Information Technology Reference
In-Depth Information
If a value close to zero is chosen then no change could effectively take place; if 1
were chosen there could only ever be an exchange of one viewpoint for another and
no negotiation of belief would be possible. We opted for
0.25
as the 'experience'
unit because if a new 'experience' arrives to compete for the Actor's attention it has
only to discard
0.25
of its current happiness. If we continue in a figurative vein by
suggesting that each new experience be mixed with those added before, we have a
situation for the gaining of experience-knowledge as follows. Actor X gains (say)
four experiences, red experience, blue experience, orange experience and purple
experience. These experiences together modify X's view such that it only has to
discard
0.25
of the overall stock of experience to make way for the new experience.
Enough of all the previous experiences remain to count as useful learning. This is an
engineering type of decision rather than one that has any scientific or strictly logical
basis.
The number of discrete values along each dimension considered by an Actor can
vary. The confidence in a particular value is affected not only by experience and
queries but also by the number of values along a dimension available for considera-
tion. In our case we have only to consider seven along a single dimension. In order
to allow for a changing number of values we used the same idea that was originally
proposed by Addis and Gooding (
1999
) for belief (also see Chap. 6). Here we intro-
duced a dynamic threshold, the indifference value 'I'. In this case the value 'I' defines
those happiness values larger than which are to be actively considered as happy. 'I'
represents the general normal happiness (contentment) of an individual actor instead
of a group confidence as in the belief model.
To calculate the indifference value 'I' a function is needed that will change
smoothly between limiting values. It should be easily calculable from any number
of different values along any particular dimension, in this case happiness. A quantity
that varies in time in this way is the inverse of entropy. Entropy is an expected measure
of the log of a range of values. This can be used as a general measure of an Actor's
general happiness while listening to the music at time
n
. In the following equation
'a' and subscript 'a' is used to denote a particular Actor. This 'Entropy(Agent
a
),' is
called the general happiness measure for an actor 'a' in the Model Entropy, where
the term model denotes the set of values that makes up the Actor's view of a piece
of music. Model Entropy is given by:
Ha
E
n
(H
a
)
=−
∗
Entropy
n
(Agent
a
)
Log
2
(E
n
(H
a
))
From this equation we can obtain an inverse of the entropy which gives an expected
value for E
n
(H
a
). This will be denoted by I
n
(H
a
). I
n
(H
a
) will be called an Indifference
Threshold for the actor '
a
' at event '
n'
:
log
2
−
1
(Entropy
n
(Agent
a
))
Indifference Threshold(a, n)
=
=
I
n
(H
a
)
The expression E
n
(H
a
) is the expected probability of a dimension (happiness) value
for actor 'a' at conversation moment 'n'. The measure of a value E
n
(H
a
) above I
n
(H
a
)
