Information Technology Reference
In-Depth Information
experimental subject in this scenario. We organise the model so that the actors are in
the same groupings as the human subjects.
In the model the actors enter into a conversation with the other actors in the group
with a limited questioning repertoire, that is only being able to ask if they like a
particular piece more than another piece, and which they liked least or most. Ordinal
relationships such as more or less are one of Wittgenstein's fundamental objects as
defined in his Tractatus; that is a primitive object that cannot be defined in terms of
any other primitive objects. This would suggest that denotational semantics is being
used here despite the lack of direct observation.
The choice of to whom the question is directed, what music is compared and what
question is framed, is left to the model of the actor's choice. The choice is based upon
the degree of uncertainty of another actor's view and the purpose of the conversation.
This purpose is to reduce the actor's degree of uncertainty about its view of all the
other actors in the group. The mechanism uses game theory as described in Chap. 6
(also Addis and Gooding
1999
and
2008
). In the model the other actors don't have
access to individual conversations, i.e. they are not listening in; they only know about
the answers to their own questions.
Each actor has their own separate sub-model for each of the other actor's internal
views and these sub-models are modified according to the answers they receive in
response to their questions. Based on these four sub-models (one is the actor's own
scale and the other three are scales from its belief model of the other actors), we can
calculate the distance measurement for each actor as perceived by any other actor
through their sub-model. In comparing the actual mentor choice order with the model
distance, we should be able to get similar results to those above (Fig.
14.3
)ifthe
actor's belief model does represent other's internal view to some degree.
14.6
Model Data Input and Results Analysis
Seven sets of actual experimental data are used as input. For each set, the model is
run 300 times to get the actors' expected belief model result. The Fig.
14.4
shows
how Actor 1's view of Actor 2 changes through the conversation.
It can be seen that as the model runs and actor 1 asks questions of actor 2, actor
1 homes in on actor 2's order of evaluation but not exactly how much it 'likes'
each piece. The realisation of sequence occurs fairly fast, by less than 30 runs, and
thereafter actor 1's belief only strengthens as to sequence and broadens as to the
absolute scaling he believes to be the view of actor 2. (See Fig.
14.3
where the
bracketed numbers in the key are the actual values for actor 2). Note that actor 1
does not always ask actor 2 at each cycle and in this case only asks actor 2 about
ten question before deciding its order. The run of questions (see Fig.
14.4
, Music 2
between 50 and 100 runs) is caused, in part, by the relative uncertainty actor 1 has
about actor 2 compared with the other actors in the group.
Actor 1's belief model of actor 2 is given in Table
14.1
. This includes 4 ranges of
belief (series 1-4) for each of the four music objects. The belief concerns the scales
