference
In-Depth Information
ð
ð
M δ
nþ1
ð
Þ¼μ=
C θ, δ
nþ1
ð
ÞμðÞdθ < min
i¼1
...
n
C θ, δ
i
ð
ÞμðÞdθ
Θ
Θ
An even more robust solution would be independent of μ. It means that whatever
the belief on the states of nature—ie even for states of nature considered extremely
low-, the alternative δ
n+1
is the better. This can be written as:
ð
ð
8i ¼ 1
...
n, 8μ,
C θ, δ
nþ1
ð
ÞμðÞdθ <
C θ, δ
i
ð
ÞμðÞdθ
ð13:3Þ
Θ
Θ
Example: Raincoat Cap
Let's illustrate what it means on a simple example of decision making situation
(see Fig.
13.1
). Suppose that the decision maker wants to have a walk and his
decision space is D¼ {d
1
: take a cap to protect against the sun; d
1
, take a raincoat to
protect against the rain}; the states of natures are Θ¼ {θ
1
, sunny weather; θ
2
, rainy
weather}; and subjective probability are, for instance {μ(θ
1
) ¼0,51; μ(θ
2
) ¼1-
μ(θ
1
) ¼0,49}. The utility function is U N
2
!R, for instance : U(θ
1
; d
1
) ¼100;
U(θ
2
; d
2
) ¼10; U(θ
1
, d
2
) ¼100; U(θ
2
; d
1
) ¼10. This situation is usually
represented by a decision-hazard tree (see Fig.
13.1
). One computes the expected
utility associated to each decision. With these data, the decision-maker should
choose d
1
—and one also understands the fragility of this choice, due to the
proximity between μ(θ
1
) and μ(θ
2
). This remark usually leads to increase knowl-
edge (e.g. look at weather forecast even if it reduces the walk time).
If we add the hypothesis that the actor can design a new solution, then the dominat-
ing solution can be designed as d
3
such that U(θ
1
; d
3
) ¼100; U(θ
2
; d
3
) ¼100.
The design of d
3
might lead to a kind of “raincoat-cap”.
Note that industrial history is actually full of such design. The graph in Fig.
13.2
illustrates the fact
that some technologies are for
instance independent of
q
1
U = 100
µ
= 0,51
q
1
U = 100
µ
= 0,51
E
1
= 56
E
1
= 56
U = 10
µ
= 0,49
U = 10
µ
= 0,49
q
2
q
2
d
1
d
1
U = 10
µ
= 0,51
U = 10
µ
= 0,51
q
1
q
1
d
2
d
2
E
2
= 54
E
2
= 54
U = 100
µ
= 0,49
U = 100
µ
= 0,49
q
2
q
2
d
3
U = 100
µ
= 0,51
q
1
E
1
= 100
U = 100
µ
= 0,49
q
2
Fig. 13.1 Raincoat-cap example. (a) Selection of the solution with the best expected utility;
Search WWH ::
Custom Search