Information Technology Reference
In-Depth Information
We can thus define a concept
flexibility
which ranges from 0 to 1 for each agent such
that:
flexibility
=
1/N
Flexibility represents an agent's responsiveness to new observational evidence. Flex-
ibility is considered to approximate to a window of N events within which belief has
adapted to the current position. The larger the window the less responsive the agent
becomes to the current event.
6.4
The Impact of Opinion: Consultation
We assume that one agent consulting another is equivalent to accessing the consulted
agent's complete range of confidence values. This access has the effect of modifying
the consulting agent's confidence in each hypothesis as if it had performed its own
experiment. (Bayes' Rule is not applied in consultations; it is needed only to update
confidences based on experimental results). The confidence value of each of the
hypotheses, which make up the consulting agents belief profile, will be modified
according to the following equation:
E
n
−
1
(H
Consultee
)]
/
M
M ranges from 1 to infinity. The larger M, the smaller the effect any evidence has on
the change in confidence. We can thus define concept receptivity. This ranges from
0 to 1 for each agent as:
E
n
(H)
=
[(M
−
1)E
n
−
1
(H)
+
receptivity
=
1/M
Receptivity reflects the consulting agent's receptiveness to the beliefs of any
consultee.
6.5
A Simple Example of Confidence Adjustment
Given a coin that is to be tossed, we might consider two possible hypotheses:
•
H1
. The coin is good (e.g., has both a head and a tail)
•
H2
. The coin is double headed (or tailed)
Result
P (Result/H1)
P (Result/H2)
Heads
0.5
1.0
Tails
0.5
0.0
According to Peirce, knowing that one of these hypotheses is true “makes the world
a less surprising place”. Treating entropy as a form of a measure of surprise, as did
